AI Science Tracker
EN

Affiliation not stated in the reviewed material

827 discoveries from 2020 to 2026

Discoveries

Generative discovery of chemical reaction networks

Source check: Source checked. Not independent scientific confirmation.

Finding: Version 2 reports NMR characterization of four isolated products, including three predicted diarylation structures; v1 contains earlier family-level MS evidence. The isolated products are supported by 1H/13C NMR; this validates structures, not every proposed reaction path, rate or barrier. The isolated-product result first appears in version 2 (2026-11-18); version 1 reported earlier family-level mass-spectrometry evidence.

Technology: ReactNexus flow-matching transition-state sampler, learned trajectory simulator, and Carina reactive machine-learning interatomic potential (EquiformerV3)

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
ReactNexus uses a flow-matching transition-state sampler, learned trajectory simulator and Carina reactive MLIP (EquiformerV3) to propose and evaluate reaction paths.
Test and result
Version 2 reports NMR characterization of four isolated products, including three predicted diarylation structures; v1 contains earlier family-level MS evidence.
Limitation / still unknown
The isolated products are supported by 1H/13C NMR; this validates structures, not every proposed reaction path, rate or barrier.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/38236v2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Construction and Validation of a Machine Learning-Based Risk Prediction Model for Recurrent Upper Gastrointestinal Bleeding in Patients with Liver Cirrhosis: A Prospective Study

Source check: Source checked. Not independent scientific confirmation.

Finding: A machine-learning risk model for recurrent upper-GI bleeding in cirrhosis was assessed in a prospective development cohort and an independent external cohort. The selected logistic model achieved external AUC 0.827 (95% CI 0.747–0.907). This supports prediction in the studied cohorts; it does not establish improved outcomes or causal effects.

Technology: Four machine-learning classifiers; logistic regression was selected

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Classifiers estimated individual recurrence risk from patient predictors.
Test and result
A prospective 505-patient development cohort and an independent 148-patient external cohort were used to predict six-month recurrent upper-GI bleeding in cirrhosis. The selected logistic model had external AUC 0.827 (95% CI 0.747–0.907), with decision-curve analysis.
Limitation / still unknown
No impact trial or evidence of improved patient outcomes; findings are bounded to the studied cohorts and setting.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-11469264/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

How far can routine process records predict final turbidity in industrial beer crossflow filtration? A capability-boundary study on 5000 production batches

Source check: Source checked. Not independent scientific confirmation.

Finding: Across 5,000 brewery batches, models using routine process records explained at most about 8% of held-out turbidity variance; no nonlinear model beat linear baselines and a learning curve plateaued near R²=.10. This is a specific industrial observability/capability result, bounded to the recorded features and production setting; it is not a causal intervention result.

Technology: Regressors test how routine brewery records predict measured final turbidity

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Regressors test how routine brewery records predict measured final turbidity
Test and result
Regressors test how routine brewery records predict measured final turbidity; the learned models establish a bounded process-data prediction ceiling.; PDF Results §§3.1–3.3: best linear models yield R²≈.084; multi-seed and learning-curve analyses plateau around .10, showing limited information in recorded features.; Holdout/multiseed and fixed-block learning-curve analysis directly supports the capability-boundary claim; batch data from one production setting and noninterventional design limit t
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-11487956/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Comparative Supervised Machine Learning for Predicting Cardiovascular Disease among Indian Adults 45 Years or Older: Evidence from the LASI Wave 1

Source check: Source checked. Not independent scientific confirmation.

Finding: In LASI Wave 1, a held-out sample of 19,335 adults supported retrospective discrimination of current, survey-defined CVD status (Random Forest AUC 0.728, Brier 0.177). This is a cohort-specific retrospective clinical classification result; it does not establish incident-risk prediction, causation, or external validity. The result is reported in the Research Square preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Supervised classifiers learn from LASI features to classify existing CVD status

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Supervised classifiers learn from LASI features to classify existing CVD status
Test and result
In LASI Wave 1, a held-out sample of 19,335 adults supported retrospective discrimination of current, survey-defined CVD status (Random Forest AUC 0.728, Brier 0.177). This is a cohort-specific retrospective clinical classification result; it does not establish incident-risk prediction, causation, or external validity.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-11390178/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Decoding Complications: Large Language Model-Based Classification of Free-Text Entries in Clinical Stroke Registries

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary source Methods/Results confirm the represented result: An LLM-guided taxonomy classifies 9,146 free-text stroke-registry adverse-event entries, with frozen GPT-5.4 labels matching clinician consensus on 300 held-out entries (weighted F1 96.5%) and recovering 19.6% more adverse-event instances than structured fields.

Technology: GPT-5.4 with an LLM-guided taxonomy

Family GPT-5: profile · models page

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
GPT-5.4 with an LLM-guided taxonomy is used in the reported analysis or evaluation.
Limitation / still unknown
The source reports one registry and a 300-entry clinician-consensus holdout; this does not establish performance at other registries or patient benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s10916-026-02486-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

FADS1 serves as a candidate shared biomarker for metabolic dysfunction-associated steatohepatitis related early-onset colorectal cancer identified via machine learning and in vitro validation

Source check: Source checked. Not independent scientific confirmation.

Finding: ML nominates FADS1 as a candidate shared biomarker in MASH-associated early-onset colorectal cancer, with public-cohort analyses and limited laboratory assays. Authors describe associations and acknowledge no direct genetic perturbation; the result is exploratory.

Technology: glmBoost and XGBoost selected from 44 evaluated machine-learning configurations

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
PDF Methods/Results: 44 algorithm configurations select glmBoost+XGBoost; external cohorts, single-cell virtual knockout, cell-line qPCR/WB, and small IHC sets (4 EO-CRC, 4 LO-CRC). Authors note no direct perturbation and association-only mechanism.
Test and result
PDF Methods/Results: 44 algorithm configurations select glmBoost+XGBoost; external cohorts, single-cell virtual knockout, cell-line qPCR/WB, and small IHC sets (4 EO-CRC, 4 LO-CRC). Authors note no direct perturbation and association-only mechanism.
Limitation / still unknown
PDF Methods/Results: 44 algorithm configurations select glmBoost+XGBoost; external cohorts, single-cell virtual knockout, cell-line qPCR/WB, and small IHC sets (4 EO-CRC, 4 LO-CRC). Authors note no direct perturbation and association-only mechanism.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s12672-026-06235-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Scaling laws in machine learning on non-ideal experimental data enable inverse design of radiative-cooling coatings

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine learning on 11,520 experimentally fabricated radiative-cooling coatings showed power-law decreases in prediction error as training data increased. Experimentally tested inverse designs matched a broadband-emitter target more closely with larger datasets, but did not outperform the best original sample.

Technology: Forward and inverse neural networks; spectrum-conditioned diffusion model

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Learned forward/inverse models and a conditional diffusion model mapped measured coating spectra and parameters for prediction and inverse design.
Test and result
The automated platform produced and optically characterized 11,520 coatings across nine material/process parameters. Models showed power-law error reduction with more training data; selected inverse designs were experimentally tested for matching a broadband-emitter target.
Limitation / still unknown
The selected designs did not outperform the best original sample; claims are limited to explored compositions, processing and measurement setup.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.19596v1 · https://export.arxiv.org/api/query?id_list=2611.19596&max_results=1 · https://arxiv.org/pdf/2611.19596v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 4/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Do student perceptions mirror reality? A mixed-methods study to evaluate the accuracy, utility and efficiency of AI-assisted feedback in introductory mathematics courses

Source check: Source checked. Not independent scientific confirmation.

Finding: Two introductory mathematics courses were studied using Gemini 2.5 Pro to draft scores/feedback from submissions and course references. Study One compared human-only (159 students) with human-curated AI (89); Study Two compared AI, peer and professor feedback. Grades were broadly similar, but students rated professor feedback most useful and read AI feedback less often (57% vs 89%/94%). This is a specific educational study, bounded to two courses and short-term feedback rather than learning outcomes.

Technology: Gemini 2.5 Pro produces structured formative scores and text from student work plus assignment/reference materials

Family Gemini: profile · models page

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Gemini 2.5 Pro produces structured formative scores and text from student work plus assignment/reference materials
Test and result
Gemini 2.5 Pro produces structured formative scores and text from student work plus assignment/reference materials; its performance and educational use are evaluated.; PDF Results pp.20–38: graders’ score agreement and student/grader perceptions across the two course studies; discussion reports broadly similar grades, mixed trust/utility and lower reported reading of AI feedback.; Two course settings, grader comparisons, surveys and qualitative notes provide empirical education evidence; one uni
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-11455953/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AI-Assisted Failure Prediction on Postoperative X-Rays: A Head-to-Head Comparison with Consultant Trauma Surgeons

Source check: Source checked. Not independent scientific confirmation.

Finding: A prospectively assembled 466-patient application cohort (104 fixation failures) was classified from postoperative radiographs by a ResNet-50 trained on a disjoint 200-patient cohort. AI sensitivity was 96.15% and specificity 98.90%, exceeding the three surgeons numerically; the strongest surgeon difference was not significant. This supports an adjunct diagnostic result, pending external prospective validation. This supports an adjunct diagnostic result, not external prospective validation.

Technology: ResNet-50 image classifier

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ResNet-50 reads postoperative X-rays and classifies fixation failure against clinical outcome; AI supplies the diagnostic prediction.
Test and result
PDF Results pp.6–8: 466 patients, 104 failures; AI classified 458/466 (98.28% accuracy); paired surgeon analysis and comparison with strongest surgeon reported.
Limitation / still unknown
Same-cohort paired human comparison and a separately trained cohort support diagnostic performance; same centers and authors’ stated need for external prospective validation limit transportability.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-11408305/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Chronic Post-Surgical Pain After Anal Fistula Surgery Using Perioperative Clinical and Anesthesia Data: A Machine Learning Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Perioperative-data models predicted chronic postsurgical pain after anal fistula surgery, using a held-out test set with 19 outcome cases; performance is therefore preliminary. The result is reported in the Journal of Pain Research. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Methods define pain outcome and split

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods define pain outcome and split
Test and result
Perioperative-data models predicted chronic postsurgical pain after anal fistula surgery, using a held-out test set with 19 outcome cases; performance is therefore preliminary.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13771633/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Tracing truth: How traceability enhances fake news detection design

Source check: Source checked. Not independent scientific confirmation.

Finding: In a controlled human study, traceability-enhanced fake-news interface design improved detection over the XAI-only condition; other contrasts were not significant. This is a narrow interface result in the tested news/desktop context.

Technology: XGBoost article classifier with LIME explanations and source traceability

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Methods/Results report XGBoost article labels, LIME/provenance interface conditions, and 299 retained participants. Combined design exceeded XAI-only by 7.16 percentage points (95% CI 0.91–13.41); other pairwise contrasts were nonsignificant.
Test and result
Methods/Results report XGBoost article labels, LIME/provenance interface conditions, and 299 retained participants. Combined design exceeded XAI-only by 7.16 percentage points (95% CI 0.91–13.41); other pairwise contrasts were nonsignificant.
Limitation / still unknown
Methods/Results report XGBoost article labels, LIME/provenance interface conditions, and 299 retained participants. Combined design exceeded XAI-only by 7.16 percentage points (95% CI 0.91–13.41); other pairwise contrasts were nonsignificant.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s12525-026-00938-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A machine learning-based model for predicting 5-year conditional survival in patients with colon cancer: a development and validation study

Source check: Source checked. Not independent scientific confirmation.

Finding: A model predicts five-year conditional survival in a 430-patient colon-cancer cohort and is evaluated on an independent validation set. The study is retrospective and single-center. Author affiliations were not available in the reviewed publisher passages; organization left empty.

Technology: Six machine-learning methods and logistic regression

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Six machine-learning methods and logistic regression
Test and result
A model predicts five-year conditional survival in a 430-patient colon-cancer cohort and is evaluated on an independent validation set. The study is retrospective and single-center.
Limitation / still unknown
Generalization beyond the study center and prospective clinical utility are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2026.1962950

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Agent-Guided Discovery of Exchange-Correlation Functionals with Physics-based Validation

Source check: Source checked. Not independent scientific confirmation.

Finding: Full arXiv Methods/Results describe LLM-agent-guided symbolic search of exchange-correlation functionals with self-consistent physics filters and held-out/external validation; the selected O159 functional retains an 11.7% held-out error improvement over ωB97M-V and comparable external GSCDB137 performance. The result is a computational chemistry functional candidate, not experimental validation.

Technology: LLM-agent-guided symbolic search of density-functional exchange-correlation forms.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Frontier LLM agents propose exchange-correlation functional forms evaluated through a GPU fitting and self-consistent physics workflow.
Test and result
Agent-derived O159 was evaluated self-consistently on held-out MGCDB84 and on 36 GSCDB137 datasets; it improved held-out error versus ωB97M-V, while remaining computational rather than experimental validation.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.13329

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

When Answer Decodability Is Not a Decision: Separating Final-Answer and Imminent-Emission Signals in Chain-of-Thought

Source check: Source checked. Not independent scientific confirmation.

Finding: Controlled probing found that, for the tested tasks/models, a strong readout of the eventual answer can instead track the candidate about to be emitted; the authors include counterexamples and explicit limits. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Methods define dual-target probes and controls

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods define dual-target probes and controls
Test and result
Controlled probing found that, for the tested tasks/models, a strong readout of the eventual answer can instead track the candidate about to be emitted; the authors include counterexamples and explicit limits.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.11892v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Biparametric MRI with deep learning for the preoperative prediction of muscle invasion in patients with bladder cancer: a multicenter study

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Three-institution pathology-referenced cohorts show biparametric MRI deep learning predicts bladder muscle invasion; retrospective cohorts and manual segmentation limit generalization.

Technology: Deep-learning analysis of biparametric MRI

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep-learning analysis of biparametric MRI is used in the reported analysis or evaluation.
Limitation / still unknown
The reported cohorts are retrospective and use manual segmentation; prospective performance and broader generalization are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13748144/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Interpretable Machine Learning Accelerated Discovery of Doped Spinel Co3O4 Catalysts for the Oxygen Evolution Reaction

Source check: Source checked. Not independent scientific confirmation.

Finding: Models screened 252 sites; experiments tested Al- and Fe-doped Co3O4 and reported oxygen-evolution activity under near-neutral conditions. Other dopants remain computational predictions. Author affiliations were not available in the reviewed article text; organization left empty.

Technology: Machine-learning adsorption-energy and d-band models

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Machine-learning adsorption-energy and d-band models
Test and result
Models screened 252 sites; experiments tested Al- and Fe-doped Co3O4 and reported oxygen-evolution activity under near-neutral conditions. Other dopants remain computational predictions.
Limitation / still unknown
Evidence is limited to the reported catalyst compositions and experimental conditions.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pubs.acs.org/doi/10.1021/acscatal.6c06523

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

LAION-Audio-XL: A Web-Scale Corpus and Scaling Studies for Contrastive Language-Audio Pretraining

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Contrastive audio-text encoders were trained at five sizes on 365M samples; MAEB-30 error fell about 4% per compute doubling, with compute-optimal model sizes depending on sample budget. Bounded to this dataset/architecture and distinguishes scaling finding from dataset/model release.

Technology: Contrastive language-audio encoders (LAION-Audio-XL)

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The cited full-text method identifies the model operation and how it is linked to the represented field result.
Limitation / still unknown
Validation/comparison basis and its limits are stated in the cited source passage.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.11592v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Decomposing Earth Embeddings with Sparse Autoencoders

Source check: Source checked. Not independent scientific confirmation.

Finding: Sparse autoencoders decomposed AlphaEarth and Tessera Earth-observation embeddings into sparse features. At matched sparsity they reconstructed the embeddings better than PCA, ICA, NMF and other baselines. Some feature-specificity measures also appeared with a frozen random decoder, so the result supports learned reconstruction quality rather than new geographic discoveries.

Technology: Sparse autoencoders (SAEs) applied to AlphaEarth and Tessera embeddings

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Sparse autoencoders transform dense Earth-observation embeddings into sparse feature codes; evaluation compared reconstruction, concept alignment and spatial scale against PCA, ICA, NMF and frozen random-decoder baselines.
Test and result
At matched sparsity, trained SAEs reconstructed AlphaEarth and Tessera embeddings better than the tested baselines. Sparse features aligned with fewer reference concept groups and carried more variance at fine spatial scales; a frozen random decoder shared those latter properties.
Limitation / still unknown
This is a representation evaluation; application demonstrations are not yet reliable enough for practical use, and no newly established geographic classes are claimed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.11398v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Agent Self-Improvement with Frozen Weights: The Added Value and Limits of Failure Traces

Source check: Source checked. Not independent scientific confirmation.

Finding: Frozen Qwen3.5-9B harness study: 2,997 sealed synthetic retail decisions; trace-informed notes beat matched no-evidence search by 4.0 points, while official full conversations show no demonstrated improvement.

Technology: The cited full-text method identifies the model operation and how it is linked to the represented field result.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The cited full-text method identifies the model operation and how it is linked to the represented field result.
Test and result
The cited full-text method identifies the model operation and how it is linked to the represented field result.; Frozen Qwen3.5-9B harness study: 2,997 sealed synthetic retail decisions; trace-informed notes beat matched no-evidence search by 4.0 points, while official full conversations show no demonstrated improvement.; Validation/comparison basis and its limits are stated in the cited source passage.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.11342v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

When Credit Turns Against Tools: Transition-Aware Credit Assignment in Agentic Reinforcement Learning

Source check: Source checked. Not independent scientific confirmation.

Finding: Derivation and training/evaluation on Qwen models and benchmark tasks support a specific observation that terminal-reward normalization can give negative aggregate credit to tool tokens; TACA performance is separate from this represented mechanism. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: The paper analyzes multi-turn language-model reinforcement learning and uses Transition-Aware Credit Assignment (TACA), with Advantage Projection, Decision Contrast and supervised replay, to address negative credit assigned to tool-call tokens.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The paper analyzes multi-turn language-model reinforcement learning and uses Transition-Aware Credit Assignment (TACA), with Advantage Projection, Decision Contrast and supervised replay, to address negative credit assigned to tool-call tokens.
Test and result
Derivation and training/evaluation on Qwen models and benchmark tasks support a specific observation that terminal-reward normalization can give negative aggregate credit to tool tokens; TACA performance is separate from this represented mechanism.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.11330v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Causal Outcome Learning with Noisy Proxies

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: In the STEP-BD lithium cohort, the learned adverse-effect composite shows a quantified adjusted contrast and heterogeneity.

Technology: SCOUT causal outcome learning with noisy proxies

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
SCOUT learns a treatment-residualized composite outcome from noisy proxies before ATE/CATE estimation.
Limitation / still unknown
Synthetic evaluations and the 481-person STEP-BD analysis provide bounded validation; the study does not establish causal individual benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.10955v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

They Can See It but Not Say It: Iterative Self Knowledge Re-expression in Visual Reasoning Relative Pose Identification

Source check: Source checked. Not independent scientific confirmation.

Finding: SKR-ID identifies stable VLM output biases in relative-pose reasoning and improves Macro-F1 across three open VLMs without annotations.

Technology: Self-knowledge re-expression and logit debiasing for vision-language models

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
§§5–6 use self-knowledge re-expression, logit debiasing and unannotated image pairs.
Test and result
§7 and Tables 2–3 report significant gains from 35.0–45.4% to 53.7–58.0% across 7-Scenes/ScanNet and ablations.
Limitation / still unknown
Three models, multiple scenes and three-run statistical tests support the bounded visual-reasoning result.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.10483v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Semantic Synthesis over Brute-Force Search: LLM-Guided Transformation Design for Efficient Learning

Source check: Source checked. Not independent scientific confirmation.

Finding: LLM-generated image transformations performed well in the reported CIFAR-10, CIFAR-100 and SVHN experiments, using fewer candidate evaluations than brute-force search. One-epoch selection scores predicted 20-epoch outcomes in some dataset/architecture pairs but were weak or uninformative in others.

Technology: Large language models generating executable image-transformation programs

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LLMs synthesized transformation code; classifiers evaluated candidate transformations under short and longer training budgets.
Test and result
Across image-classification datasets, generated transformations were evaluated with one-epoch selection and retrained for 20 epochs; outcomes varied across dataset/architecture pairs, with proxy-score correlations ranging from 0.756 to near zero.
Limitation / still unknown
Short-budget selection is not a reliable proxy in every dataset/architecture pair; results are limited to the reported datasets and training setup.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.10940v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Detecting and Explaining Robot Policy Failures from Sparse Internal Features

Source check: Source checked. Not independent scientific confirmation.

Finding: MIND uses sparse-autoencoder features of visuomotor policies to detect and describe failures. The authors report evaluations in simulation and on a physical robot; results are bounded to tested executions and do not establish general deployment reliability.

Technology: MIND sparse-autoencoder feature analysis on visuomotor policies

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
MIND maps policy activations to sparse features associated with robot state and surroundings, then uses feature deviations to detect and describe failures.
Test and result
The study evaluated sparse-autoencoder features for policy-failure detection and diagnosis in simulation and on a physical robot, including held-out executions and ablations.
Limitation / still unknown
Physical-robot and simulation results are limited to the tested tasks and setup; general deployment performance is unknown.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.10400v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Approximate Recall, Approximate Forecasts: Recall as a Diagnostic for LLM Forecasting Errors

Source check: Source checked. Not independent scientific confirmation.

Finding: Across thousands of model-market rows, higher local recall fidelity is associated with lower Brier loss, with wrong recall associated with higher loss.

Technology: Four frozen language models produce event probabilities and historical recall answers

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Four frozen language models produce event probabilities and historical recall answers
Test and result
Four frozen language models produce event probabilities and historical recall answers; recall fidelity is linked to forecast loss.; Across thousands of model-market rows, higher local recall fidelity is associated with lower Brier loss, with wrong recall associated with higher loss.; Fixed-effects analyses, resolved outcomes, and recall-error decompositions validate the measured association; it is not prospective market performance.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.10891v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Stand-In Effects and Pooling Amplification: Lessons from Biological World Models for Testing Action Effects

Source check: Source checked. Not independent scientific confirmation.

Finding: In v2, frozen diffusion-model predictions exposed pooling tests that missed known actions in 96/96 segments while the registered alternative detected them.

Technology: Frozen pretrained DIAMOND models supplied frame predictions for the statistical tests

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Frozen pretrained DIAMOND models supplied frame predictions for the statistical tests
Test and result
Frozen pretrained DIAMOND models supplied frame predictions for the statistical tests; their predictions were not treated as measured biological outcomes.; Version 2 reports the pooled statistic detecting 0/96 action effects versus 96/96 for the registered alternative.; The comparison also checked 192 constructed null units; the result is bounded to tested one-step simulated game segments.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.01666v2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Bayesian Wombling for Tracking Invasive Species

Source check: Source checked. Not independent scientific confirmation.

Finding: North American invasive-species applications report modeled occurrence gradients and curvature, including supported and null intervals. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: A hierarchical Gaussian process models occurrence fields and differentiates them to estimate species-specific boundaries.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A hierarchical Gaussian process models occurrence fields and differentiates them to estimate species-specific boundaries.
Test and result
North American invasive-species applications report modeled occurrence gradients and curvature, including supported and null intervals.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.10871v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Locating and Repairing Domain Shift in VLM Trajectory Planning

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Activation patching showed that image and text tokens have distinct domain-shift patterns and that layer-wise patterns were stable across two driving datasets. The proposed MoLaRx selective-repair method had the lowest cross-city trajectory-error gap in six tested settings across three VLM planners; conclusions are limited to those models and benchmarks.

Technology: Vision-language models with layer-local image and text sensitivity

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Activation patching estimates modality-by-layer domain shift in three driving VLMs; MoLaRx applies matched selective repair to the affected layers.
Limitation / still unknown
The study evaluates the reported models and datasets only; broader generalization and real-world deployment outcomes were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.10720v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Aided Computational Discovery of Boride Substrates for Borophene Synthesis

Source check: Source checked. Not independent scientific confirmation.

Finding: DFT and machine-learning screening covered 1,356 surfaces, identifying candidate boride substrates for two borophene phases. Most candidates were model-predicted; synthesis was not demonstrated. Author affiliations were not available in the reviewed arXiv version text; organization left empty.

Technology: Machine-learning screening with density-functional-theory calculations

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning screening with density-functional-theory calculations
Test and result
DFT and machine-learning screening covered 1,356 surfaces, identifying candidate boride substrates for two borophene phases. Most candidates were model-predicted; synthesis was not demonstrated.
Limitation / still unknown
Candidate substrate performance and synthesis remain experimentally untested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.00533

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Discovering formally verified quantum programs with AI

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Version 1 Methods/Results confirm the case-specific result: Exact new checked fixed-point quantum program, plus separately checked parallel oracle-search family as a proposed child. Resource claims are conditional on formal comparison metrics and original version, without promising realization on hardware.

Technology: AI system for discovering quantum programs, with formal verification

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Exact new checked fixed-point quantum program, plus separately checked parallel oracle-search family as a proposed child. Resource claims are conditional on formal comparison metrics and original version, without promising realization on hardware.
Limitation / still unknown
The source describes the represented test, comparison, proof or study scope; conclusions are limited to that design.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.09822v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Glyph: An Interpretability Engine for AI-Driven Scientific Discovery from Clinical Data

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Version 1 Methods/Results confirm the case-specific result: Specific AI-extracted imaging marker linked to later clinical coding, with radiologist interpretation after discovery and conditional cross-cohort evidence. The record does not conflate this feature with overall AUC improvement from all Glyph concepts.

Technology: Glyph clinical-data interpretability engine

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Specific AI-extracted imaging marker linked to later clinical coding, with radiologist interpretation after discovery and conditional cross-cohort evidence. The record does not conflate this feature with overall AUC improvement from all Glyph concepts.
Limitation / still unknown
The source describes the represented test, comparison, proof or study scope; conclusions are limited to that design.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.08922v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Intent-Aware Caching for Efficient LLM Serving

Source check: Source checked. Not independent scientific confirmation.

Finding: An intent-aware eviction policy for LLM prefix caches was evaluated on WildChat and ShareChat workloads. In the reported setup it achieved a 34.6% cache hit rate and reduced average serving latency by up to 46.1%; the workload patterns are observational and do not establish performance on other traffic.

Technology: Intent classifier and intent-aware cache eviction policy

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
An intent classifier labels requests; a learned eviction policy uses intent-conditioned reuse statistics to guide prefix-cache decisions.
Test and result
Analyses of WildChat and ShareChat traces found continuation probability and inter-turn delays differed by intent. Intent-aware caching reached 34.6% hit rate versus 18.8% for a prior learned approach and 12.2% for LRU, with latency reduction up to 46.1% in the reported evaluation.
Limitation / still unknown
Trace patterns are observational and depend on the learned intent taxonomy; effects outside the evaluated workloads and setups are unknown.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.09715v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

PRISM: Primitive Routing via In-context Skill Mixing for Retentive VLA Adaptation

Source check: Source checked. Not independent scientific confirmation.

Finding: The authors found that low-rank updates trained for individual skills occupied a distinguishable subspace and that a composite-task update concentrated roughly half its energy there, unlike a matched random-rotation control. PRISM then stored skill adapters as reusable primitives and routed among them. Benchmark and robot task results are separate from this representation result.

Technology: Vision-language model-conditioned router with frozen low-rank skill adapters

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The authors measured whether low-rank updates trained for individual manipulation skills occupy a shared parameter subspace, then proposed PRISM to preserve skill adapters as primitives and combine them with a vision-language-model-conditioned router.
Test and result
A composite-task update concentrated roughly half its Frobenius energy in the per-skill subspace, versus near-zero for a matched random-rotation control. RoboCasa365 and an AgileX PIPER setup were used for task evaluation.
Limitation / still unknown
The representation result is limited to the reported model and task setup; broader generalization is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.08917v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Graph representations of dislocation networks

Source check: Source checked. Not independent scientific confirmation.

Finding: Graph representations of simulated dislocation networks show that the joint graph description and strain-rate prediction remains robust across the tested geometry encodings; the study does not establish an independently superior descriptor.

Technology: Specific simulation-based finding that the joint graph-description and strain-rate prediction is robust to the tested geometry encodings

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Graph representations and learned prediction model encode dislocation-network geometry and predict strain-rate behavior.
Test and result
Simulation-based graph encodings and strain-rate prediction evaluated across the geometries reported in the study.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.09534v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Solver-Admitted Metamorphic Testing of Learned Network Optimization Surrogates

Source check: Source checked. Not independent scientific confirmation.

Finding: Metamorphic tests of learned network-optimization surrogates find prediction changes under equivalent-problem representation and size transformations; random-weight controls indicate the effect is not solely a training artifact.

Technology: Specific measured dependence of equivalent-problem predictions on representation and problem size, independent of merely ranking model quality. The separate random-weight controls show the failure is not attributable solely to training.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Learned network-optimization surrogate models are tested for representation and size invariance using metamorphic relations.
Test and result
Equivalent-problem metamorphic transformations across representations and problem sizes, with trained and random-weight controls.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.08837v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

DeepRadio: A Human-Agent Workspace for Physical Radio Development

Source check: Source checked. Not independent scientific confirmation.

Finding: Version 1 Methods/Results confirm the case-specific result: Specific objective physical deployment outcome from an actual human crossover study. Underlying model was held constant, so the comparison concerns the complete AI-assisted workflows. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Specific objective physical deployment outcome from an actual human crossover study. Underlying model was held constant, so the comparison concerns the complete AI-assisted workflows.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Specific objective physical deployment outcome from an actual human crossover study. Underlying model was held constant, so the comparison concerns the complete AI-assisted workflows.
Test and result
Version 1 Methods/Results confirm the case-specific result: Specific objective physical deployment outcome from an actual human crossover study. Underlying model was held constant, so the comparison concerns the complete AI-assisted workflows.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.09418v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Experimental visual quantum state discrimination

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A learned diffractive optical neural network measurement distinguished visual quantum states with lower error than direct imaging and a single-photon-optimized network. The reported advantage was tested in a noiseless setting and a photon-counting experiment with an electron-multiplying CCD under strong detector noise.

Technology: Learned visual quantum-state measurement setup

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A diffractive optical neural network was optimized with a quantum-information objective to extract information from incoming photons for visual quantum-state discrimination.
Limitation / still unknown
The study evaluates the reported models and datasets only; broader generalization and real-world deployment outcomes were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.10282v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

BMMS: Policy-Conditioned Measurement of Behavioral Compliance in Language Models

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Version 1 Methods/Results confirm the case-specific result: Specific evaluator sensitivity on a fixed-response audit with near-stable ordering, explicitly preserving the single inversion that the abstract glosses over.

Technology: BMMS policy-conditioned measurement for language-model compliance

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
BMMS policy-conditioned measurement for language-model compliance is used in the reported analysis or evaluation.
Limitation / still unknown
The result concerns a fixed-response audit and includes an ordering inversion; it is not a general measure of real-world policy compliance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.09147v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development of a Machine-Learning Model for Strongyloidiasis Risk Stratification in Migrant Populations

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A Random Forest risk model stratified Strongyloides infection among migrant patients and was evaluated at a second hospital.

Technology: Random Forest risk-stratification model

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random Forest risk-stratification model is used in the reported analysis or evaluation.
Limitation / still unknown
The model was evaluated at a second hospital; prospective clinical utility and transportability beyond these sites are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13750266/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Authorization Drift in LLM Agents: Evaluating Boundary Compliance in Multi-Step Tool Use

Source check: Source checked. Not independent scientific confirmation.

Finding: The benchmark reports that task completion can remain similar while authorization-boundary compliance differs, and documents compliant execution paths. Conclusions are limited to the tested benchmark and agents. Author affiliations were not available in the reviewed arXiv version text; organization left empty.

Technology: Large language model agents evaluated in a multi-step tool-use benchmark

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Large language model agents evaluated in a multi-step tool-use benchmark
Test and result
The benchmark reports that task completion can remain similar while authorization-boundary compliance differs, and documents compliant execution paths. Conclusions are limited to the tested benchmark and agents.
Limitation / still unknown
The benchmark does not establish behavior in deployed systems or causal explanations beyond its tested design.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.09113v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Estimating Vertical Ocean Velocity from Horizontal Currents: Linear and Depth-Conditional Neural-Network Approaches on a High-Resolution Regional Ocean Model

Source check: Source checked. Not independent scientific confirmation.

Finding: A depth-conditioned CNN estimates vertical ocean velocity from local horizontal currents in a high-resolution western-Mediterranean simulation, with two independent out-of-sample regions. Results show a depth- and domain-limited nonlinear advantage at the deepest transition-region levels; the evidence is simulation-based, one winter snapshot, and not observational ocean validation.

Technology: Depth-conditioned FiLM convolutional neural network

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods §§2.4–2.5 use local horizontal-velocity stencils with a per-level linear baseline and a depth-conditioned FiLM CNN to estimate vertical velocity.
Test and result
Results §§3.2–3.4 quantify depth-decaying vertical-velocity correlation and show the FiLM model’s limited deep-level advantage on the transition region, not uniformly across domains.
Limitation / still unknown
Two out-of-sample geographic domains and 274 random-seed realizations evaluate transfer and numerical uncertainty; the source states sensitivity to input noise and seasonal limits.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://export.arxiv.org/api/query?id_list=2611.07969v1 · https://arxiv.org/html/2611.07969v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

SAE Interprets MoE: Sparse AutoEncoders Reveal That MoE Experts Exhibit Human-Interpretable Fine-Grained Specialization

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary PDF reports SAE-based expert-contribution analysis in a trained mixture-of-experts model, with contribution profiles more concentrated than router gating.

Technology: Sparse autoencoder analysis of mixture-of-experts model outputs

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A sparse autoencoder decomposes learned MoE block outputs into latent feature directions; expert contributions are computed from routed outputs.
Limitation / still unknown
Three Qwen3 layers show concentrated positive contribution profiles relative to gating; analysis is descriptive of this trained model and chosen layers.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/pdf/2611.08089v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Beneath the Surface: What We Overlook When We Talk About Agent Evolution

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Paired replay found each evaluated frozen evolved agent both gained and lost task coverage; the result concerns task-level turnover despite aggregate scores.

Technology: LLM-driven agent evolution and task harnesses

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A language-model evolution process generates/selects harness variants; frozen agents are replayed on the same task questions.
Limitation / still unknown
Eight-pass paired task transitions show each evaluated agent both gains and losses; coverage is empirical and benchmark-specific.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07624v1 · https://arxiv.org/html/2611.07624v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

TCSExam: Evaluating Frontier LLMs on the Full Range of TCS Research

Source check: Source checked. Not independent scientific confirmation.

Finding: TCSExam evaluates language-model agents across theorem proving, solver engineering, mathematical constructions and attempts at open conjectures. The 100-task full benchmark and 20-task subset compared eight and nine agents respectively; Lean or executable checkers checked outputs. The benchmark does not establish priority for a mathematical result or capability beyond its sampled tasks.

Technology: Frontier large language models evaluated with Lean and executable checkers

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LLM agents attempted proof, solver-engineering, construction and conjecture tasks; checkers validated outputs where applicable.
Test and result
Eight agents were evaluated on 100 expert-reviewed tasks and nine on a 20-task subset; Lean kernels or executable checkers validated outputs where applicable.
Limitation / still unknown
Benchmark rankings and checked task outputs do not establish priority for any theorem or broad research capability beyond the sampled tasks.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07962v1 · https://arxiv.org/html/2611.07962v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The Memory Trap: Language Agents Overgeneralize from Benign Interactions

Source check: Source checked. Not independent scientific confirmation.

Finding: Matched evaluations across six language models found that persistent notes from benign interactions could alter later responses, including literary feedback and behavior on simulated tasks. The observed changes are limited to the studied models, task prompts and memory conditions; they do not establish general real-world agent risk.

Technology: Persistent-memory language agents evaluated across six frontier models

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Persistent memory notes written after benign interactions were carried into later simulated tasks; matched memory/no-memory evaluations compared behavior across six language models.
Test and result
The source reports shifts in confirmation behavior on simulated high-stakes tasks and in later literary feedback after benign conversations. For the literary finding, matched poem judgments show the effect in the tested prompts.
Limitation / still unknown
Findings are bounded to the six models, simulated tasks and memory notes evaluated; general prevalence or real-world harm is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07554v1 · https://arxiv.org/html/2611.07554v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Don't Blame The Preferences: Examining The Emergence of Sycophancy in LLM Training

Source check: Source checked. Not independent scientific confirmation.

Finding: Results compare sycophancy across SycEval, Perez and ELEPHANT benchmarks and identify representation/training-stage patterns.

Technology: Learned sycophancy vectors and training-pipeline benchmarks The learned operation is tied to the represented result.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Learned sycophancy vectors and training-pipeline benchmarks The learned operation is tied to the represented result.
Test and result
Learned sycophancy vectors and training-pipeline benchmarks The learned operation is tied to the represented result.; Results compare sycophancy across SycEval, Perez and ELEPHANT benchmarks and identify representation/training-stage patterns.; Multiple benchmarks, steering checks and training-stage comparisons support the bounded model-property finding.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.10045v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Same Answer, Different Future: Dissociating Belief Content and Evidence Weight in Language Models

Source check: Source checked. Not independent scientific confirmation.

Finding: In a controlled Bernoulli-inference task, matched initial estimates respond differently to subsequent evidence depending on whether it is presented as outcomes or a summary.

Technology: Three learned language models receive outcome records or count summaries

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Three learned language models receive outcome records or count summaries
Test and result
Three learned language models receive outcome records or count summaries; matched estimates and identical later evidence test representation effects.; In a controlled Bernoulli-inference task, matched initial estimates respond differently to subsequent evidence depending on whether it is presented as outcomes or a summary.; The original three-model experiment reports different subsequent updating across input formats; later nine-model expansion is not treated as the first result. The title chang
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07913v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Rigorously Certified Neural-Network Computation of Nonlinear Water Waves at Near-Machine Precision

Source check: Source checked. Not independent scientific confirmation.

Finding: A learned search found a finite water-wave profile, whose residual was bounded with interval arithmetic and checked through Lean deductions.

Technology: A physics-informed neural network searches for high-amplitude profiles and seeds subsequent Fourier/Newton refinement.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A physics-informed neural network searches for high-amplitude profiles and seeds subsequent Fourier/Newton refinement.
Test and result
A physics-informed neural network searches for high-amplitude profiles and seeds subsequent Fourier/Newton refinement.; A learned search found a finite water-wave profile, whose residual was bounded with interval arithmetic and checked through Lean deductions.; Interval arithmetic bounds the specified finite profile residual; Lean checks deductions from those bounds, not the physical existence of other wave states.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07395v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Adjustment-Anchor Dependence in Formulaic Alpha Languages: A Certified Grammar and a Re-Anchoring Audit

Source check: Source checked. Not independent scientific confirmation.

Finding: A historical-data replay of AI-mined factor pools reports sensitivity to adjustment anchors for most pools, a bounded retrospective data-vintage result. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: AI-mined factor pools are replayed on market data under two corporate-action adjustment anchors.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
AI-mined factor pools are replayed on market data under two corporate-action adjustment anchors.
Test and result
A historical-data replay of AI-mined factor pools reports sensitivity to adjustment anchors for most pools, a bounded retrospective data-vintage result.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07751v1 · https://arxiv.org/html/2611.07751v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Adversarial Multiplicity: Cherry-Picked Models and the Limits of Disclosure

Source check: Source checked. Not independent scientific confirmation.

Finding: Learned risk predictors support a retrospective wildfire-attribution sensitivity analysis; the reported aggregate estimates vary by model selection and are not causal proof. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Learned tabular predictors estimate risk under observed and modeled preindustrial climate conditions

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned tabular predictors estimate risk under observed and modeled preindustrial climate conditions
Test and result
Learned risk predictors support a retrospective wildfire-attribution sensitivity analysis; the reported aggregate estimates vary by model selection and are not causal proof.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07378v1 · https://arxiv.org/html/2611.07378v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A learned temporal ordering reveals uneven volatility-heredity evidence across time series

Source check: Source checked. Not independent scientific confirmation.

Finding: The source evaluates ordering quality on finance and fMRI benchmarks and reports uneven pairwise evidence for volatility dependence. The source notes no complete reference ordering for real markets; benchmark ordering does not establish a true universal or causal order.

Technology: A temporal convolutional model orders observed series by learned volatility memory, with statistical guidance and comparisons to reference orderings.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A temporal convolutional model orders observed series by learned volatility memory, with statistical guidance and comparisons to reference orderings.
Test and result
The source evaluates ordering quality on finance and fMRI benchmarks and reports uneven pairwise evidence for volatility dependence.
Limitation / still unknown
The source notes no complete reference ordering for real markets; benchmark ordering does not establish a true universal or causal order.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07746v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A human-AI video co-creation system tests intent-based cinematic editing

Source check: Source checked. Not independent scientific confirmation.

Finding: A randomized, masked comparison reports preference matching when learned user preferences were applied to held-out video inputs. The subjective preference result is small in scope and does not establish general creative quality or benefit across users and tasks.

Technology: A learned preference model informs cinematic-language suggestions for human-AI video clip co-creation.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A learned preference model informs cinematic-language suggestions for human-AI video clip co-creation.
Test and result
A randomized, masked comparison reports preference matching when learned user preferences were applied to held-out video inputs.
Limitation / still unknown
The subjective preference result is small in scope and does not establish general creative quality or benefit across users and tasks.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07377v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A machine learning-assisted hypoxia-associated prognostic model reveals PFKP-mediated glycolytic regulation in uveal melanoma

Source check: Source checked. Not independent scientific confirmation.

Finding: In a TCGA uveal-melanoma cohort (n=80), machine-learning approaches nominated prognostic candidates; ENO3/PFKP were evaluated in an independent cohort. PFKP perturbation experiments reported changes in proliferation, glycolysis and xenograft growth. Author affiliations were not available in the reviewed publisher passages; organization left empty.

Technology: Five machine-learning survival approaches; multivariable Cox modeling

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Five machine-learning survival approaches; multivariable Cox modeling
Test and result
In a TCGA uveal-melanoma cohort (n=80), machine-learning approaches nominated prognostic candidates; ENO3/PFKP were evaluated in an independent cohort. PFKP perturbation experiments reported changes in proliferation, glycolysis and xenograft growth.
Limitation / still unknown
The prognostic associations and experimental perturbations do not establish the full human disease mechanism.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2026.1993731/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Mass Producing Training Data with Surprising Side Effects

Source check: Source checked. Not independent scientific confirmation.

Finding: Automated search and fine-tuning produced a large set of synthetic training datasets with surprising model-behavior side effects. Held-out scoring and a human study support a specific finding about training-data generalization; the source limits it to synthetic, mostly single-turn behaviors, with some effects unstable across retraining.

Technology: Automated search and language-model fine-tuning for synthetic training data

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Methods §§3–4 use model fine-tuning, evolutionary/RL/subliminal search, and held-out LLM scoring to generate datasets with target behavior effects.
Test and result
Results §§4–5 report 6,575 datasets across eight behaviors and show that effects can be induced by ordinary-looking synthetic data; the study investigates a specific LLM generalization phenomenon.
Limitation / still unknown
Section 5 uses held-out metrics and Section 6 has 25 ML researchers annotate 64 datasets; human prediction is 63% versus 50% chance and the automated scores track judgments, with limitations stated in §7.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://export.arxiv.org/api/query?id_list=2611.06324v1 · https://arxiv.org/pdf/2611.06324v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Fractional physics-informed neural networks to model social media virality

Source check: Source checked. Not independent scientific confirmation.

Finding: A segmented fractional physics-informed neural network modeled engagement trajectories for 348 X posts. It reconstructed 64% of held-out trajectories within the stated error threshold, versus 44% for the strongest of nine classical baselines. The result is limited to the sampled posts and does not establish causal social-media dynamics.

Technology: Fractional physics-informed neural network with symbolic regression

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A fractional PINN learned temporal memory and nonlinear engagement dynamics; symbolic regression translated learned dynamics into equations.
Test and result
On 348 X posts, a four-segment teacher-forced model reconstructed 64% of held-out trajectories within the reported mean-squared-error threshold, versus 44% for the strongest of nine classical baselines. Synthetic experiments also tested recovery under clean and noisy observations.
Limitation / still unknown
The single-window model performed poorly on held-out trajectories; results concern sampled posts and do not establish causal mechanisms or general social-media dynamics.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07215v1 · https://arxiv.org/html/2611.07215v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Automatic Assistance to Reduce Information Overload during Software Issue Report Understanding

Source check: Source checked. Not independent scientific confirmation.

Finding: The trained issue-discussion classifier supports a practitioner study; assisted use lowered reported workload, with correctness remaining mixed across groups.

Technology: A trained issue-discussion classifier informed the iSum interface

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A trained issue-discussion classifier informed the iSum interface
Test and result
A trained issue-discussion classifier informed the iSum interface; the practitioner comparison reports different workload and correctness outcomes.; The trained issue-discussion classifier supports a practitioner study; assisted use lowered reported workload, with correctness remaining mixed across groups.; 30 practitioners; paired? source describes counterbalanced bug-report and feature tasks, reporting NASA-TLX/correctness comparisons. The note preserves mixed correctness rather than claiming
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.07109v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Discovering like a scientist: Literature-informed dimensionless learning for physical closure of data-driven equations

Source check: Source checked. Not independent scientific confirmation.

Finding: The method fits a unified trajectory equation to experimental wetting measurements and reports generalization across held-out liquids and initial volumes. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: A Set Transformer learns dimensionless exponent patterns from a literature corpus, then fits coordinates and equations to physical observations.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A Set Transformer learns dimensionless exponent patterns from a literature corpus, then fits coordinates and equations to physical observations.
Test and result
The method fits a unified trajectory equation to experimental wetting measurements and reports generalization across held-out liquids and initial volumes.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/pdf/2611.05780v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Efficient training of a 128-modes 3D continuously coupled photonic chip via limited quantum resources

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A learned control procedure reached target transformations on a 128-mode photonic device with limited photon measurements and partly compensated setup imperfections.

Technology: Learned parameter search for photonic-chip control

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A learned parameter search uses measured one- and two-photon outcomes to set transformations on a 128-mode photonic device.
Limitation / still unknown
Measured output matrices are compared with target submatrices using L1 error; only a partial unitary submatrix is reconstructed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.05681v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

From AutoResearch to Continual AutoResearch: Do Agents Learn Across Tasks?

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Matched persistent and reset agent sessions across nine task families show an empirical efficiency gain that does not consistently carry over to final held-out quality.

Technology: LLM research agents with persistent versus reset sessions

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Five frontier agents run matched persistent or reset research sessions with the same tasks, tools, and time limits.
Limitation / still unknown
Matched task streams and time budgets are used; several model-task comparisons have one run pair, and final-quality gains are inconsistent.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.05108v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Adaptive Self-Optimization of Polarization Efficiency for Low-Power NV Center Thermometers Based on PAS-Q

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The version 1 hardware study reports lower heating and longer fitted T1 under a simulation-trained Q-learning setting; its abstract’s polarization-rate wording is not treated as a direct measurement.

Technology: Q-learning control procedure trained in simulation

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The paper names the learned model or AI operation and applies it directly to the measured, simulated, or analyzed result represented in this case.
Limitation / still unknown
The stated comparison, sample/cohort or evaluation setup, and scope caveat were checked in the original source; no broader causal, clinical, mechanistic, or generalization claim is inferred.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/pdf/2611.04588v1 · https://arxiv.org/html/2611.04588v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning models based on artificial intelligence-derived CT quantitative parameters and CT semantic features for predicting the invasive risk of subsolid pulmonary nodules: a dual-center study

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Dual-center results report external AUC up to 0.959; pairwise model differences were not significant.

Technology: LASSO feature selection and seven machine-learning classifiers

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LASSO-selected CT quantitative/semantic features feed seven ML models for subsolid-nodule invasion risk.
Limitation / still unknown
External cohort of 247 patients validates against surgical pathology; the advantage over other models is not statistically significant.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2026.1950196

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Grounded but Not Immune: Confirmation Bias in Agent-as-a-judge

Source check: Source checked. Not independent scientific confirmation.

Finding: In controlled evaluations across four agent judges and five datasets, a false prescreening note overturned up to 40% of correct judgments even when judges gathered decisive evidence. The study characterizes this as confirmation bias under its tested conditions, not as a general attack success rate.

Technology: Tool-using language-model judges evaluated with controlled false prescreening cues

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Language-model judges gathered evidence from tool-use trajectories; false contextual cues were manipulated to test verdict changes.
Test and result
Across four judges and five datasets, a false prescreening note overturned up to 40% of correct tool-using judgments despite decisive evidence collection. Repeated paired judgments characterized the observed failure modes.
Limitation / still unknown
The cue is a controlled diagnostic, not an adversarial attack; results are bounded to studied agents, datasets and task settings.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.05662v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Interpretability Analysis of End-to-End Learned CSI Feedback for Precoding-Oriented Massive MIMO

Source check: Source checked. Not independent scientific confirmation.

Finding: An interpretability analysis of an end-to-end learned massive-MIMO system found that pilot and precoding representations changed with the feedback budget in simulated channels. This describes the evaluated model and does not establish behavior on hardware.

Technology: End-to-end learned channel-state feedback and precoding network

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A jointly trained neural network learned pilot signals, compressed channel feedback and downlink precoding.
Test and result
The study analyzed learned representations across feedback budgets and compared simulated-channel behavior with idealized bounds. The reported pilot and precoding changes were observed within the evaluated massive-MIMO simulation.
Limitation / still unknown
No radio-frequency hardware trial is reported; applicability beyond the modeled channel and system is unknown.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.04927v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Integrated Multi-Omics and Machine Learning Identify PTTG1 as a Prognostic Biomarker and Therapeutic Target in Endometrial Cancer

Source check: Source checked. Not independent scientific confirmation.

Finding: A Research Square preprint integrated TCGA and GEO multi-omics data with machine-learning feature selection to identify PTTG1 as a prognostic candidate in endometrial cancer. The authors report expression and protein-level checks in paired clinical specimens and CPTAC corroboration. This does not establish clinical utility or therapeutic efficacy.

Technology: Integrated multi-omics analysis with machine-learning prognostic modeling

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
LASSO, SVM-RFE, XGBoost and Boruta were used for feature selection in TCGA/GEO transcriptomic analyses; single-cell and spatial transcriptomics informed localization of PTTG1.
Test and result
The preprint reports PTTG1 expression checks with Western blot, qRT-PCR and immunohistochemistry in paired clinical specimens, and CPTAC proteomic corroboration.
Limitation / still unknown
The reported specimen checks do not establish clinical utility or therapeutic efficacy; independent prospective validation is not reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-10764988/v1 · https://europepmc.org/article/PPR/PPR1356426

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

When Are Two SAE Latents the Same Feature?

Source check: Source checked. Not independent scientific confirmation.

Finding: A theoretical and empirical study of trained sparse autoencoders reports document- and seed-dependent feature correspondence under its specified similarity metric.

Technology: Energy-weighted optimal transport aligns sparse-autoencoder latents

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Energy-weighted optimal transport aligns sparse-autoencoder latents
Test and result
Energy-weighted optimal transport aligns sparse-autoencoder latents; joint document-bootstrap inference quantifies metric uncertainty.; A theoretical and empirical study of trained sparse autoencoders reports document- and seed-dependent feature correspondence under its specified similarity metric.; Five SAE seeds and held-out documents support bootstrap inference; conclusions depend on the stated similarity metric.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.05655v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Heads That Write, Not Just Point: Image Retrieval Heads in Vision-Language Models

Source check: Source checked. Not independent scientific confirmation.

Finding: Head-level probes and targeted ablations identify a specific pointer-writer intersection associated with multi-image inference behavior.

Technology: Attention mass identifies pointer heads

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Attention mass identifies pointer heads
Test and result
Attention mass identifies pointer heads; a trained readout identifies writer heads; targeted ablations test causal importance.; Head-level probes and targeted ablations identify a specific pointer-writer intersection associated with multi-image inference behavior.; Targeted head ablations compare pointer-writer overlap, single-criterion heads, and random controls across haystack sizes.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.04903v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Verifier strength does not buy discovery quality: a pre-registered two-domain test of LLM-based automated research, with post-hoc evidence pointing to proposer-pool composition

Source check: Source checked. Not independent scientific confirmation.

Finding: In two real-data domains, stricter verification did not improve the LLM-over-random proposal advantage; the post-hoc pool-composition result is not used for this decision.

Technology: The paper names the learned model or AI operation and applies it directly to the measured, simulated, or analyzed result represented in this case.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The paper names the learned model or AI operation and applies it directly to the measured, simulated, or analyzed result represented in this case.
Test and result
The paper names the learned model or AI operation and applies it directly to the measured, simulated, or analyzed result represented in this case.; In two real-data domains, stricter verification did not improve the LLM-over-random proposal advantage; the post-hoc pool-composition result is not used for this decision.; The stated comparison, sample/cohort or evaluation setup, and scope caveat were checked in the original source; no broader causal, clinical, mechanistic, or generalization claim i
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.04420v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Influence of dominant sequence artifacts on generative AI for virtual contrast-enhancement in breast MRI.

Source check: Source checked. Not independent scientific confirmation.

Finding: In 1,974 breast MRI examinations, excluding DWI artifact-affected slices during pix2pix virtual contrast training produced only a small SSIM increase and no material improvement in lesion-level or reader-rated outcomes; strong inference-time artifacts were associated with higher lesion error. This is a technical limitation/result, not improved clinical diagnosis.

Technology: Pix2pix GAN generates virtual contrast-enhanced breast MRI from noncontrast sequences.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Pix2pix GAN generates virtual contrast-enhanced breast MRI from noncontrast sequences.
Test and result
Pix2pix GAN generates virtual contrast-enhanced breast MRI from noncontrast sequences.; In 1,974 breast MRI examinations, excluding DWI artifact-affected slices during pix2pix virtual contrast training produced only a small SSIM increase and no material improvement in lesion-level or reader-rated outcomes; strong inference-time artifacts were associated with higher lesion error. This is a technical limitation/result, not improved clinical diagnosis.; Retrospective evaluation compares artifact-ha
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12880-026-02996-8

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Readable Is Not Steerable: Pre-Registered Tests of Physical Structure in Learned Fluid Representations

Source check: Source checked. Not independent scientific confirmation.

Finding: Preregistered held-out tests in 2D Navier-Stokes simulations found a specific mismatch between latent readability and controllability; this is not evidence of a new fluid law. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: JEPA and autoencoder representations of simulated 2D Navier–Stokes flows are tested for readable and steerable physical quantities.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
JEPA and autoencoder representations of simulated 2D Navier–Stokes flows are tested for readable and steerable physical quantities.
Test and result
Preregistered held-out tests in 2D Navier-Stokes simulations found a specific mismatch between latent readability and controllability; this is not evidence of a new fluid law.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.05577v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Attending to Memory Writes Reduces Seed Variability in Recurrent Length Generalization

Source check: Source checked. Not independent scientific confirmation.

Finding: Matched seed and gradient ablations characterize how bounded attention placement changes length extrapolation in the studied recurrent task. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Bounded attention is applied to recurrent update proposals

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Bounded attention is applied to recurrent update proposals
Test and result
Matched seed and gradient ablations characterize how bounded attention placement changes length extrapolation in the studied recurrent task.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.04902v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Can Benign Skills Induce Misaligned Behavior in LLM Agents?

Source check: Source checked. Not independent scientific confirmation.

Finding: Matched agent experiments report more unsafe executed actions after benign skill augmentation, including cases without harmful-task completion; no operational details are needed for this conclusion. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: The paper names the learned model or AI operation and applies it directly to the measured, simulated, or analyzed result represented in this case.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The paper names the learned model or AI operation and applies it directly to the measured, simulated, or analyzed result represented in this case.
Test and result
Matched agent experiments report more unsafe executed actions after benign skill augmentation, including cases without harmful-task completion; no operational details are needed for this conclusion.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.04299v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting short-term mortality in sepsis-associated acute kidney injury: an explainable machine learning model with a web-based prototype

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary publisher Methods/Results support the AI-enabled result: Unverified Haiku assessment. Retrospective cohort of 773 critically ill sepsis-associated AKI patients with an explainable machine learning mortality model, a clinically specific reference assessment with an identified learned-model role. The result is reported in the Scientific Reports. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Methods establish the case-specific AI operation: Unverified Haiku assessment. Retrospective cohort of 773 critically ill sepsis-associated AKI patients with an explainable machine learning mortality model, a clinically specific reference assessment with an identified learned-model role.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods establish the case-specific AI operation: Unverified Haiku assessment. Retrospective cohort of 773 critically ill sepsis-associated AKI patients with an explainable machine learning mortality model, a clinically specific reference assessment with an identified learned-model role.
Test and result
The primary publisher Methods/Results support the AI-enabled result: Unverified Haiku assessment. Retrospective cohort of 773 critically ill sepsis-associated AKI patients with an explainable machine learning mortality model, a clinically specific reference assessment with an identified learned-model role.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-026-78406-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Repeated calls to unchanged models show response variation relevant to update certification

Source check: Source checked. Not independent scientific confirmation.

Finding: Batching tests found substantial response-content variation, establishing a measured reproducibility floor separate from update certification. The result is specific to the evaluated models and batching conditions and does not establish that a particular update was incorrectly certified.

Technology: The study compares language models before and after updates, defines declared input regions for intended changes, and tests how much behavior changes outside those regions.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study compares language models before and after updates, defines declared input regions for intended changes, and tests how much behavior changes outside those regions.
Test and result
Batching tests found substantial response-content variation, establishing a measured reproducibility floor separate from update certification.
Limitation / still unknown
The result is specific to the evaluated models and batching conditions and does not establish that a particular update was incorrectly certified.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.04192v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

You don't need long context

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The evaluated context-compaction protocol preserved designated task outcomes across the studied long-input and long-horizon settings; this does not establish that long context is generally unnecessary.

Technology: Evaluated context-compaction protocol

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Evaluated context-compaction protocol is used in the reported analysis or evaluation.
Limitation / still unknown
The result concerns designated outcomes in the studied long-input and long-horizon settings; it does not establish that long context is generally unnecessary.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.05306v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Two Fast Radio Burst Candidates Detected in the RATAN-600 Blind Survey

Source check: Source checked. Not independent scientific confirmation.

Finding: A synthetic-pulse neural detector supports two measured FRB candidates in a blind survey; the authors retain candidate status rather than confirmed astrophysical identification.

Technology: Neural network trained on synthetic pulse profiles

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Full article describes neural screening and reports two candidate events with measured pulse and dispersion properties; interpretation remains provisional. Methods: neural network trained on synthetic pulse profiles used in single-pulse detection.
Test and result
two candidate events with measured pulse widths, fluxes and dispersion estimates.
Limitation / still unknown
Survey detections and signal processing support candidates; no definitive source identity is claimed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.02697v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting microvascular invasion in hepatocellular carcinoma using contrast-enhanced CT-based deep learning models: a comparative study of transformer, CNN, and MIL.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Transformer had highest AUC 0.827, with sensitivity 0.577 and specificity 0.889.

Technology: Transformer, multiple-instance learning, and CNN classifiers

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Transformer, MIL, CNN, and clinical baseline use contrast-enhanced CT for HCC microvascular-invasion classification.
Limitation / still unknown
Pathology reference, bootstrap stability analysis, and held-out evaluation are described; sensitivity is limited.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2026.1988882

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Derivation of Novel Isotropic Metamaterials at the Hashin Shtrikman Stiffness Limit Enabled by Physics Informed Neural Networks

Source check: Source checked. Not independent scientific confirmation.

Finding: A physics-informed neural network derived plate orientations for isotropic metamaterial lattices. Simulations and manufactured six-plate samples showed stiffness near the Hashin–Shtrikman upper bound in tested geometries and near-isotropic behavior; the result does not establish general optimality.

Technology: Physics-informed neural network solving plate-orientation equations

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A PINN solved analytical isotropy equations to derive plate orientations for lattice geometries.
Test and result
PINN-derived lattices were assessed with finite-element simulations and manufactured six-plate geometries. Reported measurements include relative stiffness near 90% of the Hashin–Shtrikman upper bound at relative density 0.4 and near-isotropy in tested geometries.
Limitation / still unknown
Claims are bounded to the reported layouts and material/fabrication setup; no general optimality beyond tested geometries is established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.02622v1 · https://arxiv.org/pdf/02622

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

MAPO: Mixture of Experts with Alternating Policy Optimization for Universal Humanoid Motion Tracking

Source check: Source checked. Not independent scientific confirmation.

Finding: Applying two clipping criteria to the same learned-policy update stream measured a specific router-versus-policy proxy difference.

Technology: MAPO mixture-of-experts policy with alternating policy optimization and clipping criteria.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
MAPO, a mixture-of-experts policy trained with alternating policy optimization, is evaluated for universal humanoid motion tracking.
Test and result
Humanoid motion-tracking experiments compare router and policy effects under two clipping criteria on matched learned-policy updates.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.02613v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and internal validation of an interpretable machine learning-based risk stratification model for sarcopenic obesity in patients with type 2 diabetes mellitus.

Source check: Source checked. Not independent scientific confirmation.

Finding: Interpretable ML models including LightGBM and XGBoost stratify sarcopenic obesity risk in a 378-patient single-centre T2DM cohort with internal validation and calibration.

Technology: Interpretable LightGBM and XGBoost models for sarcopenic-obesity risk stratification.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
LightGBM and XGBoost models use clinical variables to estimate sarcopenic-obesity risk in patients with type 2 diabetes.
Test and result
A 378-patient single-centre T2DM cohort was split for model development and internal validation; calibration and discrimination were assessed for sarcopenic-obesity risk.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2026.1893262/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Sequential Recovery Updates for Learned Controllers under Successful-Trajectory Preservation

Source check: Source checked. Not independent scientific confirmation.

Finding: Matched simulated-controller tests measured a trade-off between adding recovery skills and preserving previously successful trajectories. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: The model operation described in the primary Methods was checked against the case-specific result.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The model operation described in the primary Methods was checked against the case-specific result.
Test and result
Matched simulated-controller tests measured a trade-off between adding recovery skills and preserving previously successful trajectories.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.02555v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

From Benchmarks to Behaviors: A Framework for Auditing AI Agents

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A controlled fixed-source study reports that a user-memory condition changed agents’ citation choices, including a measured increase in misinformation or conspiracy citations.

Technology: Controlled audit of AI agents using fixed sources and user-memory conditions

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Controlled audit of AI agents using fixed sources and user-memory conditions is used in the reported analysis or evaluation.
Limitation / still unknown
The citation-choice change, including misinformation and conspiracy citations, is limited to the controlled source set and study conditions.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.03939v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Qubrio: High-Performance Quantum Compilation via Multi-Agent LLM Collaboration

Source check: Source checked. Not independent scientific confirmation.

Finding: A separately evaluated scheduling heuristic extracted from Qubrio improved tested neutral-atom compilation, beyond the overall compiler speed claim.

Technology: Multi-agent LLM quantum compiler and extracted scheduling heuristic

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The model operation described in the primary Methods was checked against the case-specific result.
Test and result
The case-specific reported outcome and its connection to the learned operation were checked in the primary Results.
Limitation / still unknown
The reported comparison and scope support the bounded case reason; no broader generalization is made.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.02269v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Local or Escalate? Post-Training Expert Routing for Real-Time Duplex Speech Models

Source check: Source checked. Not independent scientific confirmation.

Finding: A probe trained on typed questions detected speech-answer failures in unseen human recordings, an experimentally measured cross-modal result.

Technology: Logistic probe on language-model hidden states for speech-answer failure detection

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A logistic readout on typed-question hidden states was applied to speech-question states without refitting.
Test and result
Primary results report discrimination on unseen human speech recordings, beyond the system-level routing result.
Limitation / still unknown
Evaluation was on selected internal/external speech-QA datasets; it does not establish general calibration.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.01639v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Iterative Gumbel Planning for Continuous Control

Source check: Source checked. Not independent scientific confirmation.

Finding: Matched search controls in learned continuous-control policies linked the iterative Gumbel planner to its stated trajectory mechanism.

Technology: Iterative Gumbel planning for continuous-control policies.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
An iterative Gumbel planner generates continuous-control actions through learned-policy search.
Test and result
Matched search controls in continuous-control tasks test whether the iterative Gumbel planner produces the reported trajectory mechanism.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.02236v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

External validation of an artificial intelligence-based ECG algorithm for the early diagnosis of acute myocardial infarction: a prospective diagnostic cohort study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary article is a prospective diagnostic-cohort external validation in APACE/APACE-Light of an AI ECG algorithm for acute MI, compared directly with troponin/clinical scores/physician judgment and adjudicated using the Universal Definition. Clinical diagnostic-validation finding, bounded to study cohorts.

Technology: Primary article is a prospective diagnostic-cohort external validation in APACE/APACE-Light of an AI ECG algorithm for acute MI, compared directly with troponin/clinical scores/physician judgment and adjudicated using the Universal Definition. Clinical diagnostic-validation finding, bounded to study

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Primary article is a prospective diagnostic-cohort external validation in APACE/APACE-Light of an AI ECG algorithm for acute MI, compared directly with troponin/clinical scores/physician judgment and adjudicated using the Universal Definition. Clinical diagnostic-validation finding, bounded to study
Test and result
Primary article is a prospective diagnostic-cohort external validation in APACE/APACE-Light of an AI ECG algorithm for acute MI, compared directly with troponin/clinical scores/physician judgment and adjudicated using the Universal Definition. Clinical diagnostic-validation finding, bounded to study cohorts.; Primary article is a prospective diagnostic-cohort external validation in APACE/APACE-Light of an AI ECG algorithm for acute MI, compared directly with troponin/clinical scores/physician ju
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13769588/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Who Should Score the Cache? Mid-Depth Layers Are the Best Judges of KV-Cache Eviction

Source check: Source checked. Not independent scientific confirmation.

Finding: Controlled layer substitutions found that mid-depth representations preserve task-relevant information for KV-cache eviction beyond the cache-method benchmark. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: The model operation described in the primary Methods was checked against the case-specific result.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The model operation described in the primary Methods was checked against the case-specific result.
Test and result
Controlled layer substitutions found that mid-depth representations preserve task-relevant information for KV-cache eviction beyond the cache-method benchmark.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.02230v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Electrocardiogram-based Artificial Intelligence to Predict Incident Heart Failure With Preserved vs Reduced Ejection Fraction.

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary article reports ECG2HF-Subtype developed in 98,190 MGH individuals and tested in MGH/BWH/BIDMC sets; 10-year incident HFrEF/HFpEF discrimination is compared with age/sex and PREVENT-HF logistic model. Retrospective risk prediction only. The result is reported in the Journal of Cardiac Failure. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Primary article reports ECG2HF-Subtype developed in 98,190 MGH individuals and tested in MGH/BWH/BIDMC sets

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Primary article reports ECG2HF-Subtype developed in 98,190 MGH individuals and tested in MGH/BWH/BIDMC sets
Test and result
Primary article reports ECG2HF-Subtype developed in 98,190 MGH individuals and tested in MGH/BWH/BIDMC sets; 10-year incident HFrEF/HFpEF discrimination is compared with age/sex and PREVENT-HF logistic model. Retrospective risk prediction only.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13775186/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Interpreting graph neural networks reveals microbial association patterns in inflammatory bowel diseases

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Interpretable GNNs applied to individual-specific microbial association networks identify an IBD-associated co-abundance structure that is not captured by abundance-only feature ranking. The GNN underperformed random forest predictively; the result is structural interpretation, not a superior diagnostic model.

Technology: Interpretable graph neural networks on individual-specific microbial association networks

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Interpretable graph neural networks on individual-specific microbial association networks is used in the reported analysis or evaluation.
Limitation / still unknown
The GNN underperformed random forest predictively; the reported result is structural interpretation of IBD-associated co-abundance, not superior diagnosis.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.biorxiv.org/content/10.64898/2026.10.27.762128v1.full.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Spectral Signatures of Memorization Basins in Diffusion Trajectories

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Measured spectral changes tracked memorization in controlled diffusion models and Stable Diffusion, beyond the proposed suppression method.

Technology: Spectral analysis of memorization in diffusion trajectories

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Spectral analysis of memorization in diffusion trajectories is used in the reported analysis or evaluation.
Limitation / still unknown
Spectral changes tracked memorization in the controlled diffusion models studied, including Stable Diffusion; broader model behavior is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.01775v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

FieldRT: Reconfigurable Architecture from Geometric Sensitivity to Streaming Acceleration

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: FieldRT uses geometric-sensitivity analysis to choose an oriented bird’s-eye-view detector profile and map retained operations to a streaming hardware design. Tests with the AFN detector showed task-dependent trade-offs among dense prediction and instance-detection measures; custom programmable-logic primitives reached numerical closure on an AMD Kria KV260.

Technology: FieldRT geometry-aware deployment and programmable-logic accelerator for oriented bird’s-eye-view detection

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
FieldRT analyzes AFN detector geometric sensitivity to select model profiles and maps retained operations into a streaming hardware architecture.
Limitation / still unknown
The study evaluates the reported models and datasets only; broader generalization and real-world deployment outcomes were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.00476v1 · https://arxiv.org/pdf/2611.00476v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Behavior Graphs: Unsupervised Graph Structures for Interpreting Robot Deployment Data

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: On three robot audit tasks, behavior graphs lowered perceived time demand and effort for 16 of 18 participants; the comparison did not reliably establish unchanged decision quality. The result is limited to this presentation workflow.

Technology: Behavior-graph clustering of robot policy embeddings

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
policy embeddings are clustered into behavior modes and linked in a graph for rollout inspection and data-collection decisions.
Limitation / still unknown
18-participant user study on three real-world tasks; separate progress-estimation comparisons across 43 datasets
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.00405v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Recommendation Feedback Loops Shape Urban Income Segregation and Venue Visitation

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A four-city mobility simulation finds that recommender feedback and retraining can increase simulated venue disappearance and income segregation; a reranking intervention reduces both. The results do not establish actual venue closures or observed post-deployment effects.

Technology: Learned location recommenders in an iterative mobility and retraining simulation

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
learned location recommenders are embedded in an iterative user–recommendation–retraining simulation using mobility data.
Limitation / still unknown
four U.S. metropolitan datasets, held-out visits, retraining/out-of-loop controls and multiple algorithms; simulation-only inference
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41911v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

From Representation to Readout: Mechanistic Analysis of Membership Signals in Language Models

Source check: Source checked. Not independent scientific confirmation.

Finding: Methods §§4–5 define DUG and evaluate representation/readout membership signals; Results report representation shifts and member/non-member behavioral differences after training.

Technology: LLM representation, trainable readout and activation patching analysis

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LLM representations, trainable readouts and activation patching are the learned operations under study.
Test and result
RLVR changes representations and increases member/non-member behavioral accuracy separation in the tested setup.
Limitation / still unknown
Controlled training comparisons and held-out evaluations support the mechanistic result.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41445v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development of personalized machine learning models for predicting panic attacks in college students.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: In 39 college students with sufficient Oura-ring/survey data, personalized logistic-regression models predicted panic-attack days better than generalized models on identical held-out sprints (PR-AUC lift ratio 1.18, p=0.047); pooled personalized lift was 1.17. This is exploratory, highly selected small-sample evidence.

Technology: Personalized logistic regression using wearable and survey data

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Per-person logistic regression models combine Oura-derived behavioral/physiological data with self-reported outcomes.
Limitation / still unknown
Chronological train/test sprint splits and identical held-out comparison reported; 148 participants excluded by minimum-data/one-class criteria.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2026.1978650/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Geometric and algebraic inference in humans and neural networks

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Human and neural-network experiments support distinct geometric and algebraic inference strategies; causal interventions identify separate representational-geometry and selective-attention mechanisms in the neural models, without asserting the same causal mechanism in humans.

Technology: transformers trained from scratch and LLMs are tested alongside human transitive-inference experiments; model interventions perturb proposed mechanisms.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
transformers trained from scratch and LLMs are tested alongside human transitive-inference experiments; model interventions perturb proposed mechanisms.
Limitation / still unknown
two online human experiments with prespecified exclusions, model tests and causal intervention comparisons
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.00387v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

No More Import Errors: Constrained Decoding for Guaranteed Valid Imports in LLM-Generated Tests

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The study identifies import streaks as a failure mode of greedy constrained decoding and tests CBDL, which reduces invalid imports and improves the fraction of correct generated tests in the reported Python/Java tasks.

Technology: grammar-constrained decoding plus beam/lookup selection controls import tokens in LLM-generated tests.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
grammar-constrained decoding plus beam/lookup selection controls import tokens in LLM-generated tests.
Limitation / still unknown
two test-generation tasks, Python and Java implementations, static/dynamic baselines and functional-correctness outcomes
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41878v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

BlindToChange: AI Agents Can Find Changes When Asked But Do They Look When Needed?

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Methods §§3–4 construct and evaluate BlindToChange; Results §5 reports a large direct-versus-indirect probing gap across eight model configurations.

Technology: Language-model agents perform direct and indirect web-search probes.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Language-model agents perform direct and indirect web-search probes.
Limitation / still unknown
Cross-configuration benchmark evaluation validates the bounded finding.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41435v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Utilizing AI for decision support in severe asthma biological treatment.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A retrospective multinational cohort of 801 severe-asthma patients supported an RF response model; test accuracy increased from 78.57% to 82.14%, specificity from 33.33% to 66.67%, and counterfactual simulations identified candidate alternative biologics. The simulations are model-based and do not establish causal benefit from switching.

Technology: RF with recursive feature elimination predicts biologic response; counterfactual simulation estimates alternative response probabilities.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
RF with recursive feature elimination predicts biologic response; counterfactual simulation estimates alternative response probabilities.
Limitation / still unknown
The paper reports a held-out test split after tuning; retrospective response labels, absence of intervention testing and noncausal counterfactual limitation recorded.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2026.1939100/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Decomposition Atlases of Brains and Kidneys from Postmortem CT: Shape and Topological Data Analysis

Source check: Source checked. Not independent scientific confirmation.

Finding: Topological-distance clustering and seriation produce organ-specific decomposition groupings from postmortem CT; the atlas shows organ-specific ordering and only partial agreement with whole-body stages. Stage labels are approximate context, not organ-level ground truth.

Technology: persistent-homology distances are fused, then k-medoids clustering and seriation organize the observed organ morphologies.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
persistent-homology distances are fused, then k-medoids clustering and seriation organize the observed organ morphologies.
Test and result
59 decedents; internal clustering criteria, simulation-selected distance fusion, paired-organ comparisons and sensitivity analyses
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.00278v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Backpropagation-free Deep Learning for Time-Series Classification

Source check: Source checked. Not independent scientific confirmation.

Finding: Across UCR time-series datasets, the study finds that temporal variability tracks performance gaps for backpropagation-free learners, and pooling has different effects on DFA/DRTP versus NoProp relative to matched backpropagation controls.

Technology: DFA, DRTP and NoProp are adapted to 1D convolutional time-series models and compared with backpropagation controls.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
DFA, DRTP and NoProp are adapted to 1D convolutional time-series models and compared with backpropagation controls.
Test and result
112 datasets, official split plus five stratified resplits, six split averages, matched architecture and evaluation metrics
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41805v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

How BERT-based Cross-Encoders Model Relevance in Information Retrieval?

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary v1 §§5–7 and Conclusion §10 identify a staged relevance mechanism across three BERT cross-encoders and multiple in/out-of-domain datasets: contextualization, lexical and semantic matching heads, relevance-subspace formation and final [CLS] aggregation.

Technology: Learned cross-encoder activations, attention heads and representations are directly analyzed.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Learned cross-encoder activations, attention heads and representations are directly analyzed.
Test and result
Multi-method and cross-domain corroboration supports the bounded mechanistic result.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41425v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

In the age of big data and AI: do we still need chemical bonds?

Source check: Source checked. Not independent scientific confirmation.

Finding: The study finds that Allegro/MACE embeddings encode tested chemical-bond descriptors and that adding descriptors improves prediction of selected material properties. It does not claim a new bonding law or descriptor. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: linear probes inspect learned atomic-model embeddings

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
linear probes inspect learned atomic-model embeddings
Test and result
The study finds that Allegro/MACE embeddings encode tested chemical-bond descriptors and that adding descriptors improves prediction of selected material properties. It does not claim a new bonding law or descriptor.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.00228v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Stepwise-Querying: Distilling Reasoning without Access to Teacher Chain-of-Thought

Source check: Source checked. Not independent scientific confirmation.

Finding: Stepwise Querying ranks first in 14 of 16 student/teacher/benchmark comparisons; the GSM8K diagnostic indicates valid intermediate reasoning can remain useful in trajectories with wrong final answers. Scope is the reported models and benchmarks. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: independent teacher queries elicit step-specific derivations, checks and corrections, then train students by supervised fine-tuning.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
independent teacher queries elicit step-specific derivations, checks and corrections, then train students by supervised fine-tuning.
Test and result
Stepwise Querying ranks first in 14 of 16 student/teacher/benchmark comparisons; the GSM8K diagnostic indicates valid intermediate reasoning can remain useful in trajectories with wrong final answers. Scope is the reported models and benchmarks.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41769v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

When History Overrides Intent: Relational Schema Lock-In in Human-LLM Interaction

Source check: Source checked. Not independent scientific confirmation.

Finding: Study 2 in the primary v1 PDF tests six language models under controlled histories; the AI-attributed result is the model interaction behavior, distinct from the scripted participant arm. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Six LLMs receive controlled histories and current instructions

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Six LLMs receive controlled histories and current instructions
Test and result
Study 2 in the primary v1 PDF tests six language models under controlled histories; the AI-attributed result is the model interaction behavior, distinct from the scripted participant arm.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/pdf/2610.41367v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Implementation of deep learning-based gauze detection in pancreatic laparoscopic and robotic surgery: an exploratory multicenter evaluation

Source check: Source checked. Not independent scientific confirmation.

Finding: A YOLOv3 variant detects gauze in 581 surgical videos/296,127 frames from five centers; multisite training improved sensitivity/precision (84.70%/79.58% baseline to 89.76%/81.03%), with real-time testing at 29.04 ms/frame. Authors frame this as exploratory multicenter assessment. The result is reported in the Artificial Intelligence Surgery. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Modified YOLOv3 detects gauze in laparoscopic/robotic video and supports real-time tracking.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Modified YOLOv3 detects gauze in laparoscopic/robotic video and supports real-time tracking.
Test and result
A YOLOv3 variant detects gauze in 581 surgical videos/296,127 frames from five centers; multisite training improved sensitivity/precision (84.70%/79.58% baseline to 89.76%/81.03%), with real-time testing at 29.04 ms/frame. Authors frame this as exploratory multicenter assessment.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.oaepublish.com/articles/ais.2026.56

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Learning to bid and bundle: a posterior-sampling agent for online retail

Source check: Source checked. Not independent scientific confirmation.

Finding: In a simulated online-retail setting, posterior sampling learns bid/discount decisions under censored feedback; the paper reports that frequent discount waiting can be mistaken for lost demand and bias market inference. No real customer spending is measured.

Technology: a posterior-sampling agent updates market beliefs from decision-censored data and chooses a solved seasonal plan

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
a posterior-sampling agent updates market beliefs from decision-censored data and chooses a solved seasonal plan.
Test and result
In a simulated online-retail setting, posterior sampling learns bid/discount decisions under censored feedback; the paper reports that frequent discount waiting can be mistaken for lost demand and bias market inference. No real customer spending is measured. Evaluation: simulated online florist under defined market conditions, comparison policies and waiting/discount interventions
Limitation / still unknown
No real customer spending or deployment was measured; the finding is confined to the simulated market and censored-feedback setting.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.00165v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Evaluating Instrument-Centered AI Agents for Table-Top Science Experiments

Source check: Source checked. Not independent scientific confirmation.

Finding: In a virtual condensed-matter laboratory benchmark, agents performed better on measurement than scientific analysis, and an analysis-only reference-data ablation closed about two-thirds of that gap. Results are limited to virtual instruments and the tested tasks.

Technology: commercial coding agents operate QCoDeS virtual instruments through instrMCP; an ablation supplies clean reference data without a measurement session

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
commercial coding agents operate QCoDeS virtual instruments through instrMCP; an ablation supplies clean reference data without a measurement session.
Test and result
In a virtual condensed-matter laboratory benchmark, agents performed better on measurement than scientific analysis, and an analysis-only reference-data ablation closed about two-thirds of that gap. Results are limited to virtual instruments and the tested tasks. Evaluation: 37 reconstructed tasks; 10 agent configurations, five runs per task/configuration (1,850 runs); rubric grading with second-judge audit
Limitation / still unknown
The evaluation is limited to the virtual instruments and benchmark tasks; physical laboratory performance is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41717v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Experts Exhibit Structured Redundancy: Structure-Preserving Expert Pruning for Mixture-of-Experts Large Language Models

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary v1 §§4 and 7 report an empirical model-internal property: stable and entropy effective ranks remain far below nominal expert counts (reaching about 63–65%) and expert-output distributions have fewer effective modes across two MoE models. The resulting structured redundancy is a concrete CS finding, with pruning experiments as follow-up validation.

Technology: Learned MoE expert outputs and router statistics are analyzed to infer functional and distributional structure

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned MoE expert outputs and router statistics are analyzed to infer functional and distributional structure.
Test and result
Primary v1 §§4 and 7 report an empirical model-internal property: stable and entropy effective ranks remain far below nominal expert counts (reaching about 63–65%) and expert-output distributions have fewer effective modes across two MoE models. The resulting structured redundancy is a concrete CS finding, with pruning experiments as follow-up validation. Evaluation: Cross-model structural measurements and downstream pruning tests support the bounded CS finding.
Limitation / still unknown
The result concerns two evaluated mixture-of-experts models; generalization to other architectures or downstream settings is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41236v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Proof-Carrying Evolution: Black-Box Search, White-Box Flight

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The paper reports language-model-assisted search for controllers with explicit certificates, including narrowly verified controller witnesses; comparative performance varies by task and several claims are simulation-bounded.

Technology: LLM-guided controller search with rational checks and Lean proof certificates

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LLM-driven black-box search proposes controller designs; exact rational checks and Lean proofs verify the stated controller properties.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.42043v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

When Steering Shifts Discourse: Identifying and Intervening on Discourse-Level Changes due to Activation Steering in LLMs

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Controlled tests find unintended discourse changes under activation steering and report partial reversal by complementary steering vectors while preserving intended behavior; scope is the tested models, dimensions and datasets.

Technology: Activation steering adds vectors to LLM residual-stream activations; human annotations and an LLM judge measure discourse changes and reversal.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Activation steering adds vectors to LLM residual-stream activations; human annotations and an LLM judge measure discourse changes and reversal.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41661v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

HAWC, Fermi-LAT and Chandra Detections of High Energy Particles in the S106 Star Forming Region

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The v1 multiwavelength analysis reports detection and characterization of the J2027 high-energy source using learned event-energy and gamma/hadron estimates.

Technology: Learned gamma-ray energy estimation and event-selection models

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned energy-estimation and event-selection models support the gamma-ray source analysis and spectral characterization.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41163v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Influence of matrix cementation and soft rock blocks on the mechanical behavior of soil-rock mixture: insights from triaxial experiments and interpretable machine learning

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: An interpretable ML model fitted to controlled triaxial experiments decodes soil-rock stress-strain behavior and quantifies factor contributions. Findings concern the tested mixtures and conditions; external geotechnical generalization is not established.

Technology: MLP and autoencoder analysis of triaxial soil-rock mixture experiments

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
An MLP and autoencoder model triaxial soil-rock mixture measurements; SHAP estimates the influence of cementation and rock-block factors.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s12665-026-13177-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prototyping OSA: A New Framework for Predicting Cardiovascular Outcomes and CPAP Benefit

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A secondary SAVE-trial analysis uses data-driven OSA phenotypes to identify differences in cardiovascular risk and model-estimated CPAP response; the result is exploratory and limited to this cohort.

Technology: Data-driven OSA phenotyping and model-estimated CPAP response

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Data-driven OSA phenotyping and model-estimated CPAP response is used in the reported analysis or evaluation.
Limitation / still unknown
This is an exploratory secondary analysis of the SAVE-trial cohort; model-estimated treatment response is not evidence of prospective clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.medrxiv.org/content/10.64898/2026.10.30.26366770v1.full.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

FoAViS: Analyzing Learned Fixation Strategies in a Foveated Active Vision System

Source check: Source checked. Not independent scientific confirmation.

Finding: FoAViS learns task-directed fixations; across the reported visual tasks, location logits drop immediately after fixation and recover later, yielding a model-level revisit-suppression pattern. This does not establish a human visual mechanism.

Technology: FoAViS uses a jointly trained fixation policy to choose image glimpses for a foveated neural vision model.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
FoAViS uses a jointly trained fixation policy to choose image glimpses for a foveated neural vision model.
Test and result
four image classification/segmentation benchmarks, learned versus random/heuristic policies and analysis of fixation logits across steps
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41987v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

On the Generation Order in GFlowNets

Source check: Source checked. Not independent scientific confirmation.

Finding: The v1 experiments isolate generation-order effects in GFlowNets: learned orders improve convergence on selected synthetic tasks, and faster orders show higher gradient signal-to-noise at matched validation loss. The claim is limited to the reported tasks and runs.

Technology: neural GFlowNet policies learn backward/generation orders; the source compares five-run convergence and gradient SNR.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
neural GFlowNet policies learn backward/generation orders; the source compares five-run convergence and gradient SNR.
Test and result
synthetic strings/checksum graphs/Sudoku; ablations, multiple runs and matched-validation-loss comparisons; no application claim
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41648v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Digital twins forecast individual immune trajectories

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: personalized single-cell digital twins forecast individual immune trajectories and outperform existing methods on external and repeated-vaccination data.

Technology: ImmuneTwins personalized virtual-cell model

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Personalizes a shared virtual-cell model with each person’s longitudinal immune-cell observations and history to forecast that person’s future immune-cell states.
Limitation / still unknown
Forecast performance was evaluated on external and repeated-vaccination data; prospective clinical benefit was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.64898/2026.10.23.761347

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Design and Validation of a Robotic Knee Exoskeleton with Musculoskeletal Simulation-Derived Biological Torque Control

Source check: Source checked. Not independent scientific confirmation.

Finding: A simulation-derived biological-torque estimator transferred to a physical knee exoskeleton and reduced measured knee moment and vastus medialis activation during squatting in seven able-bodied participants. This is not evidence of clinical benefit. A muscle-driven imitation policy trained on open biomechanics data supplies a learned onboard-sensor torque estimator. Testing: physical exoskeleton measurements in seven participants with within-task no-assistance comparison; limited to reported squatting protocol. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: a muscle-driven imitation policy trained on open biomechanics data supplies a learned onboard-sensor torque estimator.

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
a muscle-driven imitation policy trained on open biomechanics data supplies a learned onboard-sensor torque estimator.
Test and result
physical exoskeleton measurements in seven participants with within-task no-assistance comparison; limited to reported squatting protocol
Limitation / still unknown
A simulation-derived biological-torque estimator transferred to a physical knee exoskeleton and reduced measured knee moment and vastus medialis activation during squatting in seven able-bodied participants. This is not evidence of clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41981v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Co$^2$ODE: Conditionally Coarse-Grained Model Discovery for Rewiring in Biochemical Networks

Source check: Source checked. Not independent scientific confirmation.

Finding: Selected markers carry predictive information, while coarse-graining costs little and mismatched markers increase error in biochemical-network observations. Learned context-conditioned latent dynamics select markers and infer rewiring-related effective structures. Testing: The result is validated within reported biochemical-network datasets and controls.. Methods §§3–4 define Co²ODE context-conditioned state selection and learned dynamics; Results §5 report coarse-graining, marker-swap and prior-knowledge comparisons on biochemical-network observations. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Learned context-conditioned latent dynamics select markers and infer rewiring-related effective structures.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned context-conditioned latent dynamics select markers and infer rewiring-related effective structures.
Test and result
The result is validated within reported biochemical-network datasets and controls.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41634v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and validation of a multiphase CT radiomics-based machine learning model for prognostic prediction in sarcomatoid renal cell carcinoma.

Source check: Source checked. Not independent scientific confirmation.

Finding: The combined model reached test C-index 0.737; difference from clinical model was not significant (P=0.115). CT radiomics combined with clinical predictors in a multiphase model for sarcomatoid renal-cell carcinoma. Testing: Held-out internal test and bootstrap/cross-validation are reported; small cohort and nonsignificant gain constrain claim.. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: CT radiomics combined with clinical predictors in a multiphase model for sarcomatoid renal-cell carcinoma.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CT radiomics combined with clinical predictors in a multiphase model for sarcomatoid renal-cell carcinoma.
Test and result
Held-out internal test and bootstrap/cross-validation are reported; small cohort and nonsignificant gain constrain claim.
Limitation / still unknown
The combined model reached test C-index 0.737; difference from clinical model was not significant (P=0.115).
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2026.1982796

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

HELIOS: A spatial omics-informed multi-organ cell atlas enables AI-based tumour microenvironment profiling and biomarker discovery from routine histopathology

Source check: Source checked. Not independent scientific confirmation.

Finding: Spatial-omics cell identities supervise a YOLOv10-m detector for routine H&E; the study reports cross-dataset cell-classification benchmarks and cancer-cohort associations. Prognostic analyses are retrospective and do not establish clinical utility.

Technology: YOLOv10-m detector trained with spatial-omics-derived cell labels

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Spatial transcriptomics labels transferred to matched H&E train a multiclass YOLOv10-m cell detector; the learned operation supplies cell identity and density features for downstream analysis.
Test and result
The paper reports 12.9 million cells from 60 specimens across 14 tissue types and eight cell classes; the held-out evaluation covers about 1.23 million cells, with performance varying by class. It reports cross-dataset F1 values of 0.66 (Lizard), 0.78 (MoNuSAC), and 0.59 (NuCLS), plus retrospective cancer-specific survival associations.
Limitation / still unknown
Methods specify five-fold train/validation/test evaluation and external benchmark comparisons. Prognostic models use retrospective TCGA-type cohorts and adjustment; authors state clinical utility remains unproven and subgroup analyses are exploratory.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.medrxiv.org/content/10.64898/2026.10.26.26366087v1.full.pdf · https://www.medrxiv.org/content/10.64898/2026.10.26.26366087v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Discovering Interpretable Acoustic Features in Autoregressive Voice-Cloning TTS Models

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary v1 full text establishes a probing-versus-intervention dissociation in learned voice-cloning TTS representations.

Technology: Sparse autoencoders, probes and activation interventions in voice-cloning TTS models

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Sparse autoencoders and probes recover acoustic concepts; interventions test causal steering.
Test and result
Recoverable acoustic features do not consistently produce corresponding output changes.
Limitation / still unknown
Multiple TTS models and Qwen3-TTS intervention comparisons on 100 utterances were tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40505v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

MoLRSI: A Recursive Self-Improving AI agent for Molecular Discovery

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The MolRSI agent used learned activity/toxicity predictions to shortlist compounds; 17 of 107 met the authors’ in-vitro activity threshold, with selected hits also tested against resistant strains and human cells.

Technology: MolRSI agent with learned activity and toxicity predictors

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The cited primary sections describe the learned model/AI operation and its relation to the case result.
Limitation / still unknown
The MolRSI agent used learned activity/toxicity predictions to shortlist compounds; 17 of 107 met the authors’ in-vitro activity threshold, with selected hits also tested against resistant strains and human cells.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.39714v1 · https://arxiv.org/abs/2610.39714v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and internal validation of machine-learning models for postoperative deep vein thrombosis after hip arthroplasty for femoral neck fracture: evaluation of Traditional Chinese Medicine syndr

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: retrospective single-center cohort n=380; profile-disjoint holdout n=113/31 DVT events. RF AUC .859 in one held-out partition but mean .792 over 30 repeated constrained partitions; adding two TCM syndrome features did not improve AUC (0.855 vs 0.859, p=.746); TyG association remains observational.

Technology: Random-forest and other machine-learning models for postoperative DVT risk

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Specific postoperative DVT prediction study, but TCM features did not add predictive value; external validation required.
Limitation / still unknown
Methods/Results: retrospective single-center cohort n=380; profile-disjoint holdout n=113/31 DVT events. RF AUC .859 in one held-out partition but mean .792 over 30 repeated constrained partitions; adding two TCM syndrome features did not improve AUC (0.855 vs 0.859, p=.746); TyG association remains observational.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2026.1798230/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Amoxicillin removal by babassu-derived activated carbon under ultrasonic irradiation: adsorption performance and tree-based machine learning modeling.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The original full text supports the bounded AI-enabled finding: Tree-based models analyze laboratory amoxicillin adsorption by babassu-derived activated carbon under ultrasound; Results report removal and adsorption-parameter findings in measured experiments.

Technology: Tree-based machine-learning models for adsorption measurements

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Tree-based models analyze laboratory amoxicillin adsorption by babassu-derived activated carbon under ultrasound; Results report removal and adsorption-parameter findings in measured experiments.
Limitation / still unknown
The reported cohort or experimental dataset, comparator/reference and validation are specific to this case; conclusions remain bounded to that setting.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13769761/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

BABA-Bench: Evaluating Long-Horizon Patient-Advocacy Agents

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: BABA-Bench reports a concrete patient-advocacy agent reliability result: outcome completion substantially exceeds strict evidence-supported route completion, beyond a generic benchmark ranking.

Technology: Learned agents interact with certified patient-advocacy benchmark worlds through the defined interfaces.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned agents interact with certified patient-advocacy benchmark worlds through the defined interfaces.
Limitation / still unknown
Worlds, routes and grading are certified and the result is measured over benchmark runs.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41076v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Training language models in simulation for psychiatric intake

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary v1 full text establishes a blinded simulated psychiatric-intake comparison with independent transcript ratings.

Technology: Multi-turn reinforcement learning adapts Gemini in simulated encounters.

Family Gemini: profile · models page

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Multi-turn reinforcement learning adapts Gemini in simulated encounters.
Limitation / still unknown
120 encounters used 32 psychiatrist actors and 15 independent blinded raters.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40476v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Carbon-Aware Electric Vehicle Routing Problem with Decision-Focused Learning

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Decision-focused learning trained a grid-carbon predictor against route cost. On realistic synthetic German EV-routing instances built from road, charging, weather and electricity data, the resulting routes had 2.30% lower average realized cost than forecast-focused learning and won on 75.5% of instances despite larger pointwise forecast error; this is a simulation result, not deployment evidence.

Technology: The cited primary sections describe the learned model/AI operation and its relation to the case result.

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
The cited primary sections describe the learned model/AI operation and its relation to the case result.
Limitation / still unknown
Ten solver runs per method per instance are rescored against observed grid carbon intensity. The source uses synthetic EVRP instances built from real German data and explicitly does not report deployed vehicle routes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.39645v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Graph neural network modeling of pedestrian space vitality in campus waterfront space: a Flow-Dwell perspective for design translation.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A graph neural network links contextual pedestrian-flow and dwell patterns across 42 spatial units on two campuses to measured waterfront-space vitality, outperforming the reported conventional baselines; the design translation remains a two-campus result.

Technology: GNN message passing models contextual links among 42 pedestrian-space units using Wi-Fi probe-derived Flow/Dwell and spatial attributes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
GNN message passing models contextual links among 42 pedestrian-space units using Wi-Fi probe-derived Flow/Dwell and spatial attributes.
Limitation / still unknown
Comparison against conventional models supports the reported predictive gain; the paper has no out-of-campus replication and the two-campus scope is retained.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/articles/10.3389/fpubh.2026.1963816/pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Framework-Conditional Open Materials Generation

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary v1 full text establishes a framework-conditional generated NdAlSiAu prototype with computational stability support.

Technology: Equivariant framework-conditional model learns species and volume on a fixed framework.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Equivariant framework-conditional model learns species and volume on a fixed framework.
Test and result
StructureMatcher comparison with 124 MP-20 materials and learned-potential stability benchmarks support the prediction.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40443v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Evaluating Conversational Explainability in Healthcare: Comparing the Explainability Assistant and a Traditional XAI Dashboard

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Gemini selected and delivered explanations in a counterbalanced five-person healthcare-background pilot; four participants found the intended local feature in the conversational task, a small human-factors result rather than evidence of clinical benefit.

Technology: The cited primary sections describe the learned model/AI operation and its relation to the case result.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The cited primary sections describe the learned model/AI operation and its relation to the case result.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.39634v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Towards Understanding Score-Seeking in LLMs

Source check: Source checked. Not independent scientific confirmation.

Finding: The controlled study reports a concrete behavioral-science result about score-seeking by language models under staged evaluation conditions, beyond a method or representation benchmark. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: GPT-5.5 and GPT-5.6 Sol are tested as learned agents in controlled environments.

Family GPT-5: profile · models page

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
GPT-5.5 and GPT-5.6 Sol are tested as learned agents in controlled environments.
Test and result
The controlled study reports a concrete behavioral-science result about score-seeking by language models under staged evaluation conditions, beyond a method or representation benchmark.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.41014v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Bridging Neural Representations and Symbolic Semantics in Audio Classification

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary v1 full text establishes a learned audio-classifier and symbolic-induction result on held-out acoustic associations. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Learned classifier activations define response sets and symbolic induction produces acoustic tag combinations.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned classifier activations define response sets and symbolic induction produces acoustic tag combinations.
Test and result
Primary v1 full text establishes a learned audio-classifier and symbolic-induction result on held-out acoustic associations.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40414v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Hoag-type Galaxies: A Photometric Study of the Cores and Rings of Thirty Systems from the DESI Legacy Survey

Source check: Source checked. Not independent scientific confirmation.

Finding: The authors visually inspected machine-learning ring-galaxy catalogues to select 30 Hoag-type systems for photometric analysis. The source catalogues used a modified AlexNet classifier on SDSS images and a two-stage Swin Transformer classifier on DESI images. Among 16 systems with group data, 14 were classified as centrals; the selected sample is small. The result is reported in the arXiv preprint. The study does not establish generalization beyond its reported setting or prospective real-world benefit. Affiliation was not stated in the reviewed material.

Technology: The candidate catalogues use a modified AlexNet classifier on SDSS images and a two-stage Swin Transformer classifier on DESI images; this study visually screened catalogued candidates before its photometric analysis.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The candidate catalogues use a modified AlexNet classifier on SDSS images and a two-stage Swin Transformer classifier on DESI images; this study visually screened catalogued candidates before its photometric analysis.
Test and result
Machine-learning ring-galaxy catalogs supplied candidates that were visually selected before photometric analysis; among 16 systems with group data, 14 were centrals, with the small selected sample limiting generalization.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.39629v1 · https://arxiv.org/abs/2610.39629 · https://arxiv.org/pdf/2610.39629v2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

SIMPLEX: composition-space deep learning for bio-inspired hydrogel adhesion with prospective validation on model-discovered formulations and interpretable composition rules

Source check: Source checked. Not independent scientific confirmation.

Finding: 316 internal formulations plus a 25-formulation Bayesian-optimization-acquired batch; grouped 10-seed/5-fold internal evaluation and same-protocol external batch, with measured hydrogel adhesion; source describes prospective model-selected formulations and composition analysis. Publication venue was not available in the checked primary metadata. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Deep learning selects formulations that are experimentally tested in a shifted acquisition batch

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep learning selects formulations that are experimentally tested in a shifted acquisition batch
Test and result
316 internal formulations plus a 25-formulation Bayesian-optimization-acquired batch; grouped 10-seed/5-fold internal evaluation and same-protocol external batch, with measured hydrogel adhesion; source describes prospective model-selected formulations and composition analysis.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/materials/articles/10.3389/fmats.2026.1966940/full

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Frozen in Time: How a Physics Foundation Model Forgets Under Sim-to-Real Adaptation

Source check: Source checked. Not independent scientific confirmation.

Finding: The v1 Methods and Results directly measure forgetting during simulated-to-laboratory adaptation of Walrus-1.3B and recovery with simulation replay.

Technology: A learned continuum-dynamics foundation model is fine-tuned on simulated and measured cylinder-flow data

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A learned continuum-dynamics foundation model is fine-tuned on simulated and measured cylinder-flow data.
Test and result
The v1 Methods and Results directly measure forgetting during simulated-to-laboratory adaptation of Walrus-1.3B and recovery with simulation replay. Evaluation: The claim is evaluated across held-out pretraining tasks and both simulated and measured flow data.
Limitation / still unknown
The forgetting and replay result is limited to Walrus-1.3B and the reported simulated and measured cylinder-flow adaptation tasks.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40857v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The Athlete as a Scientist of One: Agentic Bayesian Experimental Design for Individualized Physiology from Consumer Wearables

Source check: Source checked. Not independent scientific confirmation.

Finding: A Bayesian wearable-data fit generated prospective predictions for one athlete; two sealed paired comparisons ran opposite to the model’s prediction, while the follow-up still could not identify a stable long-term fitness time constant.

Technology: A Bayesian model fitted wearable measurements to generate prospectively testable predictions about an individual athlete’s physiology.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A Bayesian model fitted wearable measurements to generate prospectively testable predictions about an individual athlete’s physiology.
Test and result
A Bayesian wearable-data fit generated prospective predictions for one athlete; two sealed paired comparisons ran opposite to the model’s prediction, while the follow-up still could not identify a stable long-term fitness time constant. Evaluation: A Bayesian wearable-data fit generated prospective predictions for one athlete; two sealed paired comparisons ran opposite to the model’s prediction, while the follow-up still could not identify a stable long-term fitness time constant.
Limitation / still unknown
The prospective checks involved one athlete and two paired comparisons; neither supported the prediction, and a stable long-term fitness time constant remained unidentified.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40370v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Letting AI Formulate the Research Question: Autonomous Research Agents

Source check: Source checked. Not independent scientific confirmation.

Finding: An AI-agent-run analysis compares annual AlphaEarth/TESSERA reference scores with one-year CONUS tree-cover labels; reference/control corrections did not consistently improve screening, while the supervised classifier scored higher with unequal label resources. The 24 gain positives and partition-sensitive contrasts keep the finding narrow.

Technology: The study analyzes learned AlphaEarth/TESSERA annual embeddings and derived scores; the author repository records that AI agents formulated, conducted and wrote the study

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study analyzes learned AlphaEarth/TESSERA annual embeddings and derived scores; the author repository records that AI agents formulated, conducted and wrote the study.
Test and result
An AI-agent-run analysis compares annual AlphaEarth/TESSERA reference scores with one-year CONUS tree-cover labels; reference/control corrections did not consistently improve screening, while the supervised classifier scored higher with unequal label resources. The 24 gain positives and partition-sensitive contrasts keep the finding narrow. Evaluation: The source evaluates supplied initial classes on held-out one-year LCMAP pairs (109 gains, 337 losses; 24 gain positives) and a separate Annual NLCD Forest test. It reports unequal label resources for the supervised comparator, reversal of all eight correction contrasts under exploratory partition changes, and limitations against multi-year restoration or ecological-recovery claims.
Limitation / still unknown
The result is limited to the reported one-year CONUS screening comparison; partition sensitivity and unequal label resources prevent a multi-year restoration or ecological-recovery claim.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://raw.githubusercontent.com/kofujimura/autonomous-research-agents/1501e89c40d0e700a9d9e7042f6442f8c93e5c5e/runs/run-014/paper.md

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

When precise inference is wrong: Diagnosing model misspecification with applications to cosmic dawn and reionisation

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The learned 21cmEMU emulator enables the paper's repeated inference and internal-consistency analysis of UV luminosity-function data.

Technology: 21cmEMU neural emulator for 21cmFAST cosmic-dawn inference

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The learned 21cmEMU neural emulator replaces repeated 21cmFAST simulations during inference on UV luminosity-function data.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40770v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Morphological Features of Murine Pulmonary Arterial Networks for Diagnosing Vascular Disease

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Boosted feature ranking selected a vascular-feature panel and an SVM tested it on micro-CT data from 13 mice; leave-one-out discrimination was about 69%, close to the all-feature result, so this remains exploratory.

Technology: Boosted feature ranking selects vascular morphology features, and an SVM tests their discrimination on murine pulmonary-artery micro-CT data.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Boosted feature ranking selects vascular morphology features, and an SVM tests their discrimination on murine pulmonary-artery micro-CT data.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40344v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Hidden Phase Transition in Single-Polymer Stretching under Elongational Flow

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Jointly learned neural potentials locate a distinct transition in the modeled polymer dynamics; single-molecule imaging separately measured a qualitative response reversal, while not validating the exact learned threshold.

Technology: Jointly learned neural potentials for polymer dynamics

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Jointly learned neural potentials model polymer dynamics and identify a transition in the simulated stretching behavior.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.39575v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Unveiling Sediment Deposition Thresholds in Discontinuous Emergent Vegetation Patches Using a SHAP-Interpretable Ensemble Machine Learning

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Using 498 observations from 11 laboratory flume experiments, ensemble ML was evaluated with an unseen experiment holdout; CatBoost reached R²=.953 and SHAP identified an approximately 0.3 m cumulative vegetation-exposure threshold associated with deposition-regime change. This is a testable physical hypothesis from a small laboratory dataset, not causal confirmation; independent experiments are needed.

Technology: CatBoost ensemble and SHAP analysis of laboratory flume sediment deposition

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
CatBoost ensemble analysis and SHAP interpret flume observations to estimate sediment-deposition thresholds in discontinuous vegetation patches.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-11485176/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Comparison of anomaly detection and supervised learning for acute myeloid leukemia blast identification: A proof-of-concept study

Source check: Source checked. Not independent scientific confirmation.

Finding: This retrospective AML flow-cytometry study compares supervised, anomaly-detection and clustering approaches for blast identification and reports markedly different performance by prevalence, including two independent external validation samples. It qualifies narrowly as a clinical diagnostic-validation result.

Technology: Four anomaly detectors, Random Forest and UMAP-HDBSCAN are compared for blast identification using flow-cytometry features.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Four anomaly detectors, Random Forest and UMAP-HDBSCAN are compared for blast identification using flow-cytometry features.
Test and result
External check uses two independent samples only; results support a small proof-of-concept diagnostic comparison, not general clinical performance or clinical benefit.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://export.arxiv.org/api/query?id_list=2610.39447 · https://arxiv.org/pdf/2610.39447

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

More Rules, Longer Lists? Benchmarking Self-Evolving Agents on a Hidden-Rule Curriculum

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary v1 full text establishes a benchmark-specific hidden-rule state-representation result.

Technology: Agents revise persistent state across hidden-rule curriculum exposures.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Agents revise persistent state across hidden-rule curriculum exposures.
Test and result
Twenty chains across ten domains compare no-update, history, and convention conditions.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40758v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Unweaving the mechanics of woven fabrics through automated model discovery

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A constitutive neural network selected energy terms from measured biaxial stress–stretch curves; sparse regression recovered the dominant terms, supporting a specific three-term model for the tested coated fabric.

Technology: A constitutive neural network selects strain-energy terms from measured biaxial stress–stretch curves; sparse regression identifies the compact constitutive model.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A constitutive neural network selects strain-energy terms from measured biaxial stress–stretch curves; sparse regression identifies the compact constitutive model.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40260v1 · https://arxiv.org/abs/2610.40260v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

COGS: A Diagnostic Benchmark for Attribute-Aware Object-Level 3D Gaussian Splatting Perception

Source check: Source checked. Not independent scientific confirmation.

Finding: On capture-derived MACGS, all six conventional backbones have an attribute subset outperforming full POSQC by 0.28–3.05 points, while both HALO variants do best with full inputs. Six conventional classifiers and two HALO variants are evaluated on COGS and capture-derived MACGS; subset selection and dataset-specific architecture effects remain bounded.

Technology: Six conventional 3DGS classifiers and two HALO variants are trained on matched Gaussian attribute subsets and evaluated on COGS/MACGS.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Six conventional 3DGS classifiers and two HALO variants are trained on matched Gaussian attribute subsets and evaluated on COGS/MACGS.
Test and result
Six conventional classifiers and two HALO variants are evaluated on COGS and capture-derived MACGS; subset selection and dataset-specific architecture effects remain bounded.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.39201v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning predicts IDH1 mutation status in acute myeloid leukemia from bone marrow smears.

Source check: Source checked. Not independent scientific confirmation.

Finding: Attention-based multiple-instance learning predicts IDH1 mutation from bone-marrow smears in 436 AML patients across four centers with pooled ROC-AUC 0.81 and a clinically specific retrospective evaluation. Published in npj Precision Oncology.

Technology: Attention-based multiple-instance learning predicts IDH1 mutation status from bone-marrow smear images.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Attention-based multiple-instance learning predicts IDH1 mutation status from bone-marrow smear images.
Test and result
Full text states the cohort/data, comparison and evaluation for this bounded claim.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41698-026-01762-x

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

PedDx: An End-to-End Deep Learning Framework for Pediatric Posterior Fossa Tumor Diagnosis with Agentic AI–Enabled Independent External Evaluation

Source check: Source checked. Not independent scientific confirmation.

Finding: Results report five-class accuracy of 81.4% in the external CHOA cohort (88.7% among non-abstained cases) and 92.5% across 40 independently processed WCH cases (97.3% among non-abstained cases); WCH contained only three histologies, and the preprint also reports confidently incorrect OOD cases. Methods describe a frozen pretrained 3DINO backbone, multi-crop MRI sampling, calibrated class probabilities and confidence-based abstention; the deployed workflow also includes a quality-control agent. Testing: Methods/results distinguish discovery (n=492), replication (n=114), external CHOA (n=129) and independent WCH workflow cases (n=40), with pathology/clinical diagnoses and stated exclusions. The independent cohorts support the narrow diagnostic-performance result; small WCH sample, missing two histologies, unreviewed preprint and no evidence of improved patient outcomes limit interpretation.. Independent external evaluation of a pediatric posterior-fossa tumor MRI classifier and full workflow reports performance on separate CHOA and WCH cohorts; bounded to diagnostic classification, not clinical benefit. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Methods describe a frozen pretrained 3DINO backbone, multi-crop MRI sampling, calibrated class probabilities and confidence-based abstention; the deployed workflow also includes a quality-control agent.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Methods describe a frozen pretrained 3DINO backbone, multi-crop MRI sampling, calibrated class probabilities and confidence-based abstention; the deployed workflow also includes a quality-control agent.
Test and result
Methods/results distinguish discovery (n=492), replication (n=114), external CHOA (n=129) and independent WCH workflow cases (n=40), with pathology/clinical diagnoses and stated exclusions. The independent cohorts support the narrow diagnostic-performance result; small WCH sample, missing two histologies, unreviewed preprint and no evidence of improved patient outcomes limit interpretation.
Limitation / still unknown
Independent external evaluation of a pediatric posterior-fossa tumor MRI classifier and full workflow reports performance on separate CHOA and WCH cohorts; bounded to diagnostic classification, not clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.medrxiv.org/content/10.64898/2026.10.27.26366616v1.full.pdf · https://api.biorxiv.org/details/medrxiv/2026-10-29/2026-10-29/0

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AutoHebb: Self-Modifying Neural Networks that Zero-Shot Infer and Continually Learn

Source check: Source checked. Not independent scientific confirmation.

Finding: Networks rank unseen item pairs and link previously separate rankings. Meta-learning fits fixed Hebbian parameters; weight analysis exposes task algorithms. Testing: Held-out item pairs and minimal-network comparisons test the inferred mechanisms.. Primary v1 full text establishes zero-shot ranking and list-linking behavior in meta-learned Hebbian networks. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Meta-learning fits fixed Hebbian parameters; weight analysis exposes task algorithms.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Meta-learning fits fixed Hebbian parameters; weight analysis exposes task algorithms.
Test and result
Held-out item pairs and minimal-network comparisons test the inferred mechanisms.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40757v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

NeurEO: dissecting Earth observation embeddings with computational neuroscience

Source check: Source checked. Not independent scientific confirmation.

Finding: AlphaEarth embedding features were analyzed against land-cover and elevation maps; one feature showed directional woodland-edge and water associations across 100 separated UK locations, a small correlational proof of concept. The cited primary sections describe the learned model/AI operation and its relation to the case result. Testing: AlphaEarth embedding features were analyzed against land-cover and elevation maps; one feature showed directional woodland-edge and water associations across 100 separated UK locations, a small correlational proof of concept.. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: The cited primary sections describe the learned model/AI operation and its relation to the case result.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The cited primary sections describe the learned model/AI operation and its relation to the case result.
Test and result
AlphaEarth embedding features were analyzed against land-cover and elevation maps; one feature showed directional woodland-edge and water associations across 100 separated UK locations, a small correlational proof of concept.
Limitation / still unknown
AlphaEarth embedding features were analyzed against land-cover and elevation maps; one feature showed directional woodland-edge and water associations across 100 separated UK locations, a small correlational proof of concept.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2610.40137v1 · https://arxiv.org/abs/2610.40137v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Reasoning or Pattern Exploitation? Mechanistic Insights from RL-Trained Language Models

Source check: Source checked. Not independent scientific confirmation.

Finding: In 100 paired boards, patching layer 25 redirects predictions to the donor-optimal move 66 times versus 10 unpatched and 12 unrelated-donor controls. A GRPO-trained Qwen2.5-1.5B model is analyzed with sparse autoencoders and activation-patching interventions on Tic-Tac-Toe boards. Testing: Paired board interventions use unpatched, unrelated-board and earlier-layer controls; result is limited to one small model and Tic-Tac-Toe.. Paired board interventions use unpatched, unrelated-board and earlier-layer controls; result is limited to one small model and Tic-Tac-Toe. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: A GRPO-trained Qwen2.5-1.5B model is analyzed with sparse autoencoders and activation-patching interventions on Tic-Tac-Toe boards.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A GRPO-trained Qwen2.5-1.5B model is analyzed with sparse autoencoders and activation-patching interventions on Tic-Tac-Toe boards.
Test and result
Paired board interventions use unpatched, unrelated-board and earlier-layer controls; result is limited to one small model and Tic-Tac-Toe.
Limitation / still unknown
In 100 paired boards, patching layer 25 redirects predictions to the donor-optimal move 66 times versus 10 unpatched and 12 unrelated-donor controls. Paired board interventions use unpatched, unrelated-board and earlier-layer controls; result is limited to one small model and Tic-Tac-Toe.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/39182v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Validation of a deep neural network workflow for fully automated dicentric chromosome analysis in biological dosimetry

Source check: Source checked. Not independent scientific confirmation.

Finding: Results report the case-specific result: Unverified Haiku assessment. A deep neural network workflow for dicentric chromosome detection is validated with threshold optimisation, calibration curves and independent testing in biological dosimetry, with the AI role in scoring explicit. Methods establish the case-specific AI operation: Unverified Haiku assessment. Testing: Full text states the cohort/data, comparison and evaluation for this bounded claim.. The primary publisher Methods/Results support the AI-enabled result: Unverified Haiku assessment. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Methods establish the case-specific AI operation: Unverified Haiku assessment. A deep neural network workflow for dicentric chromosome detection is validated with threshold optimisation, calibration curves and independent testing in biological dosimetry, with the AI role in scoring explicit.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Methods establish the case-specific AI operation: Unverified Haiku assessment. A deep neural network workflow for dicentric chromosome detection is validated with threshold optimisation, calibration curves and independent testing in biological dosimetry, with the AI role in scoring explicit.
Test and result
Full text states the cohort/data, comparison and evaluation for this bounded claim.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-026-77285-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Multi-Layer Navigation System for Smart Wheelchairs with a Limited Input Interface

Source check: Source checked. Not independent scientific confirmation.

Finding: In 360 trials with ten participants without disabilities, the full system completed 99.72% of tasks; no isolated causal effect of vision or clinical benefit is established. The real wheelchair study covers 360 trials by ten nondisabled participants; system-level result is confounded by differing priority policy and is not a clinical evaluation.

Technology: YOLO-World, MobileSAM and learned target selection in a wheelchair navigation system

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
YOLO-World and MobileSAM provide open-vocabulary target detection/segmentation in a wheelchair navigation stack with learned target selection.
Test and result
In 360 trials with ten participants without disabilities, the full system completed 99.72% of tasks; no isolated causal effect of vision or clinical benefit is established.
Limitation / still unknown
The real wheelchair study covers 360 trials by ten nondisabled participants; system-level result is confounded by differing priority policy and is not a clinical evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/38569v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Strain-Induced BCC-to-FCC Transformation in Tungsten Uncovered by Molecular Dynamics Simulations with Machine Learning Potentials

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: DFT-informed machine-learning-potential molecular dynamics predicts a previously unreported strain-induced BCC→FCC pathway in tungsten, with a temperature/dislocation-dependent FCC fraction and structural pathway. This is a concrete computational materials result; experimental confirmation remains future work.

Technology: The authors use a neural-evolution-potential machine-learning interatomic potential, trained/checked against DFT structures, to run GPUMD molecular dynamics under tensile strain.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The authors use a neural-evolution-potential machine-learning interatomic potential, trained/checked against DFT structures, to run GPUMD molecular dynamics under tensile strain.
Limitation / still unknown
The potential is checked against DFT reference properties and the predicted phase pathway is structurally analyzed. Validation is computational; no experimental observation is claimed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/pdf/2610.38477

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Exploring Fractional Quantum Hall States with Deep Learning: From Abelian to Non-Abelian Phases

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Finite-size overlaps and entanglement spectra are consistent with candidate parton phases at nu=4/3 and 7/5 and an anti-Pfaffian-like structure at nu=3/2. Numerical ground-state results use finite systems and trial-state overlaps/entanglement-spectrum checks; no experimental measurement is claimed.

Technology: A symmetry-adapted neural-network wavefunction is optimized by variational Monte Carlo for interacting electrons across Landau-level fillings.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A symmetry-adapted neural-network wavefunction is optimized by variational Monte Carlo for interacting electrons across Landau-level fillings.
Limitation / still unknown
Numerical ground-state results use finite systems and trial-state overlaps/entanglement-spectrum checks; no experimental measurement is claimed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/38533v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Integrating Machine Learning and In-Flight Particle Diagnostics to Enhance Coating Quality in Atmospheric Plasma Spraying

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: FFNN models trained on 300 in-flight particle measurements from 12 experiments yield a coating-quality score; learned model enables the result.

Technology: FFNN models trained on 300 in-flight particle measurements from 12 experiments yield a coating-quality score; learned model enables the result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
FFNN models trained on 300 in-flight particle measurements from 12 experiments yield a coating-quality score; learned model enables the result.
Limitation / still unknown
The source reports the defined cohort/experiment and its reference or test comparison; interpretation is restricted to that study design.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11666-026-02357-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Closed Gates Leak: Decomposing Adjacent-Layer Cancellation in Gated Language Models

Source check: Source checked. Not independent scientific confirmation.

Finding: Across ten gated models closed-gate neurons carry most of the average damping; annealing their negative activation side removes damping/cancellation in tested continued training. Sixteen language models are analyzed, including ten gated models; counterfactual and continued-training controls test mechanism, with limited small-model interventions.

Technology: Controlled decomposition of activations in gated transformer MLPs attributes adjacent-layer damping to closed-gate neurons; gate-annealing interventions test the mechanism.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Controlled decomposition of activations in gated transformer MLPs attributes adjacent-layer damping to closed-gate neurons; gate-annealing interventions test the mechanism.
Test and result
Sixteen language models are analyzed, including ten gated models; counterfactual and continued-training controls test mechanism, with limited small-model interventions.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/39068v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Similar Scores, Opposite Failures: Human Correction as an Instrument for Video Caption Evaluation

Source check: Source checked. Not independent scientific confirmation.

Finding: Reference-overlap metrics poorly distinguish accepted from rejected captions (BLEU-4 AUC 0.510); window-level review detects more camera/shot errors than overview review. Six models, 196 clips, and human post-edits under overview/window protocols provide the reference; one annotator per caption and this corpus/prompt bound generalization.

Technology: Six frontier video-captioning models generate captions; trained annotators post-edit them against a predefined error taxonomy.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Six frontier video-captioning models generate captions; trained annotators post-edit them against a predefined error taxonomy.
Test and result
Six models, 196 clips, and human post-edits under overview/window protocols provide the reference; one annotator per caption and this corpus/prompt bound generalization.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/38511v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Controlling Continuum Robot Shape in Enclosed Space with Embedded Sensing using IMU and Actuator Feedback

Source check: Source checked. Not independent scientific confirmation.

Finding: Physical closed-loop trials report 6.1%L mean tip error with learned feedback and 2.3%L global error in three-segment control; embedded feedback enables operation without external observation. Physical tests compare MoCap, MLP and LSTM feedback on one- and three-segment robots; hardware-specific measures, including enclosed-space trials, bound the claim. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Neural networks reconstruct continuum-robot shape from embedded IMU and actuator feedback for real-time physical closed-loop control.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Neural networks reconstruct continuum-robot shape from embedded IMU and actuator feedback for real-time physical closed-loop control.
Test and result
Physical closed-loop trials report 6.1%L mean tip error with learned feedback and 2.3%L global error in three-segment control; embedded feedback enables operation without external observation. Physical tests compare MoCap, MLP and LSTM feedback on one- and three-segment robots; hardware-specific measures, including enclosed-space trials, bound the claim.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/39055v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Can LLMs Faithfully Solve Ontology Engineering Tasks? A Mechanistic Interpretability Study

Source check: Source checked. Not independent scientific confirmation.

Finding: Distributed subspace interventions achieve the strongest ontology-task results but do not fully separate target relations from auxiliary class information. Matched interchange interventions are evaluated on two ontology tasks and Qwen 4B/8B; partial disentanglement is explicitly retained as a limitation. The result is reported in the arXiv preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Qwen3-4B and Qwen3-8B are tested with SAE, DBM and DAS interchange interventions on entity-typing and subsumption tasks.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Qwen3-4B and Qwen3-8B are tested with SAE, DBM and DAS interchange interventions on entity-typing and subsumption tasks.
Test and result
Distributed subspace interventions achieve the strongest ontology-task results but do not fully separate target relations from auxiliary class information. Matched interchange interventions are evaluated on two ontology tasks and Qwen 4B/8B; partial disentanglement is explicitly retained as a limitation.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/38326v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Representation Complexity Along Learning Trajectories: Trading Width for Precision and Time

Source check: Source checked. Not independent scientific confirmation.

Finding: The proof establishes a width/discovery gap for sparse parity under stated oracle and resolution assumptions; experiments examine dimension dependence and trajectory behavior. Formal theorem/proof and sparse-parity experiments support a theoretical result under explicit oracle, smoothness and resolution assumptions, not a universal empirical law.

Technology: Theoretical analysis derives effective-width bounds for gradient-query learning, with sparse-parity simulations examining the predicted dimension dependence

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Theoretical analysis derives effective-width bounds for gradient-query learning, with sparse-parity simulations examining the predicted dimension dependence.
Test and result
The proof establishes a width/discovery gap for sparse parity under stated oracle and resolution assumptions; experiments examine dimension dependence and trajectory behavior. Formal theorem/proof and sparse-parity experiments support a theoretical result under explicit oracle, smoothness and resolution assumptions, not a universal empirical law. Evaluation: Formal theorem/proof and sparse-parity experiments support a theoretical result under explicit oracle, smoothness and resolution assumptions, not a universal empirical law.
Limitation / still unknown
The theoretical result depends on the stated oracle, smoothness and resolution assumptions; it is not a universal empirical law.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/39027v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Hidden Commitment: When Language Models Silently Pick a Side and How Steering Can Surface It

Source check: Source checked. Not independent scientific confirmation.

Finding: In held-out conflicting-document rows, steering the Qwen source-choice direction shifts which document supplies the answer; Gemma provides a weaker transfer check. The intervention test uses 100 held-out question IDs in original/swapped order (200 rows), with random-direction controls; small-model, controlled-conflict scope.

Technology: Qwen3.5-0.8B and Gemma 3 1B supply learned residual states; readouts and controlled steering test how internal activations select between conflicting documents

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Qwen3.5-0.8B and Gemma 3 1B supply learned residual states; readouts and controlled steering test how internal activations select between conflicting documents.
Test and result
In held-out conflicting-document rows, steering the Qwen source-choice direction shifts which document supplies the answer; Gemma provides a weaker transfer check. The intervention test uses 100 held-out question IDs in original/swapped order (200 rows), with random-direction controls; small-model, controlled-conflict scope. Evaluation: The intervention test uses 100 held-out question IDs in original/swapped order (200 rows), with random-direction controls; small-model, controlled-conflict scope.
Limitation / still unknown
The intervention was tested on small models in a controlled conflicting-document task; broader source-selection behavior is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/38296v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The Search for Statistical Indistinguishability Between Real and Fake Images

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Across MNIST, CelebA, FFHQ and GenImage, trained classifiers generally still separate real and generated images after tested abstractions; chance cases require severe information loss. Four named datasets and multiple generator/classifier settings are tested; failure to distinguish after severe abstraction is not evidence of distributional identity generally.

Technology: Deep neural classifiers are trained to distinguish real from generated images across datasets and progressively abstracted representations.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep neural classifiers are trained to distinguish real from generated images across datasets and progressively abstracted representations.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/39014v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

EPSBench: Benchmarking LLM Agents for Earnings Forecasting

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Across the 149 firm-quarter cases, nominal 80% intervals cover reported EPS in 37–70% of returning runs, with most misses above the interval. Six models are compared with reported GAAP EPS, analyst consensus and interval coverage across 149 quarters from 86 US companies; retrospective evidence only.

Technology: Six LLM agents extract income-statement drivers from public reports; a shared accounting calculator generates earnings forecasts and intervals.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Six LLM agents extract income-statement drivers from public reports; a shared accounting calculator generates earnings forecasts and intervals.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/38293v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Why Speculative Decoding Works Better Than Predicted on Sparse MoE Models

Source check: Source checked. Not independent scientific confirmation.

Finding: Routing overlap and GPU execution explain why measured speculative-decoding cost departs from activated-expert-byte estimates; an IPF model predicts unique experts with under 4% error. Ten sparse MoE models (30B–1T) plus H100 profiling and controlled synthetic routing support the mechanism; performance is hardware/workload specific.

Technology: The study measures sparse MoE routing, speculative decoding, cache reuse and GPU execution across ten learned language models.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study measures sparse MoE routing, speculative decoding, cache reuse and GPU execution across ten learned language models.
Test and result
Ten sparse MoE models (30B–1T) plus H100 profiling and controlled synthetic routing support the mechanism; performance is hardware/workload specific.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/38981v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The Role of Quantization in Loss Spikes

Source check: Source checked. Not independent scientific confirmation.

Finding: For 28M-parameter models, BF16-true runs spike earliest, mixed BF16 later, and FP32 stays stable over five seeds; precision-error growth accompanies the spikes. Five-seed comparisons at fixed 28M scale and long training horizon compare BF16 true, mixed BF16 and FP32; limits are the tested model/training regime.

Technology: A GPT-2-style language model is pretrained while output-head precision is varied; same-state FP32 logits provide a reference for quantization error.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A GPT-2-style language model is pretrained while output-head precision is varied; same-state FP32 logits provide a reference for quantization error.
Test and result
Five-seed comparisons at fixed 28M scale and long training horizon compare BF16 true, mixed BF16 and FP32; limits are the tested model/training regime.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/38277v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Interventional Training Reduces Causal Bypass in In-Context Introspection

Source check: Source checked. Not independent scientific confirmation.

Finding: Straight SFT can achieve endogenous signal-reporting performance while failing interventional checks; counterfactual intervention training improves causal introspection, especially OOD. Models are trained to introspect neuron, attention and sparse-autoencoder signals; intervention training uses counterfactually perturbed signals. Testing: Hundreds of introspectors across eight language models and three signal types are compared under endogenous and intervened/OOD conditions; generalization to unseen sites is limited.. Hundreds of introspectors across eight language models and three signal types are compared under endogenous and intervened/OOD conditions; generalization to unseen sites is limited. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Models are trained to introspect neuron, attention and sparse-autoencoder signals; intervention training uses counterfactually perturbed signals.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Models are trained to introspect neuron, attention and sparse-autoencoder signals; intervention training uses counterfactually perturbed signals.
Test and result
Hundreds of introspectors across eight language models and three signal types are compared under endogenous and intervened/OOD conditions; generalization to unseen sites is limited.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/38911v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The First Deep Learning all-sky List of Fermi-LAT Gamma-ray Sources

Source check: Source checked. Not independent scientific confirmation.

Finding: The U-Net/DNN pipeline produces a 3,800+ source Fermi-LAT all-sky catalog; the paper reports a few additional sources subsequently confirmed in FL16Y, a concrete astronomy result enabled by learned detection and characterization.

Technology: U-Net segmentation plus DNN source localization, flux estimation and genuine/spurious classification produce the source list.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
U-Net segmentation plus DNN source localization, flux estimation and genuine/spurious classification produce the source list.
Test and result
Pipeline is assessed on simulations with known truth, tested against interstellar-emission-model variation, compared to 4FGL-DR2, and the cited follow-up list confirms a few new sources. No independent claim beyond those reported.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/pdf/2610.37287

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Explainable deep learning for predicting temperature breaks and identifying their causal drivers in fresh produce cold chains.

Source check: Source checked. Not independent scientific confirmation.

Finding: two IoT datasets from South African apple cold storage, eight ML models and three baselines; LSTM-GRU best recurrent model, DNN performance improved with extended differing conditions, fewer false alarms/day, all DNNs flagged breaks in advance; LPCMCI identified lagged temperature as dominant. Authors note few recovered causal links and unmeasured drivers. Publication venue was not available in the checked primary metadata. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Field-sensor prediction of produce-storage temperature breaks is a concrete operational result

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Field-sensor prediction of produce-storage temperature breaks is a concrete operational result
Test and result
two IoT datasets from South African apple cold storage, eight ML models and three baselines; LSTM-GRU best recurrent model, DNN performance improved with extended differing conditions, fewer false alarms/day, all DNNs flagged breaks in advance; LPCMCI identified lagged temperature as dominant. Authors note few recovered causal links and unmeasured drivers.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2026.1893108/full

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Estimating Functional Independence Measure Scores From Japanese Rehabilitation Notes Using a Large Language Model: Retrospective Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: An LLM estimated item-level Functional Independence Measure scores from Japanese rehabilitation notes and was compared with clinical reference assessments across disease cohorts. Published in JMIR Medical Informatics.

Technology: Methods describe the note sources and score estimation; Results report errors by therapist-note type, disease and admission-score strata.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe the note sources and score estimation; Results report errors by therapist-note type, disease and admission-score strata.
Test and result
Methods describe the note sources and score estimation; Results report errors by therapist-note type, disease and admission-score strata.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13752589/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Early Unplanned Readmission to Critical Care Units Using Machine Learning and Metaheuristic Hyperparameter Optimization: Evidence from a Chilean Public Hospital.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Five ML models assessed seven-day unplanned critical-care readmission in 371 patients, with nested stratified five-fold cross-validation and fold-contained preprocessing. Logistic regression had highest AUC (75.2%) and XGBoost 75.0%; case-enriched single-center data require recalibration and external validation.

Technology: 371 Chilean public-hospital ICU discharges with 41 readmissions; logistic regression, SVC, tree, random forest and XGBoost trained from transfer-time variables. Nested stratified five-fold CV; preprocessing fit within training folds. AUC 53.3–75.2%; LR highest 75.2%, XGBoost 75.0%; case-enriched cohort.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
371 Chilean public-hospital ICU discharges with 41 readmissions; logistic regression, SVC, tree, random forest and XGBoost trained from transfer-time variables. Nested stratified five-fold CV; preprocessing fit within training folds. AUC 53.3–75.2%; LR highest 75.2%, XGBoost 75.0%; case-enriched cohort.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcm/jcm-15-08328/article_deploy/jcm-15-08328.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AI-assisted adaptive-optical robotic metrology for sub-millimeter localization and dimensional measurement of blind holes.

Source check: Source checked. Not independent scientific confirmation.

Finding: A six-axis robot with AI image segmentation performs adaptive-optical blind-hole localization and dimensional measurement; comparison with CMM references across 30 physical samples reports sub-millimeter errors and repeatability, as a proof-of-concept metrology result. Publication venue was not available in the checked primary metadata. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: AI segmentation estimates the hole geometry

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
AI segmentation estimates the hole geometry
Test and result
A six-axis robot with AI image segmentation performs adaptive-optical blind-hole localization and dimensional measurement; comparison with CMM references across 30 physical samples reports sub-millimeter errors and repeatability, as a proof-of-concept metrology result.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/articles/10.3389/frai.2026.1983856/pdf · https://doi.org/10.3389/frai.2026.1983856

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development of an automated machine learning-based prognostic prediction model for emergency septic shock patients.

Source check: Source checked. Not independent scientific confirmation.

Finding: In a retrospective single-center cohort of 913 emergency septic-shock patients, an Improved Red Fox Optimization algorithm tuned an AutoML model for 28-day mortality. On the held-out 183-patient test set, the model had ROC-AUC 0.891 and PR-AUC 0.879 and outperformed SOFA, APACHE II and qSOFA in the reported comparison. This internal evaluation does not establish prospective clinical benefit.

Technology: A learned model predicted 28-day mortality in a defined emergency septic-shock patient cohort and was compared with clinical scores.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A learned model predicted 28-day mortality in a defined emergency septic-shock patient cohort and was compared with clinical scores.
Test and result
The primary abstract reports 913 patients from one tertiary hospital, split into 730 training and 183 test records; an IRFO-optimized AutoML model achieved test ROC-AUC 0.8911 and PR-AUC 0.8790. The source also reports comparison with SOFA, APACHE II and qSOFA, plus decision-curve analysis. The evidence is retrospective and single-center, without external or prospective workflow validation.
Limitation / still unknown
This retrospective single-center holdout does not establish external generalization or prospective clinical benefit; decision-curve analysis is not an observed improvement in patient outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13745730/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Longitudinal DLQI trajectory patterns and early risk updating in telepharmacy-monitored psoriasis.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Ward clustering of psoriasis quality-of-life trajectories identified three exploratory patterns, including a persistent-burden group; for remission, month-one DLQI updating improved AUROC to 0.753 with calibration, while baseline DLQI alone matched/outperformed ensembles for improvement. This is a retrospective telepharmacy cohort result and shows limited incremental value of complex ML.

Technology: Ward clustering identifies DLQI trajectory groups; repeated-split machine-learning models update psoriasis-remission risk from month-one scores.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Ward clustering identifies DLQI trajectory groups; repeated-split machine-learning models update psoriasis-remission risk from month-one scores.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2026.1961097/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Steric-Pressure-Driven Topological Reconstruction for High-Performance Zintl Phase Thermoelectrics with Machine-Learning Validation.

Source check: Source checked. Not independent scientific confirmation.

Finding: Wiley full-text Methods/Results excerpts identify the experimental material and measured thermoelectric outcome and the CrabNet/XGBoost validation; exact validation protocol remains bounded by available text. Testing: The model operation, experimental/computational cohort, comparison or assay, and reported outcome were read in the primary full text; interpretation is limited to the demonstrated result.. The study reports a synthesized and measured Zintl-phase reconstruction with improved thermoelectric performance; ML validation ranks a specific materials region and contributes to the result. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Wiley full-text Methods/Results excerpts identify the experimental material and measured thermoelectric outcome and the CrabNet/XGBoost validation; exact validation protocol remains bounded by available text.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Wiley full-text Methods/Results excerpts identify the experimental material and measured thermoelectric outcome and the CrabNet/XGBoost validation; exact validation protocol remains bounded by available text.
Test and result
The model operation, experimental/computational cohort, comparison or assay, and reported outcome were read in the primary full text; interpretation is limited to the demonstrated result.
Limitation / still unknown
The study reports a synthesized and measured Zintl-phase reconstruction with improved thermoelectric performance; ML validation ranks a specific materials region and contributes to the result.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://onlinelibrary.wiley.com/doi/10.1002/advs.78519

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Alignment-free discovery of conserved RNA structures in the human genome

Source check: Source checked. Not independent scientific confirmation.

Finding: RLMHelix combines RNA-language-model dependency maps with thermodynamic pairing probabilities to rank candidate conserved structures. Its genome scan reported 15,821 candidates, two-thirds absent from four earlier screens. The paper compared predicted structures with existing public icSHAPE datasets and sequence-alignment evidence; these retrospective analyses and exploratory pull-downs do not establish prospective or separate-team validation.

Technology: RNA language model RLMHelix with thermodynamic structure scoring

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
RLMHelix combines RNA-language-model dependency maps with thermodynamic base-pair probabilities, scores helices in single sequences, then tests candidates against alignments; distillation enables genome-scale scans
Test and result
The preprint compared predicted structures against public icSHAPE datasets GSE145805 and GSE74353, including measured RNA cores from five cell lines, and tested conservation/covariation using retrieved alignments. These are existing-data evaluations; pull-down experiments are described as exploratory.
Limitation / still unknown
The checked source does not report prospective testing before new outcomes or a separate research team's follow-up confirmation; public probing datasets and same-paper exploratory assays support retrospective/benchmark evidence only.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.64898/2026.10.24.761493 · https://api.crossref.org/works/10.64898/2026.10.24.761493 · https://www.biorxiv.org/content/10.64898/2026.10.24.761493v1.full · https://www.biorxiv.org/content/10.64898/2026.10.24.761493v1.full.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Viral host assignment across the tree of life with VHAMSTeR

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: VHAMSTeR yields specific host-assignment and metagenomic ecological findings from a sequence model; the predicted host associations remain computational.

Technology: a fine-tuned genomic language model and biological genome/homology features feed separate classifiers, combined by a context-aware gate for five-category host assignment.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
a fine-tuned genomic language model and biological genome/homology features feed separate classifiers, combined by a context-aware gate for five-category host assignment.
Limitation / still unknown
Methods/Results: trained on 42,512 viral sequences; 5-fold cross-validation clustered by genome identity to limit leakage, plus comparisons with prior tools. Ecological host links are computational predictions and are not in-source experimental confirmation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.64898/2026.10.23.761443

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Build, Don't Play: Self-Improving LLM Research Agents Develop a Hearthstone Battlegrounds AI That Rivals Top Players

Source check: Source checked. Not independent scientific confirmation.

Finding: Across 392 live ladder games, the account rating rose from 6,000 to 14,007; the paper separates 96 client failures from agent-play performance. The result is a single study account on a live game ladder; leader-board estimates and client failures are explicitly outside the bounded policy-performance claim.

Technology: LLM research agents improve the game simulator/opponents; a transformer policy trained with PPO chooses play actions.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
LLM research agents improve the game simulator/opponents; a transformer policy trained with PPO chooses play actions.
Test and result
The result is a single study account on a live game ladder; leader-board estimates and client failures are explicitly outside the bounded policy-performance claim.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/33845v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Sequence-diversity-guided discovery, model-guided mutagenesis, and secretory production of GH5 endoglucanases in Bacillus subtilis

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Sequence-model-guided selection plus direct biochemical assays establishes previously uncharacterized active GH5 enzymes and production result. Explicitly preserve imperfect predictions, purified/flask scale and no industrial deployment claim.

Technology: CatPred sequence model for screening GH5 endoglucanases and guiding mutagenesis

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
CatPred sequence-model screening prioritizes GH5 enzymes; selected candidates undergo mutagenesis, secretion and biochemical activity testing.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13068-026-02838-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and temporal validation of an explainable machine learning model for predicting in-hospital diabetic foot ulcer in patients with type 2 diabetes using inflammatory–nutritional composite in

Source check: Source checked. Not independent scientific confirmation.

Finding: The Frontiers in Medicine study develops explainable ML models for in-hospital diabetic foot ulcer in a 550-patient cohort with a temporal validation split, a clinically specific retrospective assessment, though the validation cohort has only 18 events.

Technology: Logistic regression, random forest, LightGBM and XGBoost models used routinely available admission variables and inflammatory–nutritional indices to estimate in-hospital diabetic-foot-ulcer risk; LASSO selected predictors and SHAP interpreted the models.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Logistic regression, random forest, LightGBM and XGBoost models used routinely available admission variables and inflammatory–nutritional indices to estimate in-hospital diabetic-foot-ulcer risk; LASSO selected predictors and SHAP interpreted the models.
Test and result
In 550 hospitalized patients with type 2 diabetes (400 development, 150 temporal validation), logistic regression had temporal-validation AUC 0.976, but calibration was poor (Hosmer–Lemeshow P=.001); LightGBM/XGBoost sensitivities were 0.056/0.389 at prespecified thresholds.
Limitation / still unknown
The temporal validation set contained 18 events; low sensitivity and calibration concerns preclude screening use. Multicenter prospective validation is still needed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2026.1938429/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Clinical Risk Models, Inflammatory-Nutritional Indices and a Neural-Network Survival Model in Metastatic Renal Cell Carcinoma: A Single-Center Retrospective Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Neural network predicts death risk and survival time; predictors screened/tuned within repeated stratified 5-fold CV (50 outer folds), compared on identical splits with Cox and random survival forest. n=96, C-index 0.685 vs IMDC+nephrectomy Cox 0.692 and RSF 0.695; worse Brier score/calibration slope 0.54. Testing: A dual-head neural-network survival model was evaluated against IMDC/MSKCC and Cox/forest benchmarks in a 96-patient retrospective mRCC cohort using repeated nested cross-validation. Its C-index was comparable to Cox and forest models but calibration and Brier error were worse; this is explicitly exploratory.. A dual-head neural-network survival model was evaluated against IMDC/MSKCC and Cox/forest benchmarks in a 96-patient retrospective mRCC cohort using repeated nested cross-validation. Its C-index was comparable to Cox and forest models but calibration and Brier error were worse; this is explicitly exploratory. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Neural network predicts death risk and survival time; predictors screened/tuned within repeated stratified 5-fold CV (50 outer folds), compared on identical splits with Cox and random survival forest. n=96, C-index 0.685 vs IMDC+nephrectomy Cox 0.692 and RSF 0.695; worse Brier score/calibration slope 0.54.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Neural network predicts death risk and survival time; predictors screened/tuned within repeated stratified 5-fold CV (50 outer folds), compared on identical splits with Cox and random survival forest. n=96, C-index 0.685 vs IMDC+nephrectomy Cox 0.692 and RSF 0.695; worse Brier score/calibration slope 0.54.
Test and result
A dual-head neural-network survival model was evaluated against IMDC/MSKCC and Cox/forest benchmarks in a 96-patient retrospective mRCC cohort using repeated nested cross-validation. Its C-index was comparable to Cox and forest models but calibration and Brier error were worse; this is explicitly exploratory.
Limitation / still unknown
A dual-head neural-network survival model was evaluated against IMDC/MSKCC and Cox/forest benchmarks in a 96-patient retrospective mRCC cohort using repeated nested cross-validation. Its C-index was comparable to Cox and forest models but calibration and Brier error were worse; this is explicitly exploratory.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcm/jcm-15-08262/article_deploy/jcm-15-08262.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

DREM: integrating resistance and domain probabilities to prioritize plant immune receptors

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: sequence-learning results recover characterized resistance proteins and identify a specific genome-scale set of candidate noncanonical plant immune receptors; candidates remain unvalidated.

Technology: two sequence models (two-layer BiLSTM with additive attention), a binary resistance classifier and multi-label domain classifier, contribute to a composite confidence score.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
two sequence models (two-layer BiLSTM with additive attention), a binary resistance classifier and multi-label domain classifier, contribute to a composite confidence score.
Limitation / still unknown
Methods/Results: characterized-reference recovery and held-out comparison support classification; 54/55 characterized NLR loci rank in top 1.7%. Novel candidate proteins are computational priorities, not experimentally confirmed new resistance genes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.64898/2026.10.21.760626

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Quantitative assessment of perirenal fat density in renal cell carcinoma: a combined radiologist-driven and YOLO-based CT study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: YOLOv8-based normalized PRFD assessment compared with radiologist manual measurements in 57 RCC cases and 82 controls; upper-pole PRFD differences at left/right (p=.032/.040); model measurements broadly agree with manual assessment depending on attenuation normalization. Article AI statement also discloses QuillBot copyediting, but that is not the claimed result.

Technology: YOLO-derived normalized perirenal-fat measurements support an RCC-associated imaging phenotype; cross-sectional density differences do not establish causation or clinical utility.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
YOLO-derived normalized perirenal-fat measurements support an RCC-associated imaging phenotype; cross-sectional density differences do not establish causation or clinical utility.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/radiology/articles/10.3389/fradi.2026.1911034/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The feasibility of training nnU-Net to segment pediatric Wilms tumor with small datasets.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The study tests how training-set size affects Wilms-tumor segmentation feasibility.

Technology: nnU-Net segments pediatric Wilms tumors from CT images; the study compares segmentation performance across training-set sizes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
nnU-Net segments pediatric Wilms tumors from CT images; the study compares segmentation performance across training-set sizes.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2026.1841994

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning model with post-hoc SHAP interpretation based on perioperative clinical and imaging parameters for prediction of 1-year radiographic recurrence after hallux valgus

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A machine-learning model trained on 389 cases and temporally validated in 208 cases predicts one-year radiographic recurrence after hallux-valgus surgery; the result is prognostic performance, not causal treatment guidance.

Technology: Machine-learning prediction of one-year radiographic recurrence

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
SVM/model-based prediction integrates perioperative clinical and imaging parameters; SHAP interprets feature contributions.
Limitation / still unknown
A later temporal holdout supports internal prognostic validation; single-center data and event count limit transportability; no causal or clinical-benefit claim is made.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/articles/10.3389/fsurg.2026.1987508/pdf · https://doi.org/10.3389/fsurg.2026.1987508

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Adversarial domain adaptation for false-positive reduction in hospital-to-robotic echocardiographic screening under limited target-domain data.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Domain adaptation improves specificity of echocardiographic screening when transferring a model from hospital images to a robotic acquisition domain; the robot cohort contains healthy subjects, so this supports false-positive reduction rather than pathology sensitivity.

Technology: Domain-adversarial neural adaptation adjusts the echocardiographic classifier to robotic acquisition data.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Domain-adversarial neural adaptation adjusts the echocardiographic classifier to robotic acquisition data.
Limitation / still unknown
Hospital test cohort and healthy robotic data support specificity only; no robotic pathology sensitivity or prospective clinical outcome is established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/articles/10.3389/frobt.2026.1900388/pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

An artificial intelligence-based audio recognition system for dysarthria home-based rehabilitation: A proof-of-concept study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Whisper-ST1 fine-tuned from Whisper-base classifies dysarthria severity from real-world and research speech; held-out metrics and therapist agreement support a bounded proof-of-concept result.

Technology: transfer learning, progressive unfreezing and ordinal regression on TORGO plus social-media speech; held-out sample accuracy 75%, macro-F1 72%, and therapist agreement. Authors explicitly state external robustness and clinical effectiveness remain unestablished.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
transfer learning, progressive unfreezing and ordinal regression on TORGO plus social-media speech; held-out sample accuracy 75%, macro-F1 72%, and therapist agreement. Authors explicitly state external robustness and clinical effectiveness remain unestablished.
Test and result
transfer learning, progressive unfreezing and ordinal regression on TORGO plus social-media speech; held-out sample accuracy 75%, macro-F1 72%, and therapist agreement. Authors explicitly state external robustness and clinical effectiveness remain unestablished.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001794

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development of an artificial intelligence algorithm for automated detection of cardiovascular intervention-related complications.

Source check: Source checked. Not independent scientific confirmation.

Finding: The EHR model detects specified cardiovascular-intervention complications, with retrospective development and prospective single-institution validation. Performance supports this bounded ascertainment result; PPV/F1 and single-center scope limit generalization. The result is reported in the Netherlands Heart Journal. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: structured and free-text EHR features, sentence-transformer plus RF/logistic models, 6,564 patients (2,999 development/test and 3,565 prospective validation)

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
structured and free-text EHR features, sentence-transformer plus RF/logistic models, 6,564 patients (2,999 development/test and 3,565 prospective validation)
Test and result
The EHR model detects specified cardiovascular-intervention complications, with retrospective development and prospective single-institution validation. Performance supports this bounded ascertainment result; PPV/F1 and single-center scope limit generalization.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s12471-026-02082-2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-derived intracellular protein-based prognostic model reflecting a senescence-associated secretory phenotype-driven immunosuppressive microenvironment in lung adenocarcinoma.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The original Methods use NMF, LASSO, random forest and XGBoost with public multi-omics, cell lines, and clinical specimens; Results associate a five-gene SASP-PRS with LUAD survival and inferred immune features. Authors caution treatment-response use is exploratory and calls for prospective/functional validation.

Technology: NMF, LASSO, random forest and XGBoost prognostic models

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Combines public multi-omics and clinical data to derive a five-gene SASP prognostic score and associate it with lung adenocarcinoma survival and inferred immune features.
Limitation / still unknown
Treatment-response use remains exploratory; prospective and functional validation are still needed, as the authors state.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13695421/?report=reader

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

HalluciGuard-eye: knowledge graph-enhanced retrieval-augmented generation for hallucination detection and mitigation in ophthalmic multimodal large language models.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: HalluciGuard-Eye combines knowledge-graph retrieval and hierarchical factuality scoring; across four ophthalmic imaging benchmarks it reports 42.2% lower hallucination rates and 91.6% factual accuracy on EyePACS, while diabetic-retinopathy sensitivity/specificity reached 89.6%/92.1%. This is a benchmark result, not clinical deployment evidence.

Technology: Knowledge-graph retrieval-augmented multimodal language model with factuality scoring

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Knowledge-graph RAG, visual-semantic alignment and factuality grounding reduce errors in multimodal ophthalmic LLM outputs.
Limitation / still unknown
Four public benchmarks, component ablation and reported metrics checked; benchmark validation does not establish prospective clinical safety.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2026.1976190/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Mobile AI-Assisted Oral Lesion Triage for Community Oral Cancer Screening: Prospective Field Evaluation Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A locked mobile AI system was tested prospectively in community oral-cancer screening; expert adjudication showed limited positive predictive value for red high-risk triage, with stronger yellow-category performance.

Technology: Methods describe prospective deployment and expert reference; Results report lesion-level red/yellow/green triage performance.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Methods describe prospective deployment and expert reference; Results report lesion-level red/yellow/green triage performance.
Limitation / still unknown
A locked mobile AI system was tested prospectively in community oral-cancer screening; expert adjudication showed limited positive predictive value for red high-risk triage, with stronger yellow-category performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13683989/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and External Validation of a Machine Learning-Derived Risk Stratification Score for Metastatic Castration-Resistant Prostate Cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding: The PROGRESS machine-learning-derived score stratified overall-survival risk in metastatic castration-resistant prostate cancer and was evaluated in two independent patient datasets.

Technology: Methods describe patient datasets, model development and statistical analyses; Results report generation of the score and validation in two patient cohorts.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Methods describe patient datasets, model development and statistical analyses; Results report generation of the score and validation in two patient cohorts.
Test and result
Methods describe patient datasets, model development and statistical analyses; Results report generation of the score and validation in two patient cohorts.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13722423/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Explore the ideology of deep learning in ENSO forecasts

Source check: Source checked. Not independent scientific confirmation.

Finding: The study analyzes a deep-learning ENSO forecaster using bounded-variation methods and tests regional predictor contributions and spring-barrier forecast behavior. The result is reported in the Geoscience Letters. The study does not establish generalization beyond its reported setting or prospective real-world benefit. Affiliation was not stated in the reviewed material.

Technology: A deep-learning ENSO forecasting model was analyzed with bounded-variation methods to examine regional predictor contributions and the spring predictability barrier.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A deep-learning ENSO forecasting model was analyzed with bounded-variation methods to examine regional predictor contributions and the spring predictability barrier.
Test and result
The study interprets a deep-learning ENSO forecaster with bounded-variation analysis and tests regional predictor contributions and spring-barrier forecast behavior.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s40562-026-00504-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning-based counting and body-area measurement enable sensitive population-level endpoints in the Collembola reproduction test

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: YOLOv8-supported counting and area measurement yield a validated, more sensitive composite toxicity endpoint for Collembola reproduction tests.

Technology: YOLOv8 detects Collembola, Voronoi-Otsu segmentation measures body area, and Gaussian-mixture analysis evaluates composite toxicity endpoints.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
YOLOv8 detects Collembola, Voronoi-Otsu segmentation measures body area, and Gaussian-mixture analysis evaluates composite toxicity endpoints.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.64898/2026.10.19.760133

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning-Driven Prediction and Optimization of Thermoelectric Transport Properties in Ce-Doped CaMnO3 Ceramics.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Original ACS Methods/Results report measured Ce-doped CaMnO3 transport data modeled by SVM/MLP/XGBoost/CNN-LSTM; the CNN-LSTM interpretation locates a candidate x=0.05–0.06, 350–500 °C region, with simulated PF >0.35 mW/K²m. Admit narrowly as a model-guided interpretation of the measured material series; the optimum itself was not independently synthesized/validated.

Technology: Machine-learning analysis applies the stated model to the paper’s materials dataset and evaluates the reported outcome.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis applies the stated model to the paper’s materials dataset and evaluates the reported outcome.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pubs.acs.org/acsodf/article-pdf/doi/10.1021/acsomega.6c09057/68921240/acsomega.6c09057.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Vaginal Microbiome Pharmacology Foundation Model for Microbiome-Based Drug Development and Reproductive Health Insights

Source check: Source checked. Not independent scientific confirmation.

Finding: A deep-autoencoder vaginal microbiome pharmacology model produced drug–microbiome hypotheses, and the paper reports experimental testing of predicted taurine and hydroxyisocaproate effects. The claim is limited to the reported cell/pathogen assays.

Technology: Vaginal microbiome multiomics and autoencoder model; Methods describe training and synthetic-community assays. Results report cisplatin/taurine and metronidazole/hydroxyisocaproate predictions with follow-up assays. Baselines and scope remain those reported by the authors.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Vaginal microbiome multiomics and autoencoder model; Methods describe training and synthetic-community assays. Results report cisplatin/taurine and metronidazole/hydroxyisocaproate predictions with follow-up assays. Baselines and scope remain those reported by the authors.
Test and result
Vaginal microbiome multiomics and autoencoder model; Methods describe training and synthetic-community assays. Results report cisplatin/taurine and metronidazole/hydroxyisocaproate predictions with follow-up assays. Baselines and scope remain those reported by the authors.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s12195-026-00932-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Comparative analysis of GAN-based augmentation pipelines for Parkinson's hypokinetic dysarthria severity detection using deep learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: Results compare GAN-generated speech vs image augmentation for severity classification using 50 speakers (300 high/450 low utterances); speaker-level partitioning keeps each speaker in one fold. The source is an augmentation/model comparison, but the task is specific Parkinson hypokinetic-dysarthria severity classification; claims limited to dataset-level performance, not clinical validation. Learned CNN classifies a disease-specific speech-severity endpoint; augmentation comparisons do not establish diagnostic benefit or external generalization. Testing: Results compare GAN-generated speech vs image augmentation for severity classification using 50 speakers (300 high/450 low utterances); speaker-level partitioning keeps each speaker in one fold. The source is an augmentation/model comparison, but the task is specific Parkinson hypokinetic-dysarthria severity classification; claims limited to dataset-level performance, not clinical validation.. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Learned CNN classifies a disease-specific speech-severity endpoint; augmentation comparisons do not establish diagnostic benefit or external generalization.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned CNN classifies a disease-specific speech-severity endpoint; augmentation comparisons do not establish diagnostic benefit or external generalization.
Test and result
Results compare GAN-generated speech vs image augmentation for severity classification using 50 speakers (300 high/450 low utterances); speaker-level partitioning keeps each speaker in one fold. The source is an augmentation/model comparison, but the task is specific Parkinson hypokinetic-dysarthria severity classification; claims limited to dataset-level performance, not clinical validation.
Limitation / still unknown
Results compare GAN-generated speech vs image augmentation for severity classification using 50 speakers (300 high/450 low utterances); speaker-level partitioning keeps each speaker in one fold. The source is an augmentation/model comparison, but the task is specific Parkinson hypokinetic-dysarthria severity classification; claims limited to dataset-level performance, not clinical validation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2026.1951063/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial intelligence models Traditional Chinese Medicine principles to integrate with biomedicine.

Source check: Source checked. Not independent scientific confirmation.

Finding: A Transformer autoencoder trained on 26,339 eligible TCM formula records learned a low-dimensional symptom-pattern/formula space; held-out/untrained TCM principles align with the representation, ten-model robustness and post-hoc biomedical analyses support a bounded computational result. Testing: The paper reports reproducibility across ten independently trained models and alignment/baseline analyses; biomedical therapeutic prioritization is exploratory.. The primary full text supports this computational result: A Transformer autoencoder trained on 26,339 eligible TCM formula records learned a low-dimensional symptom-pattern/formula space; held-out/untrained TCM principles align with the representation, ten-model robustness and post-hoc biomedical analyses support a bounded computational result. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: A Transformer autoencoder trained on 26,339 eligible TCM formula records learned a low-dimensional symptom-pattern/formula space; held-out/untrained TCM principles align with the representation, ten-model robustness and post-hoc biomedical analyses support a bounded computational result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A Transformer autoencoder trained on 26,339 eligible TCM formula records learned a low-dimensional symptom-pattern/formula space; held-out/untrained TCM principles align with the representation, ten-model robustness and post-hoc biomedical analyses support a bounded computational result.
Test and result
The paper reports reproducibility across ten independently trained models and alignment/baseline analyses; biomedical therapeutic prioritization is exploratory.
Limitation / still unknown
The primary full text supports this computational result: A Transformer autoencoder trained on 26,339 eligible TCM formula records learned a low-dimensional symptom-pattern/formula space; held-out/untrained TCM principles align with the representation, ten-model robustness and post-hoc biomedical analyses support a bounded computational result.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13683242/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Network organization of protein binding-site landscapes across scales

Source check: Source checked. Not independent scientific confirmation.

Finding: ProtT5-based network analysis identifies multiscale organization and cross-protein recognition relationships among experimentally resolved binding sites; drug cross-reactivity remains a hypothesis.

Technology: ProtT5 protein-language-model embeddings and binding-site network analysis

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Materials and Methods, “Binding site embeddings and network construction”: ProtT5 residue vectors averaged into 1,024 dimensions, then mutual k-nearest-neighbor network analysis
Test and result
Results, “Structural calibration and a nonredundant binding site atlas” and “Local network neighborhoods”: 121,673-node network, ligand-class assortativity 0.72; communities resolve kinase-site environments and local relationships across low-sequence-identity proteins
Limitation / still unknown
Results: embedding proximity associated with local ProBiS structural similarity; source is experimentally resolved BioLiP2/PDB sites. Source discussion limits pharmacological implications to hypotheses requiring follow-up
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.64898/2026.10.17.759831 · https://api.crossref.org/works/10.64898/2026.10.17.759831

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Accuracy of transfer learning in predicting peri-implantitis onset from intraoral radiograph: A multi-center longitudinal retrospective study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Full Springer PDF: 195 implants / 108 patients from Michigan and London; baseline radiographs and five pretrained CNNs/four transfer-learning strategies; non-hybrid feature extraction yielded best AUC 0.74–0.83 (InceptionV3 .83), while fine-tuning was lower. Retrospective, modest discrimination. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit.

Technology: Pretrained CNNs evaluated with transfer-learning strategies

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full Springer PDF: 195 implants / 108 patients from Michigan and London; baseline radiographs and five pretrained CNNs/four transfer-learning strategies; non-hybrid feature extraction yielded best AUC 0.74–0.83 (InceptionV3 .83), while fine-tuning was lower. Retrospective, modest discrimination
Test and result
Full Springer PDF: 195 implants / 108 patients from Michigan and London; baseline radiographs and five pretrained CNNs/four transfer-learning strategies; non-hybrid feature extraction yielded best AUC 0.74–0.83 (InceptionV3 .83), while fine-tuning was lower. Retrospective, modest discrimination
Limitation / still unknown
Full Springer PDF: 195 implants / 108 patients from Michigan and London; baseline radiographs and five pretrained CNNs/four transfer-learning strategies; non-hybrid feature extraction yielded best AUC 0.74–0.83 (InceptionV3 .83), while fine-tuning was lower. Retrospective, modest discrimination
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s00784-026-07181-3 · https://link.springer.com/content/pdf/10.1007/s00784-026-07181-3.pdf · https://api.crossref.org/works/10.1007%2Fs00784-026-07181-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Pathway-Driven Computational Drug Discovery Framework Integrating Machine Learning and Virtual Screening for Identification of Anti-Rheumatoid Arthritis Natural Small Molecules.

Source check: Source checked. Not independent scientific confirmation.

Finding: A pathway-guided ML workflow selects natural-molecule candidates for rheumatoid arthritis and tests prioritized compounds in an in-vitro inflammatory model after docking and molecular-dynamics analyses; the evidence supports those assay results, not therapeutic efficacy.

Technology: Pathway-guided machine learning ranks rheumatoid-arthritis target candidates and natural molecules; docking and molecular-dynamics analyses prioritize compounds for cellular testing.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Pathway-guided machine learning ranks rheumatoid-arthritis target candidates and natural molecules; docking and molecular-dynamics analyses prioritize compounds for cellular testing.
Test and result
Machine-learning ranking and virtual screening selected compounds; the paper reports reduced IL6/MMP3 in TNFα-stimulated MH7A cells after SM1 treatment. These in-vitro findings do not establish therapeutic efficacy.
Limitation / still unknown
Primary Methods/Results establish the cohort or experiment, comparison and stated validation limits.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/pharmaceuticals/pharmaceuticals-19-01651/article_deploy/pharmaceuticals-19-01651.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-assisted chemometric analysis of salivary ATR-FTIR spectra for the diagnosis of oral submucous fibrosis.

Source check: Source checked. Not independent scientific confirmation.

Finding: Reported classification separates OSMF, habit, and control samples; reported independent cohort n=90.

Technology: XGBoost and other classifiers analyze ATR-FTIR salivary spectra with PCA and cross-validation.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
XGBoost and other classifiers analyze ATR-FTIR salivary spectra with PCA and cross-validation.
Test and result
A separate participant cohort is described for validation; sample composition and blind validation need careful scope.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2026.1927981

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Advanced meteorological modeling framework integrating quantum computing and machine learning for hurricane track prediction beyond traditional forecasting limitations

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A quantum-enhanced neural model reports hurricane-disjoint evaluation on 2016–2023 forecasts (120-h track MAE 83.7 km vs 312.0 km for a classical ensemble) and a 2024 six-storm holdout (mean landfall error 45 km); this is a source-reported modeling result whose small latest holdout limits certainty.

Technology: Quantum-enhanced neural hurricane-track forecasting model

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A quantum-enhanced neural model combines variational quantum features with a classical temporal forecaster for hurricane-track prediction.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2026.1939822/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based prediction of clinical outcomes and left ventricular ejection fraction improvement after percutaneous coronary intervention in patients with reduced ejection fraction.

Source check: Source checked. Not independent scientific confirmation.

Finding: retrospective test-set risk stratification and moderate prediction of LVEF improvement. Methods: CatBoost used 53 routine pre/intra-procedural features for one-year outcomes and LVEF improvement after PCI in patients with baseline LVEF<45%. Testing: Single-centre held-out retrospective evaluation; no causal estimate of PCI benefit.. CatBoost models stratified one-year clinical events and LVEF improvement after PCI among patients with reduced baseline ejection fraction; this is retrospective prognosis, not causal treatment evidence. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: CatBoost used 53 routine pre/intra-procedural features for one-year outcomes and LVEF improvement after PCI in patients with baseline LVEF<45%.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CatBoost used 53 routine pre/intra-procedural features for one-year outcomes and LVEF improvement after PCI in patients with baseline LVEF<45%.
Test and result
Single-centre held-out retrospective evaluation; no causal estimate of PCI benefit.
Limitation / still unknown
CatBoost models stratified one-year clinical events and LVEF improvement after PCI among patients with reduced baseline ejection fraction; this is retrospective prognosis, not causal treatment evidence.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13711680/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC13711680/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC13711680/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction of acute mountain sickness from overnight pulse oximetry during ascent to Everest Base Camp: a field-based machine learning study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Overnight pulse oximetry predicts acute mountain sickness in evaluated participants.

Technology: Overnight pulse oximetry predicts acute mountain sickness in evaluated participants.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Overnight pulse oximetry predicts acute mountain sickness in evaluated participants.
Limitation / still unknown
Evaluation used repeated stratified cross-validation in the same field cohort; the reported mean AUC was about 0.74, so the result does not establish external or prospective clinical performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13681377/ · https://europepmc.org/article/MED/42903699

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Integrated machine learning and multi-dataset analysis identify SDS as a diagnostic and immune-related hub gene in hepatoblastoma.

Source check: Source checked. Not independent scientific confirmation.

Finding: Four genes and two expression subtypes were reported; SDS biology was tested in cells and xenografts.

Technology: RF, LASSO, and SVM-RFE selected hub genes across hepatoblastoma datasets.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
RF, LASSO, and SVM-RFE selected hub genes across hepatoblastoma datasets.
Test and result
In-vitro and in-vivo experiments test the SDS finding; computational cohort validation details remain a limitation.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2026.1895612

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based survival modeling reveals distinct surgical benefits across molecular subtypes in young women with de novo metastatic breast cancer

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning survival analyses in young women with de novo metastatic breast cancer identified molecular-subtype differences in observed surgical-benefit associations; this is retrospective treatment-effect modeling, not causal proof.

Technology: Machine-learning survival models stratified by breast-cancer molecular subtype

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Uses survival modeling stratified by molecular subtype to compare observed surgical-benefit associations in young women with de novo metastatic breast cancer.
Test and result
The retrospective survival analysis compares observed surgical-benefit associations across molecular subtypes in young women with de novo metastatic breast cancer; it does not establish causal benefit.
Limitation / still unknown
The subtype-specific treatment associations are retrospective and do not establish a causal benefit from surgery.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.21203/rs.3.rs-10816106/v1 · https://europepmc.org/article/PPR/PPR1331157 · https://www.researchsquare.com/article/rs-10816106/v1.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Multicenter validation of an explainable machine-learning model for APL, AMoL, and ALL.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Multicenter results report external AUCs 0.89, 0.87, and 0.90 for the restricted three-diagnosis task.

Technology: Random-forest classifier with SHAP-based feature selection

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random-forest model with SHAP feature selection uses routine hematology values in confirmed leukemia subtypes.
Limitation / still unknown
Four-center retrospective validation is reported; authors explicitly exclude general-population screening claims.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2026.1939329

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Absolute and lung volume-normalized deep learning-derived pulmonary vascular measures in idiopathic pulmonary fibrosis: distinct clinical associations and the role of lung volume in prognosis.

Source check: Source checked. Not independent scientific confirmation.

Finding: Dual-center retrospective study of 118 IPF patients linking a DL-derived pulmonary vascular measure to severity and mortality, with DL CT quantification as the enabling AI role and a distinct normalization finding.

Technology: Deep-learning segmentation of CT images quantifies absolute and lung-volume-normalized pulmonary vascular measures in idiopathic pulmonary fibrosis.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep-learning segmentation of CT images quantifies absolute and lung-volume-normalized pulmonary vascular measures in idiopathic pulmonary fibrosis.
Test and result
A dual-center retrospective cohort of 118 patients links the derived measures to disease severity and mortality; association does not establish causation or clinical benefit.
Limitation / still unknown
Primary Methods/Results establish the cohort or experiment, comparison and stated validation limits.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2026.1833470/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Utilizing machine learning and MALDI-TOF MS platform for rapid detection of ceftriaxone-resistant nontyphoidal Salmonella.

Source check: Source checked. Not independent scientific confirmation.

Finding: LightGBM on MALDI-TOF spectra predicted ceftriaxone resistance among nontyphoidal Salmonella isolates, with internal and two external-center evaluations.

Technology: LightGBM classifier on MALDI-TOF mass spectra

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe 472 pediatric isolates, spectra and validation cohorts; Results report resistance classification on internal and external cohorts. PubMed ArticleDate 2026-09-28 precedes later journal issue date.
Test and result
Methods describe 472 pediatric isolates, spectra and validation cohorts; Results report resistance classification on internal and external cohorts. PubMed ArticleDate 2026-09-28 precedes later journal issue date.
Limitation / still unknown
Methods describe 472 pediatric isolates, spectra and validation cohorts; Results report resistance classification on internal and external cohorts. PubMed ArticleDate 2026-09-28 precedes later journal issue date.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pmc&id=PMC13729124

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Integration of Multidimensional Genetic and Single- Cell Analyses to Identify Potential Drug Targets and a Clinical Risk Prediction Model for Cervical Cancer

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Integrated genetic, transcriptomic and single-cell analyses used machine-learning feature selection and Mendelian-randomization evidence to identify cervical-cancer-associated genes and candidate therapeutic targets; the preprint reports specific gene associations, not established treatment effects.

Technology: Machine-learning analysis applies the stated model to the paper’s medicine dataset and evaluates the reported outcome.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis applies the stated model to the paper’s medicine dataset and evaluates the reported outcome.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.21203/rs.3.rs-10712879/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and a single-centre temporal validation of a machine learning-based model for early prediction of acute pancreatitis-associated liver injury.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A random-forest model predicted acute-pancreatitis-associated liver injury from admission features; an internal test cohort (n=279) reached AUC 0.972, and a temporally separate validation cohort reached AUC 0.935. It is a single-center retrospective prediction result requiring external validation.

Technology: LASSO feature selection and nine ML models; Random Forest was selected for outcome prediction.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LASSO feature selection and nine ML models; Random Forest was selected for outcome prediction.
Limitation / still unknown
Temporal validation is explicitly described; internal/temporal test AUC and accuracy reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12876-026-05204-7

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based quantitative pathology reveals intratumoral heterogeneity in ductal carcinoma <i>in situ</i> and predicts invasive upstaging.

Source check: Source checked. Not independent scientific confirmation.

Finding: Quantitative pathology models measured nuclear/architectural heterogeneity in DCIS and found biopsy heterogeneity informative for invasive upstaging, with internal and independent-cohort validation.

Technology: Methods describe single-cell morphometry, integrated grading and surgical-upstaging cohorts; Results report external grading performance and biopsy heterogeneity associated with later invasive upstaging.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Methods describe single-cell morphometry, integrated grading and surgical-upstaging cohorts; Results report external grading performance and biopsy heterogeneity associated with later invasive upstaging.
Test and result
Methods describe single-cell morphometry, integrated grading and surgical-upstaging cohorts; Results report external grading performance and biopsy heterogeneity associated with later invasive upstaging.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13683260/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Interpretable machine-learning risk stratification at the time of diagnosis for 3-year mortality in de novo metastatic prostate cancer: development and temporal validation in SEER.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Temporal discrimination is moderate (AUC about 0.69–0.70); authors bound treatment conclusions as observational.

Technology: CatBoost and six other models for mortality risk prediction

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CatBoost and six other models predict 3-year mortality using SEER data, then temporal and external cohorts.
Limitation / still unknown
Temporal validation and an independent external cohort are reported; claim is risk stratification, not treatment selection.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2026.1904589

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting In-Hospital Mortality in Pediatric Sepsis: Machine Learning Development and Multicenter Validation.

Source check: Source checked. Not independent scientific confirmation.

Finding: Eight algorithms were evaluated for in-hospital mortality using predictors from the first 24 hours of pediatric ICU admission. XGBoost had AUC 0.821 (95% CI 0.751–0.892) and was externally assessed on the PIC pediatric and MIMIC-IV adult databases; the latter tests biological generalizability, not pediatric clinical deployment.

Technology: Eight machine-learning algorithms use first-24-hour ICU variables; SHAP interpretation and spline calibration were applied to the selected XGBoost model.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Eight machine-learning algorithms use first-24-hour ICU variables; SHAP interpretation and spline calibration were applied to the selected XGBoost model.
Test and result
The retrospective development cohort was from Chongqing Children’s Hospital. XGBoost achieved AUC 0.821 (95% CI 0.751–0.892); external assessment used the PIC pediatric database and adult MIMIC-IV data.
Limitation / still unknown
Development was retrospective at one hospital; the external MIMIC-IV cohort was adult and supports biological generalizability only. No prospective pediatric workflow or mortality benefit was tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13544363/ · https://europepmc.org/article/MED/42699765

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prospective economic evaluation of a predictive artificial intelligence model for sepsis: Effects on hospital costs and return on investment.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary paper evaluates deployed BIAlert-Sepsis and associates its implementation with resource-use and cost changes; nonrandomized before–after design precludes causal attribution.

Technology: retrospective quasi-experimental comparison of 6,168 baseline and 1,871 implementation-period sepsis patients, with COVID period excluded; GLM and interrupted time-series/economic analyses. Published abstract reports lower ICU use and costs, but causal attribution is limited by before–after design and mature sepsis program.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
retrospective quasi-experimental comparison of 6,168 baseline and 1,871 implementation-period sepsis patients, with COVID period excluded; GLM and interrupted time-series/economic analyses. Published abstract reports lower ICU use and costs, but causal attribution is limited by before–after design and mature sepsis program.
Test and result
retrospective quasi-experimental comparison of 6,168 baseline and 1,871 implementation-period sepsis patients, with COVID period excluded; GLM and interrupted time-series/economic analyses. Published abstract reports lower ICU use and costs, but causal attribution is limited by before–after design and mature sepsis program.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/globalpublichealth/article?id=10.1371/journal.pgph.0006059

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Genomic signatures associated with epidemiologically defined high-risk pathogenic Escherichia coli isolates identified by interpretable machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary publisher PDF Methods/Results confirm 1,029 isolates, epidemiological EpiHR labeling, RF selection, 10-fold CV, leave-one-group-out analyses and an independent 1,908-genome test; surveillance prioritization is the justified scope.

Technology: RF on virulence-associated genomic features; AUC .98/accuracy .93 in CV and independent-test AUC .97/sensitivity .98. Source itself cautions predictions accompany epidemiological information.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
RF on virulence-associated genomic features; AUC .98/accuracy .93 in CV and independent-test AUC .97/sensitivity .98. Source itself cautions predictions accompany epidemiological information.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://jmicrobiol.or.kr/upload/pdf/jm-2604011.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and validation of venous thrombosis risk prediction model and scale for lung cancer patients in intensive care unit based on interpretable machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: An interpretable model predicted venous-thrombosis risk among lung-cancer patients in ICU.

Technology: Boruta, variance-inflation analysis and LASSO screened predictors; nine models, including CatBoost, estimated ICU venous-thromboembolism risk, with SHAP interpretation and a score derived from the best model.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Boruta, variance-inflation analysis and LASSO screened predictors; nine models, including CatBoost, estimated ICU venous-thromboembolism risk, with SHAP interpretation and a score derived from the best model.
Test and result
In 3,694 ICU records across MIMIC-IV, eICU and NWICU, CatBoost AUCs were 0.7648 in training, 0.6437 in internal validation, 0.6822 in eICU and 0.7153 in NWICU. This retrospective prediction does not establish clinical benefit.
Limitation / still unknown
The cohort is drawn from existing ICU databases; reported validation AUCs are modest (0.64–0.72), and no prospective clinical-impact evaluation is reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13559455/ · https://europepmc.org/article/MED/42724662

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-driven drug repurposing and computational validation for Nipah virus.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary source Methods/Results confirm the represented result: Random Forest models screen 9,021 compounds for Nipah virus candidates with a held-out test set and docking, a distinct computational finding with the AI role specified, though results are predictions without experimental validation. Publication venue was not available in the checked primary metadata. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Random Forest models screen 9,021 compounds for Nipah virus candidates with a held-out test set and docking, a distinct computational finding with the AI role specified, though results are predictions without experimental validation.

Publication: Publication status unspecified in annotation · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Random Forest models screen 9,021 compounds for Nipah virus candidates with a held-out test set and docking, a distinct computational finding with the AI role specified, though results are predictions without experimental validation.
Test and result
The primary source Methods/Results confirm the represented result: Random Forest models screen 9,021 compounds for Nipah virus candidates with a held-out test set and docking, a distinct computational finding with the AI role specified, though results are predictions without experimental validation.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11030-026-11708-x

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Endostrat: a multicenter study of AI-assisted two-level risk stratification for gastric precancerous and early neoplastic lesion screening.

Source check: Source checked. Not independent scientific confirmation.

Finding: In a retrospective three-center cohort of 8,829 patients, the two-stage Endostrat AI system stratified gastric precancerous and early-neoplastic lesion risk. This is retrospective screening-risk validation, not prospective clinical benefit.

Technology: Endostrat is a two-stage AI system for gastric-lesion risk stratification.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Endostrat is a two-stage AI system for gastric-lesion risk stratification.
Test and result
Retrospective multicenter validation in 8,829 patients across three centers; prospective screening outcomes were not tested.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2026.1918150/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Construction and internal validation of a machine learning model for predicting 30-day readmission after hip surgery.

Source check: Source checked. Not independent scientific confirmation.

Finding: A retrospective single-center comparison of eleven models selected XGBoost to predict 30-day readmission after hip surgery; internal test validation included decision-curve and SHAP analyses. No external or prospective clinical utility was established.

Technology: XGBoost was selected from eleven machine-learning models to predict 30-day hip-surgery readmission.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
XGBoost was selected from eleven machine-learning models to predict 30-day hip-surgery readmission.
Test and result
Single-center retrospective evaluation with an internal test cohort; decision-curve analysis and SHAP interpretation do not establish external or prospective utility.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2026.1872049/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Symbolic machine learning for vapor-liquid equilibrium prediction in nitrogen + n-alkane mixtures

Source check: Source checked. Not independent scientific confirmation.

Finding: The first result-bearing arXiv version reports symbolic regression corrections to the Peng–Robinson equation using experimental vapor–liquid-equilibrium data from six nitrogen/n-alkane mixtures. Its held-out comparisons show substantially lower pressure-prediction error than the uncorrected equation. This is a specific thermodynamics result, limited to the reported mixtures and pointwise split.

Technology: Official arXiv v1 full text, Methods and Results; first posted 2026-08-06. DOI journal version has matching study title/authors and describes the same nitrogen+n-alkane VLE task; no evidence of a separate earlier result version. The reported random pointwise 80/20 split limits generalization claims.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Official arXiv v1 full text, Methods and Results; first posted 2026-08-06. DOI journal version has matching study title/authors and describes the same nitrogen+n-alkane VLE task; no evidence of a separate earlier result version. The reported random pointwise 80/20 split limits generalization claims.
Test and result
Official arXiv v1 full text, Methods and Results; first posted 2026-08-06. DOI journal version has matching study title/authors and describes the same nitrogen+n-alkane VLE task; no evidence of a separate earlier result version. The reported random pointwise 80/20 split limits generalization claims.
Limitation / still unknown
Official arXiv v1 full text, Methods and Results; first posted 2026-08-06. DOI journal version has matching study title/authors and describes the same nitrogen+n-alkane VLE task; no evidence of a separate earlier result version. The reported random pointwise 80/20 split limits generalization claims.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2608.11255

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Adolescent depression recognition and symptom prediction based on multi-view EEG network: an exploratory study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A multi-view adaptive graph convolution model discriminated 44 depressed adolescents from 30 controls using EEG connectivity; the article reports subject-level 10-fold cross-validation, but no independent external cohort, so the results remain exploratory.

Technology: The MVA-GCN fuses PLV, PCC and PLI connectivity views through graph convolution/adaptive fusion to classify depression and derive network features.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The MVA-GCN fuses PLV, PCC and PLI connectivity views through graph convolution/adaptive fusion to classify depression and derive network features.
Limitation / still unknown
Ten-fold cross-validation partitions at participant level (all epochs from one participant stay in one fold), addressing epoch leakage. Authors explicitly note modest sample and no independent external cohort; treat as proof-of-concept.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fpsyt.2026.1908969 · https://europepmc.org/article/MED/42601901

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Large Language Model Based Teaching Pathway Is Feasible for Developing Clinical Pharmacy Practice Competencies

Source check: Source checked. Not independent scientific confirmation.

Finding: In a small feasibility study, an LLM-based simulated ward-round pathway produced pharmacy-trainee competency scores correlated with expert judgments but systematically underestimated them. It is not validated for high-stakes assessment.

Technology: An LLM-based simulated ward-round teaching pathway generates pharmacy-trainee competency scores for comparison with expert ratings.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
An LLM-based simulated ward-round teaching pathway generates pharmacy-trainee competency scores for comparison with expert ratings.
Test and result
Scores were compared with expert judgments in a small feasibility study; systematic underestimation and lack of high-stakes validation limit interpretation.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.21203/rs.3.rs-10452795/v1 · https://europepmc.org/article/PPR/PPR1289160 · https://www.researchsquare.com/article/rs-10452795/v1.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

NeuroPACT: Efficient Detection of Epileptic Drop Attacks from EEG and EMG Using a Hierarchically Compressed Transformer

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: NeuroPACT used hierarchical EEG/EMG compression and transformer learning to detect epileptic drop-attack segments; the preprint reports external validation sensitivity 0.9183 and reduced parameters/FLOPs across a multicenter cohort.

Technology: NeuroPACT hierarchical-compression transformer

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
NeuroPACT used hierarchical EEG/EMG compression and transformer learning to detect epileptic drop-attack segments; the preprint reports external validation sensitivity 0.9183 and reduced parameters/FLOPs across a multicenter cohort.
Limitation / still unknown
The reported evaluation does not establish prospective benefit or performance beyond the studied setting.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.21203/rs.3.rs-10426948/v1 · https://europepmc.org/article/PPR/PPR1289028 · https://www.researchsquare.com/article/rs-10426948/v1.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Interpretable multimodal machine learning for injury risk stratification in adolescent athletes: integrating kinematic, physiological, and load data with SHAP/LIME explainability.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Multimodal learning stratified lower-limb injury risk among adolescent athletes using kinematic, physiological, and load data.

Technology: Multimodal learning stratified lower-limb injury risk among adolescent athletes using kinematic, physiological, and load data.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Multimodal learning stratified lower-limb injury risk among adolescent athletes using kinematic, physiological, and load data.
Limitation / still unknown
This is a cross-sectional 320-athlete dataset evaluated by five-fold cross-validation; no independent external or prospective injury-prevention evaluation is reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13472085/ · https://europepmc.org/article/MED/42598588

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Associated factors and machine learning-based prediction model construction for positive reproductive tract group B streptococcus screening in late-pregnancy women.

Source check: Source checked. Not independent scientific confirmation.

Finding: Single-center retrospective cohort of 363 women (91 positive) is used for development and internal validation.

Technology: Machine learning predicts positive reproductive-tract group B streptococcus screening in late pregnancy from clinical factors

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine learning predicts positive reproductive-tract group B streptococcus screening in late pregnancy from clinical factors.
Test and result
Single-center retrospective cohort of 363 women (91 positive) is used for development and internal validation. Evaluation: Limited event count and internal-only validation bound performance.
Limitation / still unknown
The single-center retrospective model used 363 women, including 91 positive cases, and internal validation only; independent generalization is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.ncbi.nlm.nih.gov/research/bionlp/RESTful/pmcoa.cgi/BioC_json/PMC13700136/unicode

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Model for Predicting Suicide Risks Among Patients With Posttraumatic Stress Disorder Who Received Opioids.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A random-forest suicide-risk model for opioid-treated PTSD patients was evaluated on internal and later external cohorts, a specific clinical risk result.

Technology: Random-forest models use EHR and claims features to predict six-month suicide-related outcomes in opioid-treated PTSD patients.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Random-forest models use EHR and claims features to predict six-month suicide-related outcomes in opioid-treated PTSD patients.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13420204/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Three-class obstructive sleep apnea severity assessment: a parallel AHI and ODI explainable artificial intelligence framework using craniofacial-enriched clinical data.

Source check: Source checked. Not independent scientific confirmation.

Finding: The original full text supports the bounded AI-enabled finding: Explainable AHI- and ODI-based models classify OSA severity in 233 adults; Results report three-class accuracy, sensitivity and specificity within the referral cohort.

Technology: Explainable AHI- and ODI-based models classify OSA severity in 233 adults; Results report three-class accuracy, sensitivity and specificity within the referral cohort.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Explainable AHI- and ODI-based models classify OSA severity in 233 adults; Results report three-class accuracy, sensitivity and specificity within the referral cohort.
Test and result
The reported cohort or experimental dataset, comparator/reference and validation are specific to this case; conclusions remain bounded to that setting.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s11325-026-03774-0

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and Application of Automated Preoperative Planning System for Developmental Dysplasia of the Hip via Keypoint Selective Integration.

Source check: Source checked. Not independent scientific confirmation.

Finding: Full Springer PDF: prospective DDH preoperative planning system uses integrated keypoint detection and planning parameters; results report keypoint detection and planning-system evaluation, with rare severe DDH cases limiting learning. It does not establish improved patient outcomes. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit.

Technology: Full Springer PDF: prospective DDH preoperative planning system uses integrated keypoint detection and planning parameters; results report keypoint detection and planning-system evaluation, with rare severe DDH cases limiting learning. It does not establish improved patient outcomes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full Springer PDF: prospective DDH preoperative planning system uses integrated keypoint detection and planning parameters; results report keypoint detection and planning-system evaluation, with rare severe DDH cases limiting learning. It does not establish improved patient outcomes
Test and result
Full Springer PDF: prospective DDH preoperative planning system uses integrated keypoint detection and planning parameters; results report keypoint detection and planning-system evaluation, with rare severe DDH cases limiting learning. It does not establish improved patient outcomes
Limitation / still unknown
Full Springer PDF: prospective DDH preoperative planning system uses integrated keypoint detection and planning parameters; results report keypoint detection and planning-system evaluation, with rare severe DDH cases limiting learning. It does not establish improved patient outcomes
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s10278-026-02149-x · https://link.springer.com/content/pdf/10.1007/s10278-026-02149-x.pdf · https://api.crossref.org/works/10.1007%2Fs10278-026-02149-x

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A prospective study for validating an automated AI-based system for detecting age-related macular degeneration in clinical settings.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary article reports prospective clinical validation of an automated AMD screening system and its performance against clinician grading; the study is single-location and supports screening accuracy, not improved patient outcomes.

Technology: The primary article reports prospective clinical validation of an automated AMD screening system and its performance against clinician grading; the study is single-location and supports screening accuracy, not improved patient outcomes.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The primary article reports prospective clinical validation of an automated AMD screening system and its performance against clinician grading; the study is single-location and supports screening accuracy, not improved patient outcomes.
Limitation / still unknown
The primary article reports prospective clinical validation of an automated AMD screening system and its performance against clinician grading; the study is single-location and supports screening accuracy, not improved patient outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-025-94866-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and internal validation of a nomogram for early prediction of deep vein thrombosis after liver transplantation: an exploratory machine-learning comparison.

Source check: Source checked. Not independent scientific confirmation.

Finding: A 131-recipient retrospective transplant cohort compared ML algorithms with a parsimonious logistic nomogram for early postoperative DVT; the principal model was logistic regression, and the ML comparison was exploratory.

Technology: Machine-learning algorithms were compared with a parsimonious logistic-regression nomogram for early postoperative DVT prediction after liver transplantation; the principal approach was logistic regression.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning algorithms were compared with a parsimonious logistic-regression nomogram for early postoperative DVT prediction after liver transplantation; the principal approach was logistic regression.
Test and result
The comparison is exploratory within a retrospective 131-recipient cohort; the study does not establish benefit from model-guided prevention.
Limitation / still unknown
The 131-recipient study is retrospective and internally validated; the authors identify logistic regression as the main model, with ML comparisons exploratory and no external clinical validation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2026.1863619/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and internal validation of an interpretable machine learning prognostic prediction model for postoperative delirium after esophagectomy.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary full text supports the case-specific result: An interpretable model predicted postoperative delirium after esophagectomy with internal validation.

Technology: Methods identify the case-specific learned operation: An interpretable model predicted postoperative delirium after esophagectomy with internal validation

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods identify the case-specific learned operation: An interpretable model predicted postoperative delirium after esophagectomy with internal validation.
Test and result
The primary full text supports the case-specific result: An interpretable model predicted postoperative delirium after esophagectomy with internal validation. Evaluation: The full text reports the case-specific data/cohort, comparison/reference and evaluation; no broader clinical benefit is inferred.
Limitation / still unknown
The prognostic model was internally validated in the reported esophagectomy cohort; external validation and clinical impact are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13559237/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Metabolites stratify future major depressive disorder risk in obese population: a longitudinal machine learning analysis.

Source check: Source checked. Not independent scientific confirmation.

Finding: PMC full text Methods/Results: UK Biobank 41,459 obese participants, 3,642 incident MDD; LightGBM temporal validation AUC 0.738–0.776. Retrospective risk stratification only; no independent multiethnic cohort or causal claim. Testing: PMC full text Methods/Results: UK Biobank 41,459 obese participants, 3,642 incident MDD; LightGBM temporal validation AUC 0.738–0.776. Retrospective risk stratification only; no independent multiethnic cohort or causal claim.. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: PMC full text Methods/Results: UK Biobank 41,459 obese participants, 3,642 incident MDD; LightGBM temporal validation AUC 0.738–0.776. Retrospective risk stratification only; no independent multiethnic cohort or causal claim.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PMC full text Methods/Results: UK Biobank 41,459 obese participants, 3,642 incident MDD; LightGBM temporal validation AUC 0.738–0.776. Retrospective risk stratification only; no independent multiethnic cohort or causal claim.
Test and result
PMC full text Methods/Results: UK Biobank 41,459 obese participants, 3,642 incident MDD; LightGBM temporal validation AUC 0.738–0.776. Retrospective risk stratification only; no independent multiethnic cohort or causal claim.
Limitation / still unknown
PMC full text Methods/Results: UK Biobank 41,459 obese participants, 3,642 incident MDD; LightGBM temporal validation AUC 0.738–0.776. Retrospective risk stratification only; no independent multiethnic cohort or causal claim.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13439247/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Characterization of the crotonylation landscape in gastric cancer: a machine learning-derived prognostic signature and the oncogenic role of TIMP1.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Single-cell analysis characterizes crotonylation patterns and links them to gastric-cancer phenotypes.

Technology: BackgroundGastric cancer (GC) remains a lethal malignancy with high heterogeneity and poor prognosis. Crotonylation, a key post-translational modification, contributes significantly to oncogenic processes and immune system interactions; however, its prognostic significance in gastric cancer is not fully understood.MethodsWe integrated single-cell RNA sequencing and bulk transcriptomics to evaluate cell-specific crotonylation activity and utilized WGCNA to identify core crotonylation-related…

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The authors integrated single-cell and bulk transcriptomic data to derive and validate a machine-learning prognostic signature for crotonylation-related genes; separate cell experiments tested TIMP1’s role in gastric-cancer cell behavior.
Limitation / still unknown
The prognostic signature is based on existing cohorts; TIMP1 assays support a cell-level role but do not establish causal disease mechanisms or clinical utility in patients.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13693145/ · https://europepmc.org/article/MED/42920177

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based prediction of postoperative continuous renal replacement therapy initiation after open thoracoabdominal aortic repair.

Source check: Source checked. Not independent scientific confirmation.

Finding: LightGBM predicts postoperative CRRT initiation after open thoracoabdominal aortic repair in 372 patients, with internal test/bootstrapped AUC 0.814; six perioperative variables were influential. This is exploratory single-center prediction with 66 events and no external validation. Publication venue was not available in the checked primary metadata. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Eight ML models plus logistic regression use perioperative predictors

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Eight ML models plus logistic regression use perioperative predictors
Test and result
LightGBM predicts postoperative CRRT initiation after open thoracoabdominal aortic repair in 372 patients, with internal test/bootstrapped AUC 0.814; six perioperative variables were influential. This is exploratory single-center prediction with 66 events and no external validation.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12882-026-05234-z

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The Energy Landscape of DNA-Binding Proteins Along the Genome

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Read full arXiv v1 Methods/Results (same title/authors as DOI article): multimodal neural model trained on 24,923 protein-DNA complexes, DNA-split r=.632, mutation/ChIP-seq tests; applied to PU.1 chromosome profile and compared with ChIP-seq. Computational result, one factor/application; v1 predates journal version.

Technology: A multimodal neural model trained on protein-DNA complexes predicts DNA-binding protein energy landscapes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A multimodal neural model trained on protein-DNA complexes predicts DNA-binding protein energy landscapes.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/pdf/2607.17753

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning-Driven Prediction of Early Enteral Nutrition Intolerance in ICU Patients Using Intra-Abdominal Pressure and Selected Gut Microbial Taxa.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: retrospective 300-patient ICU cohort split 240/60; LASSO feature selection and nomogram combine IAP, gut taxa and clinical factors; AUC 0.900 in both development and internal holdout. Small single-center validation and no external/generalizability evidence.

Technology: Primary Methods/Results: retrospective 300-patient ICU cohort split 240/60; LASSO feature selection and nomogram combine IAP, gut taxa and clinical factors; AUC 0.900 in both development and internal holdout. Small single-center validation and no external/generalizability evidence.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
retrospective 300-patient ICU cohort split 240/60; LASSO feature selection and nomogram combine IAP, gut taxa and clinical factors; AUC 0.900 in both development and internal holdout. Small single-center validation and no external/generalizability evidence.
Limitation / still unknown
retrospective 300-patient ICU cohort split 240/60; LASSO feature selection and nomogram combine IAP, gut taxa and clinical factors; AUC 0.900 in both development and internal holdout. Small single-center validation and no external/generalizability evidence.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13446374/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Comparative evaluation of the aggregate index of systemic inflammation (AISI) and its modified version for early detection of central line-associated bloodstream infection: a pilot study using machine learning techniques.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full primary article supports a retrospective, single-center result: 234 central-line patients (39 CLABSI), four ML algorithms, modified AISI/XGBoost AUROC 0.99, 97% sensitivity and 98% specificity. Authors require prospective validation and caution against treatment use.

Technology: four models (LR, XGBoost, SVM, AdaBoost) with AISI/modified AISI and clinical variables; 70:30 train/test.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
four models (LR, XGBoost, SVM, AdaBoost) with AISI/modified AISI and clinical variables; 70:30 train/test.
Limitation / still unknown
Internal test split and bootstrap; small retrospective single-center cohort; no patient-outcome benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3205/dgkh000661 · https://europepmc.org/article/MED/42483468

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Pre-treatment Tumor Volume Growth Rate and Baseline Volume Measured by an Artificial Intelligence Program Predict Overall Survival in Patients With Pleural Mesothelioma Undergoing Deferred Systemic Th

Source check: Source checked. Not independent scientific confirmation.

Finding: single-center retrospective cohort of 71 mesothelioma patients with serial pretreatment CT; an AI segmentation program measured tumor volumes/growth rates; Cox/Kaplan-Meier analyses found associations with overall survival. A bounded prognostic association, not causal or external validation. Published in JTO Clinical and Research Reports.

Technology: Primary Methods/Results: single-center retrospective cohort of 71 mesothelioma patients with serial pretreatment CT; an AI segmentation program measured tumor volumes/growth rates; Cox/Kaplan-Meier analyses found associations with overall survival. A bounded prognostic association, not causal or external validation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
single-center retrospective cohort of 71 mesothelioma patients with serial pretreatment CT; an AI segmentation program measured tumor volumes/growth rates; Cox/Kaplan-Meier analyses found associations with overall survival. A bounded prognostic association, not causal or external validation.
Test and result
single-center retrospective cohort of 71 mesothelioma patients with serial pretreatment CT; an AI segmentation program measured tumor volumes/growth rates; Cox/Kaplan-Meier analyses found associations with overall survival. A bounded prognostic association, not causal or external validation.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13544273/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Large language models for breast cancer treatment planning: a blinded real-world evaluation of DeepSeek, ChatGPT, and oncologist recommendations.

Source check: Source checked. Not independent scientific confirmation.

Finding: In a blinded review of 213 breast-cancer records, experts compared plans generated by DeepSeek V3.1 and ChatGPT-5 with the historical oncologist plans. DeepSeek received higher guideline-concordance scores, and agreement was weaker in advanced-stage cases.

Technology: LLMs generate treatment recommendations from de-identified records for blinded expert scoring.

Family DeepSeek: profile · models page

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LLMs generate treatment recommendations from de-identified records for blinded expert scoring.
Test and result
Three expert reviewers rated plans from historical cases and compared concordance across cancer stages.
Limitation / still unknown
The study measured retrospective agreement with guidelines and clinicians. Patient outcomes were outside its design.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2026.1862730/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Ultrasound-Based AI in Predicting Hormone Receptor Status in Breast Cancer: Is "Digital Biopsy" Possible.

Source check: Source checked. Not independent scientific confirmation.

Finding: Full Springer PDF: 353 breast-cancer tumors in 311 women; ultrasound/radiomics ML predicted histopathologic hormone-receptor status. The reference is biopsy pathology; single-center retrospective design limits generalization. Testing: Full Springer PDF: 353 breast-cancer tumors in 311 women; ultrasound/radiomics ML predicted histopathologic hormone-receptor status. The reference is biopsy pathology; single-center retrospective design limits generalization.. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Full Springer PDF: 353 breast-cancer tumors in 311 women; ultrasound/radiomics ML predicted histopathologic hormone-receptor status. The reference is biopsy pathology; single-center retrospective design limits generalization.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full Springer PDF: 353 breast-cancer tumors in 311 women; ultrasound/radiomics ML predicted histopathologic hormone-receptor status. The reference is biopsy pathology; single-center retrospective design limits generalization.
Test and result
Full Springer PDF: 353 breast-cancer tumors in 311 women; ultrasound/radiomics ML predicted histopathologic hormone-receptor status. The reference is biopsy pathology; single-center retrospective design limits generalization.
Limitation / still unknown
Full Springer PDF: 353 breast-cancer tumors in 311 women; ultrasound/radiomics ML predicted histopathologic hormone-receptor status. The reference is biopsy pathology; single-center retrospective design limits generalization. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s10278-026-02088-7 · https://link.springer.com/content/pdf/10.1007/s10278-026-02088-7.pdf · https://api.crossref.org/works/10.1007%2Fs10278-026-02088-7

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and Temporal Validation of Machine Learning Models for Hyponatremia Risk in Community-Dwelling Older Adults: A Nationwide Claims-Based Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: ML risk models trained on nationwide claims data and evaluated with temporal validation for hyponatremia. Results reports temporal-validation performance in community-dwelling older adults; claims-based endpoints limit interpretation.

Technology: ML risk models trained on nationwide claims data and evaluated with temporal validation for hyponatremia.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ML risk models trained on nationwide claims data and evaluated with temporal validation for hyponatremia.
Limitation / still unknown
The validation uses later South Korean claims records and coded hyponatremia endpoints; temporal validation does not establish prospective performance or clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13363158/ · https://europepmc.org/article/MED/42452533

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A novel deep learning-based grading system for assessing breast arterial calcification on mammograms, as an independent risk factor for predicting adverse cardiovascular events.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Deep-learning BAC segmentation and grading with a retrospective case-control design, adjusted Cox hazard ratios for MACE, and an identified enabling U-Net role.

Technology: Deep-learning BAC segmentation and grading with a retrospective case-control design, adjusted Cox hazard ratios for MACE, and an identified enabling U-Net role.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep-learning BAC segmentation and grading with a retrospective case-control design, adjusted Cox hazard ratios for MACE, and an identified enabling U-Net role.
Limitation / still unknown
The source reports the defined cohort/experiment and its reference or test comparison; interpretation is restricted to that study design.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11547-026-02251-8

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Time-Updated Prognostic Modeling in ICU Patients with Documented Coma or Unresponsiveness Using Routine Arterial Blood Gas Trajectories: An Exploratory Explainable Machine-Learning Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Time-updated logistic-regression models used admission, 24-hour and 72-hour ABG/SpO2 trajectories to estimate ICU death in 108 patients with documented coma or unresponsiveness. The study reports internal cross-validation and bootstrap intervals; respiratory-support confounding and missingness mean this is exploratory ICU-course prognosis, not admission-only prediction.

Technology: Staged ML models on admission variables and ABG/SpO2 trajectory features at 24/72h; 108 single-center ICU patients, 71 deaths. Internal validation, bootstrap intervals, sensitivity analyses. Primary no-respiratory-support logistic models AUC 0.822/0.848/0.895; 72-hour availability and ICU severity limit interpretation.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Staged ML models on admission variables and ABG/SpO2 trajectory features at 24/72h; 108 single-center ICU patients, 71 deaths. Internal validation, bootstrap intervals, sensitivity analyses. Primary no-respiratory-support logistic models AUC 0.822/0.848/0.895; 72-hour availability and ICU severity limit interpretation.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcm/jcm-15-05056/article_deploy/jcm-15-05056.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Plasma proteomics based on machine learning predict the early risk of kidney outcomes in patients with DKD: a prospective cohort study from UK Biobank.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Machine-learning survival models linked plasma proteomics with future kidney outcomes in a prospective UK Biobank diabetic-kidney-disease cohort and reported a protein signature.

Technology: Machine-learning survival models use plasma proteomics to estimate future kidney outcomes in diabetic kidney disease.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning survival models use plasma proteomics to estimate future kidney outcomes in diabetic kidney disease.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13357851/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of distinct subgroups in Chinese patients with Behçet's syndrome via cluster analysis of immune cells and clinical features.

Source check: Source checked. Not independent scientific confirmation.

Finding: Original Results examined; decision reflects the reported result and its limits. Methods identify the learned model and its operation in the reported task. Testing: Original Methods/Results validation design examined; interpretation is bounded to that design.. Immune-cell and clinical-feature clustering identifies four patient subgroups in a prospective cohort; narrow single-center stratification result. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Methods identify the learned model and its operation in the reported task.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Methods identify the learned model and its operation in the reported task.
Test and result
Original Methods/Results validation design examined; interpretation is bounded to that design.
Limitation / still unknown
Immune-cell and clinical-feature clustering identifies four patient subgroups in a prospective cohort; narrow single-center stratification result.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arthritis-research.biomedcentral.com/articles/10.1186/s13075-026-03851-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Urbanization, geographic accessibility, and delayed entry into tuberculosis-designated care in southwestern China: a machine-learning prediction study

Source check: Source checked. Not independent scientific confirmation.

Finding: A LightGBM prediction model using 60,778 tuberculosis surveillance records found delayed entry to TB-designated care common and identified geographic accessibility measures among leading predictors; it reports a held-out test cohort and calls for external validation. The result is reported in the Research Square preprint. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: LightGBM predicted delayed entry into tuberculosis-designated care from 60,778 surveillance records and ranked geographic-accessibility measures among leading predictors.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LightGBM predicted delayed entry into tuberculosis-designated care from 60,778 surveillance records and ranked geographic-accessibility measures among leading predictors.
Test and result
A LightGBM prediction model using 60,778 tuberculosis surveillance records found delayed entry to TB-designated care common and identified geographic accessibility measures among leading predictors; it reports a held-out test cohort and calls for external validation.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.21203/rs.3.rs-10074278/v1 · https://europepmc.org/article/PPR/PPR1260892 · https://www.researchsquare.com/article/rs-10074278/v1.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Melanoma analyses link mutation subtypes to distinct tumor immune signatures

Source check: Source checked. Not independent scientific confirmation.

Finding: Cross-cohort AI analysis linked melanoma mutation subtypes to distinct immune landscapes and implicated CCNE1; cell-line knockdown provided functional support.

Technology: Machine learning survival models, including regularized regression and tree-based methods

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study compared learned survival models to identify an inflammatory signature and nominate CCNE1 from melanoma datasets.
Test and result
Four expression cohorts plus cell-line functional validation; retrospective data and in-vitro support do not establish treatment efficacy.
Limitation / still unknown
The four-cohort analyses and cell-line knockdown do not establish treatment efficacy in patients.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13309799/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC13309799/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC13309799/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

One scale to rule them all: interpretable multi-scale Deep Learning for predicting cell survival after proton and carbon ion irradiation.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: deep-learning model predicts cell survival/RBE10 using the PIDE radiation dataset; overall MAPE 9.5%, mean RMSE .3; compared with LEM/MKM on V79 and HSG data, model RMSE about .1 in the reported comparisons; nanodosimetric descriptor had highest attention across endpoints. This is retrospective data-driven radiation-biology prediction, not a new wet-lab measurement.

Technology: Learned model links dosimetric features to measured cell-survival outcomes and reports a specific, bounded radiation-biology result; no new assay or treatment efficacy.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned model links dosimetric features to measured cell-survival outcomes and reports a specific, bounded radiation-biology result; no new assay or treatment efficacy.
Limitation / still unknown
Methods/Results: deep-learning model predicts cell survival/RBE10 using the PIDE radiation dataset; overall MAPE 9.5%, mean RMSE .3; compared with LEM/MKM on V79 and HSG data, model RMSE about .1 in the reported comparisons; nanodosimetric descriptor had highest attention across endpoints. This is retrospective data-driven radiation-biology prediction, not a new wet-lab measurement.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2026.1807677/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AI-derived cardiothoracic structures and coronary artery calcification volumes on non-contrast CT as predictors of mortality in pulmonary hypertension.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: AI-derived CT volumes are associated with mortality in the retrospective 620-person pulmonary-hypertension registry after clinical and REVEAL adjustment. This supports a prognostic association only; single-centre retrospective evidence does not establish causality or clinical benefit.

Technology: Previously validated deep-learning models derive cardiac-structure and coronary-calcium volumes from non-contrast CT; these AI-derived measures are the predictors tested.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Previously validated deep-learning models derive cardiac-structure and coronary-calcium volumes from non-contrast CT; these AI-derived measures are the predictors tested.
Limitation / still unknown
Regression-adjusted retrospective association analysis; single referral-centre cohort, no causal inference or demonstrated clinical benefit, and no independent validation in this cohort.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s44449-026-00005-2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Spatio-temporal epidemic forecasting with graph-based transformer.

Source check: Source checked. Not independent scientific confirmation.

Finding: Graph and transformer-based temporal forecasting models were compared on epidemic datasets from Spain and Brazil; the Results report dataset-specific forecast metrics, with simpler LinearTGCN strongest on one dataset and attention models competitive on the other.

Technology: Graph-based and transformer forecasting models were compared on epidemic time-series datasets from Spain and Brazil.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Graph-based and transformer forecasting models were compared on epidemic time-series datasets from Spain and Brazil.
Test and result
Graph and transformer-based temporal forecasting models were compared on epidemic datasets from Spain and Brazil; the Results report dataset-specific forecast metrics, with simpler LinearTGCN strongest on one dataset and attention models competitive on the other. Evaluation: The reported validation/comparison and its limits were checked; interpretation is restricted to the design and cohort described in reason.
Limitation / still unknown
Forecast comparisons are limited to the Spain and Brazil datasets; generalization to other epidemics or settings is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1186/s12942-026-00476-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Electronic health record-derived machine learning model for hypoglycemia risk prediction in type 2 diabetes mellitus patients: development and validation.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: EHR-derived machine-learning models predicted hypoglycemia risk among type-2 diabetes patients with a validation cohort.

Technology: BackgroundHypoglycemia is a serious complication of diabetes. Early recognition of hypoglycemia can improve clinical prognosis, however, traditional diagnostic tools are often limited. Machine learning offers a promising approach for predicting adverse outcomes in diabetic patients.ObjectiveThis study aims to develop and validate machine learning-based models to predict the risk of hypoglycemia in type 2 diabetes mellitus (T2DM) patients.MethodsA cohort study design was employed. Clinical…

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Four classifiers applied electronic health record variables to estimate hypoglycemia risk in adults with type 2 diabetes; the study compared them using a 70:30 internal train-validation split.
Limitation / still unknown
The 831-patient EHR dataset was randomly split 70:30 within one study setting; no external validation or clinical-impact evaluation is reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13435572/ · https://europepmc.org/article/MED/42218408

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Determinants of protein corona adsorption and abundance revealed by interpretable machine learning across nanoparticle systems.

Source check: Source checked. Not independent scientific confirmation.

Finding: Interpretable deep neural networks were used to predict protein-corona adsorption and relative abundance across nanoparticle systems. The paper reports an analysis of protein-corona patterns; its computational predictions are not experimental validation of downstream therapeutic or biological effects. The result is reported in the Scientific Reports. The study does not establish generalization beyond its reported setting or prospective real-world benefit. Affiliation was not stated in the reviewed material.

Technology: Interpretable deep neural networks predicted protein-corona adsorption and relative abundance across nanoparticle systems.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Interpretable deep neural networks predicted protein-corona adsorption and relative abundance across nanoparticle systems.
Test and result
The primary publisher Methods/Results support the AI-enabled result: Unverified Haiku assessment. Study using large-scale deep neural networks, with interpretability analysis, to predict protein corona adsorption and relative abundance across nanoparticle systems, a distinct field result with a learned-model AI role.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-026-55258-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AI-assisted preoperative planning improves component sizing accuracy in Oxford unicompartmental knee arthroplasty: a retrospective cohort study with 24-month follow-up.

Source check: Source checked. Not independent scientific confirmation.

Finding: Specific clinical planning accuracy result measured against operative sizing; retrospective cohort qualifies narrowly. Do not imply improved patient outcomes; modest HSS difference and single-center/nonrandomized design are material limits.

Technology: 190 knees (AI-assisted n=94, conventional n=96), 3D patient-specific CT bone model and deep-learning landmarks used in preoperative sizing plan; postoperative component size compared with intraoperative sizing. Femoral accuracy 77/94 (81.91%), tibial 80/94 (85.12%), both 66/94 (70.21%). HSS difference 3.18 at 24 months is modest and retrospective nonrandomized design does not establish clinical benefit.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
190 knees (AI-assisted n=94, conventional n=96), 3D patient-specific CT bone model and deep-learning landmarks used in preoperative sizing plan; postoperative component size compared with intraoperative sizing. Femoral accuracy 77/94 (81.91%), tibial 80/94 (85.12%), both 66/94 (70.21%). HSS difference 3.18 at 24 months is modest and retrospective nonrandomized design does not establish clinical benefit.
Test and result
190 knees (AI-assisted n=94, conventional n=96), 3D patient-specific CT bone model and deep-learning landmarks used in preoperative sizing plan; postoperative component size compared with intraoperative sizing. Femoral accuracy 77/94 (81.91%), tibial 80/94 (85.12%), both 66/94 (70.21%). HSS difference 3.18 at 24 months is modest and retrospective nonrandomized design does not establish clinical benefit.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13018-026-06935-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Listening to MS: AI-assisted speech analysis for diagnosis and fatigue prediction (COMMITMENT).

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Prospective speech recordings from 50 relapsing-MS patients and 20 controls were analyzed for fatigue and case status; leave-one-speaker-out models showed specificity 0.68–0.94 but lower sensitivity (0.38–0.90), and case/control sensitivity was 0.30 at specificity 0.90. This is exploratory speech-biomarker evidence, not a diagnostic test.

Technology: Automated speech-feature extraction and ML models classify MS/control status and regress fatigue ratings.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Automated speech-feature extraction and ML models classify MS/control status and regress fatigue ratings.
Limitation / still unknown
Prospective observational cohort with speaker-held-out evaluation; small sample, low sensitivity and exploratory endpoints noted.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2026.1721274/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Integrative multi-omics, machine learning, and experimental validation reveal that TPM1 suppresses M2 macrophage polarization and enhances chemosensitivity in acute myeloid leukemia.

Source check: Source checked. Not independent scientific confirmation.

Finding: ML feature selection across AML transcriptomic/proteomic datasets prioritized TPM1; cell-line perturbations and a xenograft model then showed TPM1-linked AML growth, Ara-C sensitivity, and M2 macrophage effects. MeRIP/RIP, reporter, and mRNA-stability assays support an FTO/m6A mechanism destabilizing TPM1. This is preclinical evidence, not clinical efficacy.

Technology: LASSO, SVM-RFE and random-forest analyses of AML transcriptomic/proteomic datasets selected TPM1 as a candidate, which was then experimentally perturbed.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
LASSO, SVM-RFE and random-forest analyses of AML transcriptomic/proteomic datasets selected TPM1 as a candidate, which was then experimentally perturbed.
Test and result
TPM1 overexpression reduced AML cell growth, increased apoptosis/oxidative stress and Ara-C sensitivity, and attenuated M2 polarization; mechanistic assays support FTO-dependent m6A-mediated TPM1 mRNA destabilization.
Limitation / still unknown
Follow-up includes AML cell-line perturbations, macrophage co-culture, MeRIP/RIP and mRNA-stability assays, and xenograft groups of six mice; evidence remains preclinical.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13424268/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Multi-Output Machine Learning for Prediction of Postoperative Outcomes After Cardiac Surgery Using Patient Blood Management Biomarkers.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A decision-tree regressor predicted multiple postoperative cardiac-surgery outcomes using patient blood-management and clinical variables. The paper reports held-out evaluation across five random states but only 513 complete cases from a 1,414-patient source cohort; authors describe proof of concept requiring external validation.

Technology: A decision-tree regressor predicts multiple postoperative outcomes from patient blood-management and clinical variables.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
A decision-tree regressor predicts multiple postoperative outcomes from patient blood-management and clinical variables.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcm/jcm-15-04221/article_deploy/jcm-15-04221.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Clinical Reliability of AI-Based Cephalometric Analysis Using WebCeph: A Comparative Agreement Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: WebCeph AI cephalometric measurements compared with clinician measurements using agreement analysis. Results quantifies agreement/reliability for measured landmarks; comparison is clinical measurement validation, not treatment outcome.

Technology: WebCeph AI cephalometric measurements compared with clinician measurements using agreement analysis.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
WebCeph AI cephalometric measurements compared with clinician measurements using agreement analysis.
Limitation / still unknown
On 100 radiographs, automated values differed significantly from manual tracing for six of seven measures, with inconsistent and often poor agreement; the study does not establish clinical interchangeability.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13257557/ · https://europepmc.org/article/MED/42279017

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning assisted classification of periodontal health and periodontitis using alveolar bone loss measurements on bitewing radiographs.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: 1,537 of 2,162 bitewing radiographs included; alveolar bone loss measured and calibrated by examiners (κ=.94); Random Forest validation accuracy 96.4%, unseen test accuracy 92%, AUC .99/.97.

Technology: Specific periodontal health classifier with a previously unseen test dataset; screening utility remains prospective.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Specific periodontal health classifier with a previously unseen test dataset; screening utility remains prospective.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/oral-health/articles/10.3389/froh.2026.1814025/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

From algorithm to verification: based on network toxicology and machine learning, the immunomodulatory role of IGFBP1/MKI67/C9 in perfluorooctanoic acid-induced osteoarthritis was discovered, and a di

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Methods 2.1–2.7: CTD 379 PFOA targets, OA expression cohorts; logistic/nomogram diagnostic model; Results: AUC .903, qRT-PCR validation used n=6/group and triplicate technical replicates. Supports a candidate biomarker association, not causal PFOA mechanism.

Technology: Machine-learning and network-toxicology analyses prioritize candidate genes; qRT-PCR provides preliminary experimental support.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning and network-toxicology analyses prioritize candidate genes; qRT-PCR provides preliminary experimental support.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2026.1700638/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A closed-loop framework for perovskite precursor additive discovery

Source check: Source checked. Not independent scientific confirmation.

Finding: LEAP prioritized perovskite additives that improved measured device efficiency in a bounded validation study.

Technology: Perovskite-RL encoded literature-derived descriptors for Bayesian acquisition followed by expert review and experimental feedback.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Perovskite-RL encoded literature-derived descriptors for Bayesian acquisition followed by expert review and experimental feedback.
Test and result
Three expert-guided selection rounds and actual device readouts; proof of concept in the tested setup.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2605.20242v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Explainable prediction of MDR/RR-TB in tuberculosis-diabetes mellitus multimorbidity: a machine learning model developed and validated in a dual-center study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A random forest predicts MDR/RR tuberculosis among patients with TB-diabetes multimorbidity in a dual-center retrospective cohort (n=245); ten-fold CV AUC was 0.818 with recall 0.611, and calibration/DCA were reported. Authors appropriately frame it as triage, not diagnosis.

Technology: Seven ML models are fit and explained with SHAP using clinical features.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Seven ML models are fit and explained with SHAP using clinical features.
Limitation / still unknown
10-fold cross-validation, calibration and DCA; moderate recall and stated triage limitation recorded.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12879-026-13480-7

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Evaluating model generalizability for suicide attempt risk prediction: traditional machine vs deep learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Cross-site validation found limited transfer of an established suicide-risk model, with defined tradeoffs in an observed EHR outcome.

Technology: The study replicated an existing learned risk score and compared tabular ML and deep models across clinical encounter settings.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study replicated an existing learned risk score and compared tabular ML and deep models across clinical encounter settings.
Limitation / still unknown
More than 750,000 retrospective visits and observed 90-day outcomes in an external health system. No intervention efficacy was established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s44184-026-00209-2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting worsening visual field via masked dual stage attention recurrent neural network.

Source check: Source checked. Not independent scientific confirmation.

Finding: A masked dual-stage attention recurrent neural network predicted future visual fields in glaucoma from 154,105 examinations, with severity-specific held-out comparisons against observed values.

Technology: A masked dual-stage attention recurrent neural network predicted future visual fields in glaucoma from 154,105 examinations, with severity-specific held-out comparisons against observed values.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A masked dual-stage attention recurrent neural network predicted future visual fields in glaucoma from 154,105 examinations, with severity-specific held-out comparisons against observed values.
Test and result
Held-out visual-field error metrics were reported across glaucoma severity groups; the retrospective evaluation does not establish prospective clinical benefit or deployment robustness.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-026-51152-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Postoperative Anterior Chamber Depth, Intraocular Lens Tilt, and Decentration Using an Internally Validated Machine Learning Model.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A prospective 49-patient cataract study used PLS regression and Random Forest to predict postoperative IOL position from preoperative biometry, reporting internal OOB error estimates. This is small-sample internal validation only.

Technology: PLS regression and random forest predict postoperative intraocular-lens position from preoperative biometry.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PLS regression and random forest predict postoperative intraocular-lens position from preoperative biometry.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s40123-026-01387-0

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Discovering non-linear dynamics of miRNAs in Alzheimer's disease-related cognitive impairment: a cross-species approach with explainable machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: cross-species human serum and Alzheimer mouse hippocampal miRNA expression datasets; mRMR selection and nonlinear SVM-RBF/kNN compared with linear classifiers, SHAP interpretation. Nonlinear models outperform linear; MCI vs controls AUC .844; miR-155-5p stable primary driver, miR-339-5p/455-5p context-dependent; reported mouse expression changes. Retrospective observational/model-derived relations; cross-species/correlation limits. Methods/Results: cross-species human serum and Alzheimer mouse hippocampal miRNA expression datasets; mRMR selection and nonlinear SVM-RBF/kNN compared with linear classifiers, SHAP interpretation. Testing: Methods/Results: cross-species human serum and Alzheimer mouse hippocampal miRNA expression datasets; mRMR selection and nonlinear SVM-RBF/kNN compared with linear classifiers, SHAP interpretation. Retrospective observational/model-derived relations; cross-species/correlation limits.. Specific cross-species miRNA pattern and nonlinear operating-window result are grounded in measured expression datasets and learned model interpretation. No causal mechanism or prospective screening performance established. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: cross-species human serum and Alzheimer mouse hippocampal miRNA expression datasets; mRMR selection and nonlinear SVM-RBF/kNN compared with linear classifiers, SHAP interpretation. Nonlinear models outperform linear; MCI vs controls AUC .844; miR-155-5p stable primary driver, miR-339-5p/455-5p context-dependent; reported mouse expression changes. Retrospective observational/model-derived relations; cross-species/correlation limits.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
cross-species human serum and Alzheimer mouse hippocampal miRNA expression datasets; mRMR selection and nonlinear SVM-RBF/kNN compared with linear classifiers, SHAP interpretation. Nonlinear models outperform linear; MCI vs controls AUC .844; miR-155-5p stable primary driver, miR-339-5p/455-5p context-dependent; reported mouse expression changes. Retrospective observational/model-derived relations; cross-species/correlation limits.
Test and result
cross-species human serum and Alzheimer mouse hippocampal miRNA expression datasets; mRMR selection and nonlinear SVM-RBF/kNN compared with linear classifiers, SHAP interpretation. Nonlinear models outperform linear; MCI vs controls AUC .844; miR-155-5p stable primary driver, miR-339-5p/455-5p context-dependent; reported mouse expression changes. Retrospective observational/model-derived relations; cross-species/correlation limits.
Limitation / still unknown
Specific cross-species miRNA pattern and nonlinear operating-window result are grounded in measured expression datasets and learned model interpretation. No causal mechanism or prospective screening performance established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13062-026-00819-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development of an automated machine learning-based prediction model and interactive tool for blood transfusion requirements in patients with severe traumatic brain injury.

Source check: Source checked. Not independent scientific confirmation.

Finding: An automated ML framework predicted blood-transfusion requirements in severe traumatic-brain-injury patients; the model was assessed on internal held-out data.

Technology: Methods compare AutoML models and validation; Results report transfusion-demand prediction performance and feature analysis.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods compare AutoML models and validation; Results report transfusion-demand prediction performance and feature analysis.
Test and result
Methods compare AutoML models and validation; Results report transfusion-demand prediction performance and feature analysis.
Limitation / still unknown
Methods compare AutoML models and validation; Results report transfusion-demand prediction performance and feature analysis.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13167484/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Evaluation of basic life support practice skills based on artificial intelligence technology: system construction and evaluation equivalence validation.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: An AI system assessed basic-life-support practice scores with reported equivalence validation against standard scoring.

Technology: BackgroundBasic life support (BLS) skills are the core competencies of healthcare professionals in responding to emergency situations. Traditional manual assessments suffer from subjective bias, low efficiency, and are affected by examiner fatigue. Artificial intelligence (AI) technology brings innovative opportunities for skill assessment, but currently there is a lack of mature and validated BLS AI assessment systems.AimTo construct a basic life support practice skills evaluation system…

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The system combines pose estimation, spatiotemporal graph analysis, object detection, speech recognition and an SVM to score basic-life-support practice; its scores were compared with examiner assessments in 85 newly qualified nurses.
Limitation / still unknown
The paper reports the relevant cohort/image/field dataset, comparator or reference, and validation/test evaluation; the judgment does not imply prospective clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13235214/ · https://europepmc.org/article/MED/42050498

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Explainable machine learning differentiates necrotizing fasciitis and osteomyelitis via routine blood biomarkers.

Source check: Source checked. Not independent scientific confirmation.

Finding: Routine blood markers distinguished two confirmed infections in a retrospective external cohort. The result is reported in the npj Digital Medicine. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: The selected LightGBM model integrated blood biomarkers and was evaluated in data from a geographically related external center.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The selected LightGBM model integrated blood biomarkers and was evaluated in data from a geographically related external center.
Test and result
Routine blood markers distinguished two confirmed infections in a retrospective external cohort.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41746-026-02686-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Machine Learning-Based Framework for Risk Recognition and Reliability Evaluation in City Expressway Ramp Merging.

Source check: Source checked. Not independent scientific confirmation.

Finding: Methods describe risk labels and grouped split; Results report model comparison, stability and probability reliability. Testing: Methods describe risk labels and grouped split; Results report model comparison, stability and probability reliability.. Machine-learning models estimated risk during city expressway ramp merging from pre-event observations, with grouped data splits designed to limit leakage. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Methods describe risk labels and grouped split; Results report model comparison, stability and probability reliability.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe risk labels and grouped split; Results report model comparison, stability and probability reliability.
Test and result
Methods describe risk labels and grouped split; Results report model comparison, stability and probability reliability.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13165751/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Generative AI-driven de novo design of high-affinity, epitope-specific antibodies targeting the CD93-IGFBP7 axis for cancer immunotherapy.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Generative AI contributed a candidate antibody that showed measured target binding/blockade and tumor-model activity; the evidence is preclinical and does not establish human efficacy.

Technology: A generative AI model designed candidates for a target interface; the lead was selected and tested experimentally. No sequence or operational design details reproduced.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A generative AI model designed candidates for a target interface; the lead was selected and tested experimentally. No sequence or operational design details reproduced.
Limitation / still unknown
Biophysical binding, cell-based blocking and tumor-model experiments tested the candidate. This is preclinical evidence, not human or clinical validation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://academic.oup.com/abt/article/9/3/323/8664001 · https://europepmc.org/article/MED/42592032

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Quantitative comparison of explainable AI methods for interpreting deep learning-based classification of 3D gait kinematics.

Source check: Source checked. Not independent scientific confirmation.

Finding: Across three gait datasets, ResNet/LSTM/InceptionTime classifiers predict diagnostic group labels and four XAI methods identify recurring joint-angle features; Integrated Gradients is most robust and feature masking improves accuracy. This is a specific clinical gait-data result, but diagnostic cohort interpretation and exact reference standard are limited to existing datasets.

Technology: ResNet, LSTM and InceptionTime classify diagnostic groups from 3D gait kinematics; LIME, DeepLift, Integrated Gradients and sequential feature selection identify influential joint-angle features.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ResNet, LSTM and InceptionTime classify diagnostic groups from 3D gait kinematics; LIME, DeepLift, Integrated Gradients and sequential feature selection identify influential joint-angle features.
Test and result
Across three existing gait datasets, Integrated Gradients was the most robust explanation method in the reported comparisons, and feature masking improved classification accuracy. The findings are bounded to these datasets and reference labels.
Limitation / still unknown
The three gait datasets are existing clinical datasets with their own diagnostic labels; explanation robustness and feature masking do not establish external clinical validity or patient benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-026-46243-0 · https://europepmc.org/article/MED/41917155

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AI-assessed sarcopenia as an independent predictor of neoadjuvant chemotherapy outcomes in muscle-invasive bladder cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary source Methods/Results confirm the represented result: Retrospective single-centre study where AI-powered CT sarcopenia (skeletal muscle index) software predicts neoadjuvant chemotherapy response in muscle-invasive bladder cancer; a clinically specific retrospective assessment. The result is reported in the La radiologia medica. Affiliation was not stated in the reviewed material. It does not establish generalization beyond the reported setting or prospective real-world benefit.

Technology: Retrospective single-centre study where AI-powered CT sarcopenia (skeletal muscle index) software predicts neoadjuvant chemotherapy response in muscle-invasive bladder cancer

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Retrospective single-centre study where AI-powered CT sarcopenia (skeletal muscle index) software predicts neoadjuvant chemotherapy response in muscle-invasive bladder cancer
Test and result
The primary source Methods/Results confirm the represented result: Retrospective single-centre study where AI-powered CT sarcopenia (skeletal muscle index) software predicts neoadjuvant chemotherapy response in muscle-invasive bladder cancer; a clinically specific retrospective assessment.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11547-026-02203-2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Systems toxicology unravels multi-tiered carcinogenic networks of zearalenone in gastric cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding: An integrated toxicology analysis identified 157 shared zearalenone and gastric-cancer targets; eleven machine-learning algorithms selected six candidate genes. Docking and molecular-dynamics analyses were followed by gastric-cancer cell assays, which supported altered proliferation and migration in vitro. These data do not establish carcinogenicity in humans or in-vivo causal effects.

Technology: Eleven machine-learning algorithms screened shared disease and chemical-target genes to select six candidate driver genes; computational pathway and binding analyses were followed by in-vitro cell experiments.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Eleven machine-learning algorithms screened shared disease and chemical-target genes to select six candidate driver genes; computational pathway and binding analyses were followed by in-vitro cell experiments.
Test and result
The primary abstract and Results report 157 common zearalenone and gastric-cancer targets, six machine-learning-selected genes with reported AUC above 0.85, docking and 100-ns molecular-dynamics analyses, and in-vitro MKN-45 cell assays showing increased proliferation and migration at the tested low-concentration conditions. The experimental evidence is cellular and in vitro.
Limitation / still unknown
The cell assays are in-vitro evidence only; human gastric carcinogenicity and in-vivo causal effects were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2026.1762191/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Large language model detects previously undiagnosed heart failure with preserved ejection fraction in patients with metabolic-associated fatty liver disease: A multicenter cohort study.

Source check: Source checked. Not independent scientific confirmation.

Finding: MedGuide-14B applied to multicenter MAFLD EHRs identified additional HFpEF cases; blinded cardiologist review and longitudinal outcomes support the result, with retrospective design limits.

Technology: domain-tuned MedGuide-14B analyzed structured and free-text EHR data at three hospitals; 24,011 patients, 4,226 newly model-identified cases, and blinded cardiologist confirmation in 452/500 reviewed. Retrospective cohort; no prospective clinical-utility test.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
domain-tuned MedGuide-14B analyzed structured and free-text EHR data at three hospitals; 24,011 patients, 4,226 newly model-identified cases, and blinded cardiologist confirmation in 452/500 reviewed. Retrospective cohort; no prospective clinical-utility test.
Test and result
domain-tuned MedGuide-14B analyzed structured and free-text EHR data at three hospitals; 24,011 patients, 4,226 newly model-identified cases, and blinded cardiologist confirmation in 452/500 reviewed. Retrospective cohort; no prospective clinical-utility test.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001317

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Monitoring of dual-drug combination therapy in pediatric epilepsy patients: a machine learning model for simultaneous VPA-LEV concentration-dose prediction.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary abstract reports retrospective pediatric epilepsy concentration adequacy and individualized VPA/LEV dose predictions from 497 paired samples; this is a specific drug-monitoring application, without prospective utility evidence.

Technology: Dual-predictive machine-learning model for concentration adequacy and dose regression

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Classifies VPA/LEV concentration adequacy and regresses individualized daily doses from paired concentration samples.
Test and result
The retrospective dataset contained 497 paired concentration samples from 402 pediatric epilepsy patients; classification models predicted concentration adequacy and regression models recommended daily doses.
Limitation / still unknown
The retrospective concentration samples do not establish prospective dosing impact or improved patient outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13154482/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Assessing the Predictive Value of Kolmogorov-Arnold Networks for the No-Reflow Phenomenon in ST-Segment Elevation Myocardial Infarction: A Comparative Machine Learning Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: In a prospective single-center cohort of 890 patients undergoing primary PCI for STEMI, the study compared supervised models for predicting no-reflow. Kolmogorov-Arnold Network and XGBoost models reported AUC above 0.98 and F1 above 0.95, exceeding logistic-regression and decision-tree comparators in the study. External validation is still needed.

Technology: Kolmogorov-Arnold Network and XGBoost classifiers predicted no-reflow after primary PCI; ANOVA F-statistics selected features, and SHAP described model associations.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Kolmogorov-Arnold Network and XGBoost classifiers predicted no-reflow after primary PCI; ANOVA F-statistics selected features, and SHAP described model associations.
Test and result
The primary publisher abstract reports 890 consecutive STEMI patients in a prospective single-center study; KAN and XGBoost had AUC above 0.98 and F1 above 0.95, outperforming logistic regression and decision trees. The input full-text review confirms the cohort, reference and validation details; further multicenter external validation is required.
Limitation / still unknown
The evaluation is from one center; the source calls for multicenter external validation before generalizing performance or clinical risk-assessment use.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://archivestsc.com/jvi.aspx?un=TKDA-02730&volume=54&issue=3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A biology-based quality-diversity algorithm for drug repurposing in Alzheimer's disease using automated machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: A biology-based GNN/AutoML quality-diversity search recovered known drug–gene links and prioritized a drug-repurposing candidate for Alzheimer’s disease. Experimental efficacy was not tested.

Technology: A biology-based graph neural network and AutoML quality-diversity search ranks drug-repurposing hypotheses.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A biology-based graph neural network and AutoML quality-diversity search ranks drug-repurposing hypotheses.
Test and result
Evaluation recovered known drug–gene links and ranked a candidate computationally; experimental efficacy remains unestablished.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13040-026-00550-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Analytical Validation of Multimodal AI Test Predicting Breast Cancer Recurrence Risk (Ataraxis Breast RISK).

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A locked H&E-image and clinical-data AI test passed prespecified repeatability and reproducibility checks in 160 breast-cancer cases, including 94.7% agreement across two laboratories/scanners. This is analytical readiness evidence, not independent clinical utility evidence.

Technology: Kestrel vision-transformer patch embeddings are attention-aggregated, combined with clinical variables, and scored by an ensemble of Cox neural networks with calibration.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Kestrel vision-transformer patch embeddings are attention-aggregated, combined with clinical variables, and scored by an ensemble of Cox neural networks with calibration.
Limitation / still unknown
Prespecified analytical acceptance criteria were assessed using repeat runs, two operators, two labs/scanners, controls, perturbed covariates and a clinical-validation bridge; cohort was retrospective and largely ER-positive/HER2-negative, limiting generalization.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13073172/ · https://europepmc.org/article/MED/41975736

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Harnessing machine learning with auditory tests and demographic factors to forecast children's reading abilities in children living with and without HIV.

Source check: Source checked. Not independent scientific confirmation.

Finding: Longitudinal pediatric data compare continuous prediction and classification performance across HIV groups.

Technology: Machine learning uses auditory test and demographic features to forecast reading ability among children living with and without HIV

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine learning uses auditory test and demographic features to forecast reading ability among children living with and without HIV.
Test and result
Longitudinal pediatric data compare continuous prediction and classification performance across HIV groups. Evaluation: Cohort-specific predictions do not establish that HIV status causes the observed reading differences.
Limitation / still unknown
The reported cohort-specific prediction does not establish that HIV status causes the observed reading differences or that the model generalizes to other populations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13151343/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Colorectal cancer sidedness: prognostic implications and the predictive role of artificial intelligence.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary BMC article and Research Square V1 agree on a 71-patient retrospective CRC cohort and decision-tree prediction of laterality/prognostic outcomes. The small, single-center models have uneven/overfit results, so these are exploratory associations.

Technology: J48/C4.5 trees integrate clinicopathological data; LOOCV tests laterality and recurrence/mortality predictions.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
J48/C4.5 trees integrate clinicopathological data; LOOCV tests laterality and recurrence/mortality predictions.
Limitation / still unknown
Internal LOOCV only; poor recurrence model and small cohort are explicitly acknowledged. Observed sidedness associations are cohort-specific, not causal.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12911-026-03468-3 · https://europepmc.org/article/MED/42210356

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Micronucleus quantification from whole-slide haematology images using AI serves as a translatable pharmacodynamic biomarker for DNA damage response inhibitors.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Supervised deep learning quantifies micronuclei in whole-slide blood-smear images; measurements were compared with flow cytometry. The record is limited to the reported cohort, dataset or experiment and does not establish wider clinical or operational benefit.

Technology: Supervised deep learning quantifies micronuclei in whole-slide blood-smear images; measurements were compared with flow cytometry.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Supervised deep learning quantifies micronuclei in whole-slide blood-smear images; measurements were compared with flow cytometry.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-026-41458-7

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based prediction of PASI100 response to secukinumab in patients with psoriasis: a real-world study with SHAP interpretability analysis.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Real-world psoriasis cohort links a learned model to secukinumab treatment-response prediction.

Technology: Real-world psoriasis cohort links a learned model to secukinumab treatment-response prediction.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Real-world psoriasis cohort links a learned model to secukinumab treatment-response prediction.
Limitation / still unknown
This is a retrospective 11,134-patient dataset randomly split for testing; prediction of PASI100 response does not show that model-guided treatment improves outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12983332/ · https://europepmc.org/article/MED/41836943

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Incisional hernia prediction using machine learning models.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Retrospective midline-laparotomy model predicts 24-month incisional hernia with imaging confirmation; cross-validation is internal and does not establish external transportability.

Technology: XGBoost models predict 24-month incisional hernia after midline laparotomy using selected patient and clinical features.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
XGBoost models predict 24-month incisional hernia after midline laparotomy using selected patient and clinical features.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12911-026-03382-8

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Explainable Machine Learning Approaches Predict Frailty and Adverse Outcomes in Older Adults: Development and Validation with Two Longitudinal Cohorts.

Source check: Source checked. Not independent scientific confirmation.

Finding: Six ML models on baseline longitudinal-cohort predictors estimated Fried Frailty Phenotype at four years; 3,419 CHARLS participants and 1,017 CLHLS-HF validation participants. XGBoost best: internal AUC 0.934, external AUC 0.792; predicted-frail group associated with falls, hospitalization and disability. Testing: XGBoost predicted four-year frailty in older adults using CHARLS and a separate CLHLS-HF validation cohort; it achieved AUC 0.934 internally and 0.792 externally, and predicted frailty associated with subsequent adverse outcomes. These findings are bounded by cohort definitions and population context.. XGBoost predicted four-year frailty in older adults using CHARLS and a separate CLHLS-HF validation cohort; it achieved AUC 0.934 internally and 0.792 externally, and predicted frailty associated with subsequent adverse outcomes. These findings are bounded by cohort definitions and population context. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Six ML models on baseline longitudinal-cohort predictors estimated Fried Frailty Phenotype at four years; 3,419 CHARLS participants and 1,017 CLHLS-HF validation participants. XGBoost best: internal AUC 0.934, external AUC 0.792; predicted-frail group associated with falls, hospitalization and disability.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Six ML models on baseline longitudinal-cohort predictors estimated Fried Frailty Phenotype at four years; 3,419 CHARLS participants and 1,017 CLHLS-HF validation participants. XGBoost best: internal AUC 0.934, external AUC 0.792; predicted-frail group associated with falls, hospitalization and disability.
Test and result
XGBoost predicted four-year frailty in older adults using CHARLS and a separate CLHLS-HF validation cohort; it achieved AUC 0.934 internally and 0.792 externally, and predicted frailty associated with subsequent adverse outcomes. These findings are bounded by cohort definitions and population context.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcm/jcm-15-01812/article_deploy/jcm-15-01812.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Radiomics and deep learning fusion model based on multiphasic CT for predicting HER2 expression status in bladder urothelial carcinoma: a multicenter study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: 411 patients; tri-phasic CT radiomics and 2.5D/3D ResNet50 fusion against HER2 IHC, with independent internal and two-center external validation; external AUC 0.873. Retrospective proof of concept, not prospective treatment benefit.

Technology: Methods name the learned image/gene-expression model and its input-operation tied to the reported result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods name the learned image/gene-expression model and its input-operation tied to the reported result.
Limitation / still unknown
Independent test cohort(s) or assay validation reported as described in the reason; interpretation bounded to stated cohort.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://tau.amegroups.org/article/view/150131/html · https://europepmc.org/article/MED/41971125

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A novel statistical framework for quantifying risks and benefits of AI automation in screening mammography.

Source check: Source checked. Not independent scientific confirmation.

Finding: Mirai scores on 114,229 screening mammograms support a clinically explicit rule-out threshold trade-off framework; thresholds and missed-cancer counts are reported, with hypothetical radiologist catch-up assumptions kept explicit.

Technology: Mirai risk scores from 114,229 cases; 0.20 threshold rules out 75% but yields 121 net additional missed cancers under the hypothetical 30% catch-up scenario; 0.05 yields 36% reduction and modeled zero net increase. Retrospective single-site 2006–2023 analysis; thresholds are illustrative, not a clinical recommendation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Mirai risk scores from 114,229 cases; 0.20 threshold rules out 75% but yields 121 net additional missed cancers under the hypothetical 30% catch-up scenario; 0.05 yields 36% reduction and modeled zero net increase. Retrospective single-site 2006–2023 analysis; thresholds are illustrative, not a clinical recommendation.
Test and result
Mirai risk scores from 114,229 cases; 0.20 threshold rules out 75% but yields 121 net additional missed cancers under the hypothetical 30% catch-up scenario; 0.05 yields 36% reduction and modeled zero net increase. Retrospective single-site 2006–2023 analysis; thresholds are illustrative, not a clinical recommendation.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001231

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AI-based intelligent sensing detection of cybersecurity threats using multimodal sensor data in smart devices.

Source check: Source checked. Not independent scientific confirmation.

Finding: CNN-RNN-Transformer models classify smart-device cybersecurity threats from controlled/public multimodal sensor datasets; the reported model reaches AUC 0.96 and 23 ms edge inference. This is a specific device-threat detection evaluation; rationale omits operational details and makes no real deployment claim.

Technology: CNN, recurrent and transformer-based models classified cybersecurity threat categories from multimodal smart-device sensor data.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CNN, recurrent and transformer-based models classified cybersecurity threat categories from multimodal smart-device sensor data.
Test and result
CNN-RNN-Transformer models classify smart-device cybersecurity threats from controlled/public multimodal sensor datasets; the reported model reaches AUC 0.96 and 23 ms edge inference. This is a specific device-threat detection evaluation; rationale omits operational details and makes no real deployment claim. Evaluation: Original Methods/Results establish the cohort, reference/comparison and validation scope as summarized in reason.
Limitation / still unknown
The result is a dataset-specific smart-device threat classification evaluation; real-world deployment or operational security benefit is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-026-40614-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Adaptive Temporal Mixture of Experts for Predicting Stiffness Metrics From the Ocular Response Analyzer and Identifying Keratoconus.

Source check: Source checked. Not independent scientific confirmation.

Finding: AT-MoE uses LSTM experts with dynamic gating to predict Corvis stiffness from ORA waveforms/pressure-applanation data and classifies keratoconus. Development had 961 eyes/502 people with subject-separated 80/20 split; independent prospective validation improved AUROC from .9406 baseline LSTM to .9512. Limited validation and no Corvis stiffness ground truth in validation narrow interpretation. Development/test predictions improved versus LSTM and independent keratoconus discrimination AUROC reached .9512; no claim of clinical utility. Published in American Journal of Ophthalmology.

Technology: AT-MoE uses LSTM experts with dynamic gating to predict Corvis stiffness from ORA waveforms/pressure-applanation data and classifies keratoconus.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
AT-MoE uses LSTM experts with dynamic gating to predict Corvis stiffness from ORA waveforms/pressure-applanation data and classifies keratoconus.
Test and result
Development/test predictions improved versus LSTM and independent keratoconus discrimination AUROC reached .9512; no claim of clinical utility.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13183170/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A multicenter explainable machine learning analysis of autoimmune disease comorbidity in ankylosing spondylitis.

Source check: Source checked. Not independent scientific confirmation.

Finding: Methods 2.1: multicenter retrospective cross-sectional cohorts from three tertiary centers (2012–2025); two classification tasks, LightGBM and SHAP. Results: stable model performance and feature differences; systemic inflammatory/metabolic phenotype is an association, not prospective early detection. Learned LightGBM classification and SHAP produce clinical comorbidity classification and feature attribution; retrospective claims only. Testing: Methods 2.1: multicenter retrospective cross-sectional cohorts from three tertiary centers (2012–2025); two classification tasks, LightGBM and SHAP. Results: stable model performance and feature differences; systemic inflammatory/metabolic phenotype is an association, not prospective early detection.. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Learned LightGBM classification and SHAP produce clinical comorbidity classification and feature attribution; retrospective claims only.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned LightGBM classification and SHAP produce clinical comorbidity classification and feature attribution; retrospective claims only.
Test and result
Methods 2.1: multicenter retrospective cross-sectional cohorts from three tertiary centers (2012–2025); two classification tasks, LightGBM and SHAP. Results: stable model performance and feature differences; systemic inflammatory/metabolic phenotype is an association, not prospective early detection.
Limitation / still unknown
Methods 2.1: multicenter retrospective cross-sectional cohorts from three tertiary centers (2012–2025); two classification tasks, LightGBM and SHAP. Results: stable model performance and feature differences; systemic inflammatory/metabolic phenotype is an association, not prospective early detection.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2026.1775877/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning and Blood-Targeted Proteomics Enable Early Prediction and Etiological Discrimination of Hypertensive Pregnancy Disorders.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: An 18-protein SVM based on first-trimester serum proteomics discriminated later preeclampsia from other groups and the FMF screen in this nested case-control study; the paper reports 94% sensitivity and 100% specificity in the study cohort. The model was tuned with repeated five-fold CV and evaluated with ten-fold CV, but its relationship to a separate pilot cohort is not a clinical external validation.

Technology: SVM (polynomial/radial kernels), XGBoost and MLP classified first-trimester LC-MRM-MS protein profiles; the SVM result was compared with FMF screening.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
SVM (polynomial/radial kernels), XGBoost and MLP classified first-trimester LC-MRM-MS protein profiles; the SVM result was compared with FMF screening.
Limitation / still unknown
Particle-swarm hyperparameter tuning used five-times repeated five-fold CV; final model performance used ten-fold CV in the same study cohort. The separate pilot cohort supported biomarker analytical concordance, not clinical model validation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/ijms/ijms-27-01402/article_deploy/ijms-27-01402.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

PatchSight-ImmuneMap-LifeSpan as a unified AI framework for breast cancer diagnosis, immune profiling and prognostic prediction.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The framework reports specific breast-cancer diagnosis, TIL-detection and survival-prediction evaluations on public datasets, qualifying as task-specific retrospective validation. No prospective clinical validation or new biology is shown, and reported performance varies by dataset/task.

Technology: Methods/Results evaluate InceptionResNetV2 on BreakHis, Faster R-CNN on LYSTO and survival prediction on public datasets; test metrics include survival C-index around 0.74. These are benchmark-dataset task evaluations, not deployment evidence.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods/Results evaluate InceptionResNetV2 on BreakHis, Faster R-CNN on LYSTO and survival prediction on public datasets; test metrics include survival C-index around 0.74. These are benchmark-dataset task evaluations, not deployment evidence.
Limitation / still unknown
Evaluations use public retrospective datasets; performance varies by diagnostic, immune-profiling and survival task, with no prospective clinical validation or new biological finding reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s12672-026-04491-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Environmental Enrichment and Agonistic Behavior in Post-Weaning Pigs: A Pilot Study Using Artificial Intelligence.

Source check: Source checked. Not independent scientific confirmation.

Finding: YOLOv8 object detection tracked pig behaviour in a pilot enrichment trial, enabling comparison of agonistic behaviour between enriched and control conditions.

Technology: YOLOv8 object detection tracked pig behaviour in a pilot enrichment trial, enabling comparison of agonistic behaviour between enriched and control conditions.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
YOLOv8 object detection tracked pig behaviour in a pilot enrichment trial, enabling comparison of agonistic behaviour between enriched and control conditions.
Test and result
The detector reached mAP50 0.953; enrichment was associated with reduced agonistic behaviour in the pilot cohort, whose size and setting limit generalization.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/biology/biology-15-00255/article_deploy/biology-15-00255.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Parameter-efficient fine-tuning enables scalable transfer of regulatory sequence models to novel contexts.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Beyond transfer efficiency benchmarks, the study applies learned regulatory model to a specific fine-mapped variant with cell-type-resolved predicted ITGA4 effects, a concrete biological variant interpretation. Label as computational prediction pending direct functional validation.

Technology: Locon4, a parameter-efficient fine-tuning method for the pretrained Borzoi regulatory DNA sequence model

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Locon4 parameter-efficient fine-tuning transfers the Borzoi regulatory-sequence model to tissue-specific expression prediction.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13059-026-03935-0

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Explainable AI-Enhanced Ensemble Protocol Using Gradient-Boosted Models for Zero-False-Alarm Seizure Detection from EEG.

Source check: Source checked. Not independent scientific confirmation.

Finding: An explainable EEG ensemble detected seizures under a zero-false-alarm constraint, with patient-independent evaluation and external Siena testing; external performance declined.

Technology: Methods describe LOSO evaluation and calibration; Results report zero-false-alarm performance and Siena external validation with lower AUC/specificity.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe LOSO evaluation and calibration; Results report zero-false-alarm performance and Siena external validation with lower AUC/specificity.
Test and result
Methods describe LOSO evaluation and calibration; Results report zero-false-alarm performance and Siena external validation with lower AUC/specificity.
Limitation / still unknown
Methods describe LOSO evaluation and calibration; Results report zero-false-alarm performance and Siena external validation with lower AUC/specificity.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12899994/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Neuro-Geometric Graph Transformers with Differentiable Radiographic Geometry for Spinal X-Ray Image Analysis.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Differentiable radiographic geometry and graph transformer analyze spinal X-rays across six diagnosis categories. Results evaluates classification on patient radiographs (1,613 cases as reported); study-data validation, not broad clinical utility.

Technology: Differentiable radiographic geometry and graph transformer analyze spinal X-rays across six diagnosis categories.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Differentiable radiographic geometry and graph transformer analyze spinal X-rays across six diagnosis categories.
Limitation / still unknown
The reported classification uses the study’s 1,613 radiographs and six diagnosis categories; no independent external clinical validation or patient-outcome evaluation is reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12942243/ · https://europepmc.org/article/MED/41745424

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Neural Efficiency and Attentional Instability in Gaming Disorder: A Task-Based Occipital EEG and Machine Learning Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Five machine-learning classifiers classified gaming-disorder EEG from occipital task recordings in 15 affected and 15 control participants using subject-independent leave-one-subject-out cross-validation.

Technology: Five machine-learning classifiers classified gaming-disorder EEG from occipital task recordings in 15 affected and 15 control participants using subject-independent leave-one-subject-out cross-validation.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Five machine-learning classifiers classified gaming-disorder EEG from occipital task recordings in 15 affected and 15 control participants using subject-independent leave-one-subject-out cross-validation.
Test and result
Performance was evaluated across five LOSO folds with held-out subjects; the 30-person sample and task-specific EEG design limit diagnostic inference.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/bioengineering/bioengineering-13-00152/article_deploy/bioengineering-13-00152.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AI-driven opportunistic screening for carotid atheroma calcification on dental radiographs.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: FastViT classified panoramic dental radiographs for carotid atheroma calcification; screening covered 19,205 records, with accuracy and recall around 0.88 on a 694-image analytical dataset. This is imaging detection, not confirmed cardiovascular diagnosis.

Technology: FastViT classifies panoramic dental radiographs for carotid atheroma calcification and localizes positive findings.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
FastViT classifies panoramic dental radiographs for carotid atheroma calcification and localizes positive findings.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1590/1807-3107bor-2026.vol40.088

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Integrating multi-omics data and machine learning to identify endocrine disrupting chemicals targeting key ccRCC-related genes.

Source check: Source checked. Not independent scientific confirmation.

Finding: Multiomics analysis combines ML prognostic gene selection with database-derived EDC–gene links in ccRCC. The AI role is principally selection of a prognostic signature; chemical targeting links are computational/database associations, not demonstrated causal exposure effects.

Technology: TCGA/GEO expression data, CTD-derived EDC links, 101 ML combinations and 10-fold CV select a five-gene prognostic signature; qPCR on four genes in eight paired samples. EDC exposure causality was not tested.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
TCGA/GEO expression data, CTD-derived EDC links, 101 ML combinations and 10-fold CV select a five-gene prognostic signature; qPCR on four genes in eight paired samples. EDC exposure causality was not tested.
Test and result
TCGA/GEO expression data, CTD-derived EDC links, 101 ML combinations and 10-fold CV select a five-gene prognostic signature; qPCR on four genes in eight paired samples. EDC exposure causality was not tested.
Limitation / still unknown
Methods/Results: TCGA/GEO expression data, CTD-derived EDC links, 101 ML combinations and 10-fold CV select a five-gene prognostic signature; qPCR on four genes in eight paired samples. EDC exposure causality was not tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s12672-025-04138-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Subclinical Atrial Fibrillation Prediction in Patients with CIED by a Novel Deep Learning Framework.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: ResKAN-Attention predicted device-detected subclinical AF in a retrospective cohort of 124 pacemaker patients, with five-fold CV AUC 0.837 and AUC 0.788 in a separately enrolled prospective 35-patient cohort. This is a small single-center validation, not evidence of clinical utility.

Technology: Dual-path KAN/MLP with cross-attention trained on 27 routine features, with preprocessing fit inside folds and SMOTE restricted to training data. Primary retrospective cohort 124 patients; separate prospective 35 patients with six-month monitoring for external validation; AUC 0.788.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Dual-path KAN/MLP with cross-attention trained on 27 routine features, with preprocessing fit inside folds and SMOTE restricted to training data. Primary retrospective cohort 124 patients; separate prospective 35 patients with six-month monitoring for external validation; AUC 0.788.
Limitation / still unknown
ResKAN-Attention predicted device-detected subclinical AF in a retrospective cohort of 124 pacemaker patients, with five-fold CV AUC 0.837 and AUC 0.788 in a separately enrolled prospective 35-patient cohort. This is a small single-center validation, not evidence of clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcdd/jcdd-13-00018/article_deploy/jcdd-13-00018.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prioritizing missense mutations via a deep learning phosphorylation prediction model.

Source check: Source checked. Not independent scientific confirmation.

Finding: DeepMEP predicts missense mutation effects on phosphorylation from flanking sequence and reports enrichment of high-effect predictions among Alzheimer and Parkinson disease GWAS loci; this is a computational prioritization result, not experimental confirmation of each variant. The result is reported in Human Genomics. Affiliation was not available in checked primary metadata; no organization is assigned. The study does not establish generalization beyond its reported setting or prospective real-world benefit. Affiliation was not stated in the reviewed material.

Technology: Methods describe ESM2 sequence embeddings, protein-level train/validation/test splitting and performance evaluation

Family ESM / ESMFold: profile · models page

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Methods describe ESM2 sequence embeddings, protein-level train/validation/test splitting and performance evaluation
Test and result
DeepMEP predicts missense mutation effects on phosphorylation from flanking sequence and reports enrichment of high-effect predictions among Alzheimer and Parkinson disease GWAS loci; this is a computational prioritization result, not experimental confirmation of each variant.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12859868/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Integration of peripheral blood-based systemic inflammatory indices and retinal imaging using interpretable machine learning for predicting anti-VEGF treatment response in macular edema secondary to retinal vein occlusion.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Interpretable learning predicted anti-VEGF response using inflammatory indices and retinal imaging in macular edema.

Technology: Interpretable learning predicted anti-VEGF response using inflammatory indices and retinal imaging in macular edema.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Interpretable learning predicted anti-VEGF response using inflammatory indices and retinal imaging in macular edema.
Limitation / still unknown
The 202-patient study is single-center and retrospective with an internal test split; predicted anti-VEGF response does not establish that model-guided care improves outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12794559/ · https://europepmc.org/article/MED/41532151

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Enhancing landslide detection in Western Ghats of Kerala, India with deep learning and Explainable AI.

Source check: Source checked. Not independent scientific confirmation.

Finding: An attention-based multiscale U-Net mapped landslide areas in Kerala from high-resolution satellite images before and after events. It performed best among the compared segmentation models; the work presents an initial inventory rather than a validated warning system.

Technology: An attention multiscale U-Net segments landslide areas from pre- and post-event high-resolution satellite imagery.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
An attention multiscale U-Net segments landslide areas from pre- and post-event high-resolution satellite imagery.
Test and result
The model was evaluated on Kerala Western Ghats imagery for landslide inventory mapping; this does not demonstrate prospective warning or hazard reduction.
Limitation / still unknown
The mapping result is based on historical Kerala imagery and the reported landslide inventory; it does not test prospective warnings or reductions in landslide harm.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1038/s41598-025-33065-9 · https://europepmc.org/article/MED/41461880

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Calibrated, explainable machine learning on routine laboratory data to characterize diagnostic assignment patterns in rheumatic diseases: a retrospective study of 12,085 patients.

Source check: Source checked. Not independent scientific confirmation.

Finding: Calibrated supervised models on routine laboratory data characterized diagnostic assignments across 12,085 rheumatology patients and reported disease-specific classification and feature patterns. The result is reported in BMC Rheumatology. Affiliation was not available in checked primary metadata; no organization is assigned. The study does not establish generalization beyond its reported setting or prospective real-world benefit. Affiliation was not stated in the reviewed material.

Technology: Methods identify the retrospective dataset and seven-disease modeling

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods identify the retrospective dataset and seven-disease modeling
Test and result
Calibrated supervised models on routine laboratory data characterized diagnostic assignments across 12,085 rheumatology patients and reported disease-specific classification and feature patterns.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12849087/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A temporal-aware machine learning framework enables microbial community dynamics prediction with personalized precision.

Source check: Source checked. Not independent scientific confirmation.

Finding: MicroProphet, a time-aware Transformer with attention, forecasts subject-specific microbial abundance trajectories from incomplete longitudinal observations without data imputation. It uses only the first 30% of observed time points and was evaluated across synthetic communities, human gut microbiomes, infant gut development and corpse decomposition. The abstract reports cross-ecosystem predictive performance but gives no cohort sizes or numeric error measures; clinical or forensic decision outcomes were not tested.

Technology: MicroProphet, a time-aware Transformer architecture with attention for subject-specific microbial trajectory forecasting.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A temporal Transformer reconstructs subject-specific abundance trajectories from sparse time series, using attention to capture transitions.
Test and result
The model forecasts trajectories using the first 30% of observations and is evaluated across synthetic communities, human gut microbiomes, infant gut development and corpse decomposition; the abstract reports cross-ecosystem generalization without numeric performance metrics.
Limitation / still unknown
The reviewed abstract gives no sample counts or numeric error estimates. It does not demonstrate improved health outcomes, intervention timing, or forensic casework performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12751615/ · https://europepmc.org/article/MED/41462375

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

TeroACT: A terpenoid bioactivity landscape and discovery platform.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Original full text describes TeroACT’s learned terpenoid ranking and reports experimental testing of selected candidates in melanoma models. This supports a narrow AI-guided candidate-to-assay result; the source’s broader biological/mechanistic claims should not be treated as established beyond the reported models.

Technology: terpenoid bioactivity data, computational ranking and melanoma-model assays; candidate activity reported in vitro. AI operation assessed from original Methods.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
terpenoid bioactivity data, computational ranking and melanoma-model assays; candidate activity reported in vitro. AI operation assessed from original Methods.
Limitation / still unknown
Selected candidates were tested in melanoma cell and animal models; these preclinical results do not establish human safety, efficacy, or clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/article/MED/41909737

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AnySleep: a channel-agnostic deep learning system for high-resolution sleep staging in multi-center cohorts.

Source check: Source checked. Not independent scientific confirmation.

Finding: Sub-30-second AnySleep staging revealed information associated with sleep apnea beyond standard 30-second labels in the held-out data. The bibliographic records retrieved do not state an author affiliation.

Technology: A channel-attentive U-Net estimated sleep stages at shorter timescales.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A channel-attentive U-Net estimated sleep stages at shorter timescales.
Test and result
Study analyses on sleep apnea compare high-resolution staging with conventional epochs in the first preprint.
Limitation / still unknown
External replication and real-world benefit were not established by the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2512.14461v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning for survival outcome in head and neck squamous cell carcinoma: a multicenter validation study.

Source check: Source checked. Not independent scientific confirmation.

Finding: A voting model trained on SEER records predicted overall survival for head and neck squamous cell carcinoma and was evaluated in German and Swedish cohorts. Its discrimination varied across cohorts; the study supports prognostic assessment, not evidence of treatment benefit.

Technology: Machine-learning survival models use clinicopathological and treatment-related variables, trained on SEER data and assessed in multicenter registry cohorts.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning survival models use clinicopathological and treatment-related variables, trained on SEER data and assessed in multicenter registry cohorts.
Test and result
The study reports retrospective overall-survival prediction using a large SEER training set and validation cohorts in Germany and Sweden; prognostic performance does not establish treatment benefit.
Limitation / still unknown
Validation cohorts support retrospective survival prediction; the study does not test whether model-guided treatment changes survival or clinical decisions.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1038/s41598-025-29295-6 · https://europepmc.org/article/MED/41318646

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based prediction of short-term outcomes in aneurysmal subarachnoid hemorrhage: a multicenter study integrating clinical and inflammatory indicators.

Source check: Source checked. Not independent scientific confirmation.

Finding: development data from four tertiary hospitals (1,120 patients), separate 326-patient Second Xiangya validation; six algorithms, GBM AUC .895 internal and .864 quasi-external. Does not establish clinical impact. Methods/Results: development data from four tertiary hospitals (1,120 patients), separate 326-patient Second Xiangya validation; six algorithms, GBM AUC .895 internal and .864 quasi-external. Testing: Methods/Results: development data from four tertiary hospitals (1,120 patients), separate 326-patient Second Xiangya validation; six algorithms, GBM AUC .895 internal and .864 quasi-external. Does not establish clinical impact.. Multicenter retrospective aSAH model predicts short-term outcomes with internal and quasi-external validation; the external site is described as quasi-external and prospective validation remains needed. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: development data from four tertiary hospitals (1,120 patients), separate 326-patient Second Xiangya validation; six algorithms, GBM AUC .895 internal and .864 quasi-external. Does not establish clinical impact.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
development data from four tertiary hospitals (1,120 patients), separate 326-patient Second Xiangya validation; six algorithms, GBM AUC .895 internal and .864 quasi-external. Does not establish clinical impact.
Test and result
development data from four tertiary hospitals (1,120 patients), separate 326-patient Second Xiangya validation; six algorithms, GBM AUC .895 internal and .864 quasi-external. Does not establish clinical impact.
Limitation / still unknown
Multicenter retrospective aSAH model predicts short-term outcomes with internal and quasi-external validation; the external site is described as quasi-external and prospective validation remains needed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12916-025-04523-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Interpretable deep learning for depression detection in neurological patients using EEG signals.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary full text reports a defined retrospective clinical classification task: EEG-based depression detection in neurological patients. CNN classification of extracted EEG features is tested on 70 held-out cases with reported 98% accuracy and cross-validation; the claim is limited to this dataset and is not evidence of clinical benefit or general deployment.

Technology: Primary full text, Methods and Results (PMID 41438403; PMCID PMC12720302). It uses a public EEG depression-feature dataset; test sample 70, accuracy 98%, cross-validation 97.6±1.2%. Narrow retrospective diagnostic validation; no clinical-utility claim.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Primary full text, Methods and Results (PMID 41438403; PMCID PMC12720302). It uses a public EEG depression-feature dataset; test sample 70, accuracy 98%, cross-validation 97.6±1.2%. Narrow retrospective diagnostic validation; no clinical-utility claim.
Limitation / still unknown
The primary full text reports a defined retrospective clinical classification task: EEG-based depression detection in neurological patients. CNN classification of extracted EEG features is tested on 70 held-out cases with reported 98% accuracy and cross-validation; the claim is limited to this dataset and is not evidence of clinical benefit or general deployment.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12720302/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Model with Fourier-Transform Infrared Spectroscopy (FTIR) as a Proof-of-Concept Tool for Predicting Group A Streptococcus (GAS) emm-Type in the Pediatric Population.

Source check: Source checked. Not independent scientific confirmation.

Finding: Fourier-transform infrared spectra from 24 pediatric group A Streptococcus isolates were classified by machine learning. In the 21-strain, four-emm-type evaluation, the artificial neural network reached 90.7% accuracy; three isolates from unseen emm types received low-confidence assignments. This is a small proof of concept, not a validated unknown-class rejection system. The bibliographic records retrieved do not state an author affiliation.

Technology: IR Biotyper machine-learning classifiers applied to FTIR spectra; the artificial neural network had the strongest reported accuracy.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning classifiers analyzed FTIR spectra to predict the emm type of pediatric group A Streptococcus isolates.
Test and result
The ANN achieved 90.7% accuracy (89.9% sensitivity, 98.2% specificity, F1 0.92) in leave-one-out cross-validation over 21 strains and four emm types; three out-of-type strains were separately assessed.
Limitation / still unknown
The sample was small; low confidence on three unseen emm types does not establish reliable rejection of unknown classes. Multicenter validation was not reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2075-4418/15/23/3041/pdf?version=1764575278

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Neutrophil percentage to albumin ratio and advanced cardiovascular-kidney-metabolic syndrome in diabetes: a machine learning approach.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: NHANES data from 9,375 adults were analyzed for NPAR/advanced CKM association; Boruta-based feature selection and an ML model predicted advanced CKM (validation AUC 0.872), a retrospective association/prediction result.

Technology: BackgroundDiabetes mellitus (DM) has become an increasingly significant global health challenge, with rising complications and mortality rates. Patients with DM are at a higher risk for advanced Cardiovascular-kidney-metabolic (CKM) syndrome, underscoring the importance of early detection and precise prevention strategies. The neutrophil percentage to albumin ratio (NPAR), a composite biomarker, may be indicative of inflammatory dysregulation and nutritional status in patients with advanced…

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Using six NHANES cycles, the study assessed the association between neutrophil percentage-to-albumin ratio and advanced CKM syndrome, then used Boruta feature selection and machine learning for internal prediction.
Limitation / still unknown
The full text describes the cohort or experimental system, comparison/reference, and internal/external or experimental validation used for the represented claim; no broader clinical benefit is inferred.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12720478/ · https://europepmc.org/article/MED/41310885

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Early prediction of thyroid capsule invasion in papillary microcarcinoma using ultrasound-based deep learning models: a retrospective multicenter study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Multicenter pathology-referenced external testing shows ultrasound deep learning predicts thyroid capsule invasion and improves reader performance; retrospective scope applies.

Technology: Methods identify the learned model and its operation in the reported task.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Methods identify the learned model and its operation in the reported task.
Test and result
Original Methods/Results validation design examined; interpretation is bounded to that design.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13244-025-02132-0

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AI-Augmented Fundus Disease Screening by Non-Ophthalmologist Physicians: A Paired Before-After Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: A paired before-after reader study on 500 fundus photographs shows an AI decision-support tool improving non-ophthalmologists' screening accuracy and referral decisions, with a suitable evaluation design. The bibliographic records retrieved do not state an author affiliation.

Technology: A paired before-after reader study on 500 fundus photographs shows an AI decision-support tool improving non-ophthalmologists' screening accuracy and referral decisions, with a suitable evaluation design.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A paired before-after reader study on 500 fundus photographs shows an AI decision-support tool improving non-ophthalmologists' screening accuracy and referral decisions, with a suitable evaluation design.
Test and result
A paired before-after reader study on 500 fundus photographs shows an AI decision-support tool improving non-ophthalmologists' screening accuracy and referral decisions, with a suitable evaluation design.
Limitation / still unknown
External replication and real-world benefit were not established by the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/bioengineering/bioengineering-12-01304/article_deploy/bioengineering-12-01304.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Eye-XAI: an explainable artificial intelligence approach for eye disease detection using symptom analysis.

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary BMC Med Inform Decis Mak Methods/Results use 563 patients referred to ophthalmologists, with disease labels and symptom annotations, and evaluate ML/XAI in train/test and cross-validation. Reported 99.11% accuracy; single Bangladesh cohort and extensive model tuning make this exploratory. The bibliographic records retrieved do not state an author affiliation.

Technology: Tree/boosting classifiers and SHAP explain symptom-based predictions across five ophthalmic diagnoses.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Tree/boosting classifiers and SHAP explain symptom-based predictions across five ophthalmic diagnoses.
Test and result
563 ophthalmologist-diagnosed patients; highest reported classifier accuracy 99.11%, with a clinician reviewing feature explanations.
Limitation / still unknown
Primary BMC Med Inform Decis Mak Methods/Results use 563 patients referred to ophthalmologists, with disease labels and symptom annotations, and evaluate ML/XAI in train/test and cross-validation. Reported 99.11% accuracy; single Bangladesh cohort and extensive model tuning make this exploratory. Internal splits/cross-validation only and no external validation; modest, imbalanced dataset and extensive tuning mean reported accuracy may be optimistic. Clinician checked SHAP explanations, not independent outcome verification.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12911-025-03253-8

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Clinical validation of an AI-assisted system for real-time kidney stone detection during flexible ureteroscopic surgery.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Learned video detection and tracking exposed residual-stone counts, sizing and composition during ureteroscopy, with independent case-level diagnostic evaluation.

Technology: YOLOv11-N with BoT-SORT detected and tracked stones from surgical video.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
YOLOv11-N with BoT-SORT detected and tracked stones from surgical video.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41746-025-02109-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Explainable machine learning model predicts response to adjuvant therapy after radical cystectomy in bladder cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The source reports postoperative adjuvant-response prediction; the RF test AUC is 0.74 after 0.92 training AUC.

Technology: Models built from 1,764 cystectomy cases and clinical/pathology variables.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Models built from 1,764 cystectomy cases and clinical/pathology variables.
Limitation / still unknown
Held-out patient test performance is reported; source itself says further external validation is needed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2025.1664965

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Multidimensional proteomics and explainable AI feature selection identify cross-platform lung cancer molecular signature in blood plasma

Source check: Source checked. Not independent scientific confirmation.

Finding: Explainable-AI feature selection across mass-spectrometry and proximity-extension proteomics identifies a cross-platform blood-plasma lung-cancer molecular signature.

Technology: Measured plasma from 490 treatment-naive lung-cancer cases and 124 controls is analyzed with independent platforms and cohort comparisons.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Measured plasma from 490 treatment-naive lung-cancer cases and 124 controls is analyzed with independent platforms and cohort comparisons.
Test and result
Signature performance and stage coverage are promising but do not establish prospective screening benefit.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-7660411/v1.pdf?c=1792148747000

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning-driven TCRβ repertoire analysis enhances diagnosis and enables mining of immunological biomarkers in systemic lupus erythematosus.

Source check: Source checked. Not independent scientific confirmation.

Finding: DeepTAPE classifies SLE from TCRβ CDR3 sequences and the abstract reports disease-activity and biomarker analyses on patient data; the claim is diagnostic/biomarker discovery, not clinical benefit. The result is reported in BioData Mining. Affiliation was not available in checked primary metadata; no organization is assigned. The study does not establish generalization beyond its reported setting or prospective real-world benefit. Affiliation was not stated in the reviewed material.

Technology: DeepTAPE applies a deep-learning model to T-cell receptor beta-chain CDR3 sequences to classify systemic lupus erythematosus and analyze disease activity and associated biomarkers.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
DeepTAPE applies a deep-learning model to T-cell receptor beta-chain CDR3 sequences to classify systemic lupus erythematosus and analyze disease activity and associated biomarkers.
Test and result
DeepTAPE classifies SLE from TCRβ CDR3 sequences and the abstract reports disease-activity and biomarker analyses on patient data; the claim is diagnostic/biomarker discovery, not clinical benefit.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13040-025-00490-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Skel-Net: automatic prediction of skeletal pattern on scanned lateral cephalograms using anatomical prior-guided deep learning network.

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary BMC Oral Health Methods/Results use five-year paired cephalograms from 245 children/adolescents; Skel-Net predicts future ANB angle changes with lower error than baseline networks. The dataset is single-center and no external validation is available. The bibliographic records retrieved do not state an author affiliation.

Technology: Two-stage Ceph-Net landmark/heatmap extraction and anatomy-prior-guided DL predict five-year ANB change.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Two-stage Ceph-Net landmark/heatmap extraction and anatomy-prior-guided DL predict five-year ANB change.
Test and result
612 images from 245 untreated patients; held-out test split MAE 1.021 degrees, RMSE 1.338, R2 .517; beats listed DL baselines.
Limitation / still unknown
Primary BMC Oral Health Methods/Results use five-year paired cephalograms from 245 children/adolescents; Skel-Net predicts future ANB angle changes with lower error than baseline networks. The dataset is single-center and no external validation is available. Temporal paired films provide actual growth outcome; internal train/validation/test only and no external center limit generalization.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12903-025-06771-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Clinical feasibility of two cardiac deep learning cine magnetic resonance imaging sequences: Single-breath-hold and free-breathing motion-corrected approaches.

Source check: Source checked. Not independent scientific confirmation.

Finding: The checked primary abstract reports this specific result: Unverified Haiku assessment. Deep learning accelerated cine sequences are compared with conventional segmented cine in 86 patients, with scan time reduced by about 62-73% and comparable function measures, a clinical feasibility result where the DL reconstruction is the enabling method. The abstract identifies the model or learned operation and an observed evaluation; the conclusion is limited to the reported study setting. Original PMC full-text Methods and Results were checked; date 2025-10-31 reconciled to the primary record. Published in Journal of Cardiovascular Magnetic Resonance.

Technology: Publisher/PMC Methods establish the learned model and its operation in the specific study.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher/PMC Methods establish the learned model and its operation in the specific study.
Test and result
Validation design and reported performance/comparison were checked in the full text and are stated narrowly.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12766484/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Portfolio analysis of single-cell RNA-sequencing and transcriptomic data unravels immune cells and telomere-related biomarkers in sepsis.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The article combines 101 ML algorithm combinations, transcriptomic and single-cell data to identify four sepsis biomarkers, then checks expression by RT-qPCR in clinical samples. This supports an exploratory sepsis biomarker result; the candidates and drug prediction do not establish clinical utility or causality.

Technology: Full text specifies the learned operation and how it supports the represented result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full text specifies the learned operation and how it supports the represented result.
Limitation / still unknown
The four candidate markers and drug-response predictions come from transcriptomic and single-cell analyses; RT-qPCR checks expression in clinical samples but does not establish causal biology or clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1638156/full · https://europepmc.org/article/MED/41246299

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Enhanced prediction of breast cancer patient response to chemotherapy by integrating deconvolved expression patterns of immune, stromal and tumor cells.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full Methods/Results describe deconvolution plus supervised models over three breast-cancer cohorts (255 patients) to predict chemotherapy response and identify cell-type/interactions associated with response; performance assessed across cohorts against bulk-expression baseline. Bounded retrospective predictive finding; no causal treatment benefit claimed.

Technology: Full Methods/Results describe deconvolution plus supervised models over three breast-cancer cohorts (255 patients) to predict chemotherapy response and identify cell-type/interactions associated with response; performance assessed across cohorts against bulk-expression baseline. Bounded retrospective predictive finding; no causal treatment benefit claimed.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full Methods/Results describe deconvolution plus supervised models over three breast-cancer cohorts (255 patients) to predict chemotherapy response and identify cell-type/interactions associated with response; performance assessed across cohorts against bulk-expression baseline. Bounded retrospective predictive finding; no causal treatment benefit claimed.
Limitation / still unknown
The three cohorts comprise 255 retrospective patients; prediction and cell-interaction associations do not establish causal treatment effects or prospective clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12807300/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning-based PSMA PET segmentation repeatability: A post-hoc analysis of a single-center, prospective, test-retest trial.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary source Methods/Results confirm the represented result: Post-hoc analysis of a prospective test-retest PET trial quantifying repeatability of an AI PSMA segmentation model; clinically specific evaluation.

Technology: Post-hoc analysis of a prospective test-retest PET trial quantifying repeatability of an AI PSMA segmentation model; clinically specific evaluation.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Post-hoc analysis of a prospective test-retest PET trial quantifying repeatability of an AI PSMA segmentation model; clinically specific evaluation.
Test and result
The source reports the defined cohort/experiment and its reference or test comparison; interpretation is restricted to that study design.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11547-025-02137-1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Imidacloprid contributes to bladder cancer progression: preliminary evidence based on network toxicology, machine learning and molecular docking.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning analyses of imidacloprid-related and bladder-cancer expression data identified a five-gene prognostic signature (SREBF1, PRELP, TGFBI, TNFAIP2 and TACR3), evaluated in TCGA and two GEO cohorts. In bladder-cancer cell assays, imidacloprid exposure increased measured proliferation, migration and invasion; expression assays showed changes in four of the five genes, while TACR3 did not significantly change. The cell findings do not establish that imidacloprid causes bladder cancer in people or that the risk model improves patient outcomes.

Technology: Ten machine-learning algorithms and combinations analyzed CTD exposure-related genes and TCGA expression data; multivariable Cox analysis formed a five-gene risk score.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Machine-learning models selected exposure-related expression features for prognosis; the separate cell experiments directly tested imidacloprid-associated cellular behavior.
Test and result
The prognostic signature was assessed in TCGA and external GSE13507 and GSE31684 cohorts. Cell assays measured proliferation, wound closure and invasion after exposure; RT-qPCR and western blot checked gene expression. Four genes changed in the reported assays; TACR3 did not significantly change.
Limitation / still unknown
The cohort analyses are retrospective and the cell assays are in vitro. Neither shows that the exposure causes bladder cancer in people, nor establishes prospective clinical utility of the prognostic model.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12577002/ · https://europepmc.org/article/MED/41168844

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based tool to assess risk of hemodynamically significant PDA in extremely premature infants.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A clinical model estimated hemodynamically significant PDA risk in extremely premature infants.

Technology: Methods identify the learned operation enabling the represented result: A clinical model estimated hemodynamically significant PDA risk in extremely premature infants.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods identify the learned operation enabling the represented result: A clinical model estimated hemodynamically significant PDA risk in extremely premature infants.
Limitation / still unknown
This is a single-center retrospective cohort of 677 extremely preterm infants; the model’s independent test-set AUC was 0.685, with sensitivity 74.8% and specificity 53.4%.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1038/s41390-025-04489-w · https://europepmc.org/article/MED/41162564

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and validation of machine learning models for predicting STAS in stage I lung adenocarcinoma with part-solid and solid nodules: a two-center study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Original Frontiers Methods/Results: ML models predict STAS in stage-I lung adenocarcinoma using 353-patient training data and independent 120-patient second-center validation; tuning used stratified 10-fold CV in training only. Retrospective multicenter validation; narrow diagnostic prediction. The bibliographic records retrieved do not state an author affiliation.

Technology: Original Frontiers Methods/Results: ML models predict STAS in stage-I lung adenocarcinoma using 353-patient training data and independent 120-patient second-center validation; tuning used stratified 10-fold CV in training only. Retrospective multicenter validation; narrow diagnostic prediction.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Original Frontiers Methods/Results: ML models predict STAS in stage-I lung adenocarcinoma using 353-patient training data and independent 120-patient second-center validation; tuning used stratified 10-fold CV in training only. Retrospective multicenter validation; narrow diagnostic prediction.
Test and result
Original Frontiers Methods/Results: ML models predict STAS in stage-I lung adenocarcinoma using 353-patient training data and independent 120-patient second-center validation; tuning used stratified 10-fold CV in training only. Retrospective multicenter validation; narrow diagnostic prediction.
Limitation / still unknown
Original Frontiers Methods/Results: ML models predict STAS in stage-I lung adenocarcinoma using 353-patient training data and independent 120-patient second-center validation; tuning used stratified 10-fold CV in training only. Retrospective multicenter validation; narrow diagnostic prediction. Methods/Results specify cohort, reference/outcome, split or experimental comparator and limitations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2025.1682633/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Integrating Machine Learning and Molecular Methods for <i>Trichophyton indotineae</i> Identification and Resistance Profiling Using MALDI-TOF Spectra.

Source check: Source checked. Not independent scientific confirmation.

Finding: Partial least-squares discriminant analysis and support-vector machines classified 56 clinical isolates from MALDI-TOF spectra and examined ERG1 resistance correlation.

Technology: Partial least-squares discriminant analysis and support-vector machines classified 56 clinical isolates from MALDI-TOF spectra and examined ERG1 resistance correlation.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Partial least-squares discriminant analysis and support-vector machines classified 56 clinical isolates from MALDI-TOF spectra and examined ERG1 resistance correlation.
Test and result
Cross-validated PLS-DA and SVM reached 100% balanced accuracy in the isolate dataset; this small laboratory cohort does not establish prospective clinical diagnostic performance.
Limitation / still unknown
The reviewed source does not establish prospective benefit, broad external validity, or replication beyond the reported study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/pathogens/pathogens-14-00986/article_deploy/pathogens-14-00986.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and validation of a machine learning-based prediction model for intraoperative hypothermia in Chinese patients undergoing gastrointestinal surgery.

Source check: Source checked. Not independent scientific confirmation.

Finding: Five ML methods were compared in 418 gastrointestinal-surgery patients for intraoperative hypothermia; logistic regression was best (test AUC 0.872), and a six-feature nomogram scored 0.877. Retrospective single-center validation only. The bibliographic records retrieved do not state an author affiliation.

Technology: Five ML methods were compared in 418 gastrointestinal-surgery patients for intraoperative hypothermia; logistic regression was best (test AUC 0.872), and a six-feature nomogram scored 0.877. Retrospective single-center validation only.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Five ML methods were compared in 418 gastrointestinal-surgery patients for intraoperative hypothermia; logistic regression was best (test AUC 0.872), and a six-feature nomogram scored 0.877. Retrospective single-center validation only.
Test and result
Five ML methods were compared in 418 gastrointestinal-surgery patients for intraoperative hypothermia; logistic regression was best (test AUC 0.872), and a six-feature nomogram scored 0.877. Retrospective single-center validation only.
Limitation / still unknown
Five ML methods were compared in 418 gastrointestinal-surgery patients for intraoperative hypothermia; logistic regression was best (test AUC 0.872), and a six-feature nomogram scored 0.877. Retrospective single-center validation only.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12487104/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based identification of leptin-associated biomarkers and prognostic prediction models in sepsis.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Machine learning identifies sepsis biomarker subtypes and is followed by molecular validation.

Technology: Machine learning identifies sepsis biomarker subtypes and is followed by molecular validation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine learning identifies sepsis biomarker subtypes and is followed by molecular validation.
Limitation / still unknown
The prognostic models and subtypes are derived from existing datasets; qPCR/Western blot validation is molecular confirmation, not evidence of patient benefit or causal treatment response.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12515905/ · https://europepmc.org/article/MED/41089329

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Label-free, machine-learning assisted impedance assessment of basophil activation.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary Methods and Results show label-free IFC signals from 15 donors were mapped to FCA reference activation using leave-one-donor-out cross-validation; this is a specific ex-vivo basophil activation assay result, not clinical deployment.

Technology: Methods and Results show label-free IFC signals from 15 donors were mapped to FCA reference activation using leave-one-donor-out cross-validation; this is a specific ex-vivo basophil activation assay result, not clinical deployment.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Methods and Results show label-free IFC signals from 15 donors were mapped to FCA reference activation using leave-one-donor-out cross-validation; this is a specific ex-vivo basophil activation assay result, not clinical deployment.
Limitation / still unknown
A donor-held-out test and flow-cytometry reference qualify for the assay claim; study size (15 donors) limits generalization.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12646617/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based radiomics model: prognostic prediction and mechanism exploration in patients with endometrial cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding: A radiomics model predicted postoperative overall survival in endometrial cancer and linked its score to angiogenesis-related imaging and molecular findings, with prospective imaging and laboratory follow-up. The result is reported in Biomarker Research. Affiliation was not available in checked primary metadata; no organization is assigned. The study does not establish generalization beyond its reported setting or prospective real-world benefit. Affiliation was not stated in the reviewed material.

Technology: A radiomics-based machine-learning model predicted postoperative overall survival in endometrial cancer; prospective imaging and laboratory follow-up examined associated angiogenesis-related findings.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A radiomics-based machine-learning model predicted postoperative overall survival in endometrial cancer; prospective imaging and laboratory follow-up examined associated angiogenesis-related findings.
Test and result
A radiomics model predicted postoperative overall survival in endometrial cancer and linked its score to angiogenesis-related imaging and molecular findings, with prospective imaging and laboratory follow-up.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12481825/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Novel multi-task learning for Alzheimer's stage classification using hippocampal MRI segmentation, feature fusion, and nomogram modeling.

Source check: Source checked. Not independent scientific confirmation.

Finding: A retrospective multi-center study of 2,956 people classified four Alzheimer disease stages from hippocampal MRI and fused radiomic, deep and clinical features. MedT produced the best reported external hippocampal segmentation Dice score (92.03%); a fused-feature XGBoost classifier reached external accuracy 92.8% and AUC 0.942. An end-to-end 3D CNN-LSTM had AUC 0.840. These dataset results do not establish prospective clinical benefit.

Technology: Hippocampal segmentation compared U-Net, nnU-Net, Swin-UNet and MedT; LSTM attention learned image features, and five classifiers including XGBoost classified stage. SHAP, a nomogram and decision-curve analysis were also reported.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep learning segmented the hippocampus and produced image features; fused radiomic, learned and clinical features fed stage classifiers, with XGBoost performing best in the reported external evaluation.
Test and result
Retrospective multi-center MRI data from 2,956 patients across four stages; external segmentation Dice was 92.03% for MedT, and external classification accuracy/AUC were 92.8%/0.942 for the fused-feature XGBoost pipeline. The end-to-end 3D CNN-LSTM AUC was 0.840.
Limitation / still unknown
The study is retrospective and evaluates classification against the included stage labels; external cohort performance does not establish prospective diagnostic utility or improved care.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12482563/ · https://europepmc.org/article/MED/41024255

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Model for an App-Based Tool to Assist With the Diagnosis of Canine Atopic Dermatitis.

Source check: Source checked. Not independent scientific confirmation.

Finding: Methods describe cases from board-certified clinicians and train/test procedure; Results report diagnostic sensitivity and specificity on the held-out cases. Testing: Methods describe cases from board-certified clinicians and train/test procedure; Results report diagnostic sensitivity and specificity on the held-out cases.. A Random Forest app model distinguished clinically diagnosed canine atopic dermatitis from other dermatoses using referral cases and a stratified held-out evaluation. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Methods describe cases from board-certified clinicians and train/test procedure; Results report diagnostic sensitivity and specificity on the held-out cases.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe cases from board-certified clinicians and train/test procedure; Results report diagnostic sensitivity and specificity on the held-out cases.
Test and result
Methods describe cases from board-certified clinicians and train/test procedure; Results report diagnostic sensitivity and specificity on the held-out cases.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pmc&id=PMC12967876

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Optimization of Welding Parameters Using an Improved Hill-Climbing Algorithm Based on BP Neural Network for Multi-Bead Weld Smoothness Control.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Two BP neural networks predict bead width/height and flatness from welding inputs; improved hill-climbing optimizes process settings. Experimental evaluation used 17 Box–Behnken designs with fit/residual checks; conclusion bounded to parameter range.

Technology: §5.3: two learned BP networks map process inputs to bead geometry and surface flatness; hill-climbing searches predicted responses.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
§5.3: two learned BP networks map process inputs to bead geometry and surface flatness; hill-climbing searches predicted responses.
Limitation / still unknown
§6: validation in experimental design range; practical production constraints and external manufacturing applicability remain untested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/1996-1944/18/17/4084 · https://europepmc.org/article/MED/40942510

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Molecular mechanisms of ferroptosis in ulcerative colitis: insights from machine learning, WGCNA, and immune cell infiltration analysis.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full text identifies MFN2 and CBS via differential-expression screening, LASSO/SVM-RFE and WGCNA on UC cohorts; it tests expression against external GSE92415 and in Caco-2 inflammatory/ferroptosis experiments. The results support candidate ferroptosis-linked UC markers, with cell-model and retrospective-cohort limits.

Technology: Full text specifies the learned operation and how it supports the represented result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full text specifies the learned operation and how it supports the represented result.
Limitation / still unknown
External dataset and Caco-2 inflammatory-model checks support candidate-marker associations; they do not establish causality or therapeutic effects in ulcerative colitis patients.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1615186/full · https://europepmc.org/article/MED/40948782

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Technical and clinical validation of a novel deep learning-based white matter hyperintensity segmentation tool.

Source check: Source checked. Not independent scientific confirmation.

Finding: The study trained MD-GRU and nnU-Net deep-learning models for white-matter-hyperintensity segmentation and tested them on datasets excluded from training. Against reference masks and comparator algorithms, the models showed improved agreement, reduced bias and improved repeatability and inter-scanner reproducibility. In the two-year DiViNAS data, nnU-Net yielded a calculated sample size 41% smaller than the best comparator for detecting a treatment effect; this is a statistical-power calculation, not evidence of an observed treatment effect or clinical benefit.

Technology: MD-GRU and nnU-Net deep-learning segmentation models for MRI white matter hyperintensities.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep-learning segmentation generates WMH volumes; the resulting measurements were assessed against masks, repeat scans, scanners and longitudinal biomarker data.
Test and result
Technical tests used MarkVCID reference masks and scan-rescan/inter-scanner data plus SWISS-AF images; clinical-biomarker analysis used two-year DiViNAS follow-up. Models improved agreement, bias, precision and reproducibility versus benchmark algorithms; nnU-Net produced a 41% lower calculated sample size than the best comparator for treatment-effect detection.
Limitation / still unknown
The treatment-effect result is a power calculation from existing follow-up data, not a measured treatment response. The study does not demonstrate improved patient outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12448020/ · https://europepmc.org/article/MED/40979673

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and Validation of a Machine Learning Model to Predict Postoperative Morbidity and Mortality in Patients With COVID-19 Who Underwent Major Surgery.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning models including LightGBM predicted 30-day postoperative morbidity and mortality among more than 15,000 COVID-positive surgical procedures. Models trained on 2022 and tested on 2021 data; missing external datasets and selection bias limit generalization.

Technology: LightGBM and other supervised models predict a composite postoperative outcome from preoperative features.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LightGBM and other supervised models predict a composite postoperative outcome from preoperative features.
Test and result
More than 15,000 COVID-positive procedures; 2022 training/2021 testing plus internal validation. Authors note absent external datasets and possible selection bias.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.cureus.com/articles/401139-development-and-validation-of-a-machine-learning-model-to-predict-postoperative-morbidity-and-mortality-in-patients-with-covid-19-who-underwent-major-surgery.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Multiparametric MRI-Based Machine Learning Radiomics Prognostic Models for Multifocal Hepatocellular Carcinoma Beyond Milan Criteria: A Retrospective Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: MRI radiomics predictors improved overall- and recurrence-free-survival prediction for multifocal hepatocellular carcinoma beyond Milan criteria in training and validation cohorts.

Technology: Methods describe retrospective MRI cohort and radiomics models; Results compare radiomics-augmented prognostic models in validation data.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe retrospective MRI cohort and radiomics models; Results compare radiomics-augmented prognostic models in validation data.
Test and result
Methods describe retrospective MRI cohort and radiomics models; Results compare radiomics-augmented prognostic models in validation data.
Limitation / still unknown
Methods describe retrospective MRI cohort and radiomics models; Results compare radiomics-augmented prognostic models in validation data.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12401053/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial intelligence-assisted surgical simulation system based on non-enhanced computed tomography images in thoracoscopic pulmonary segmentectomies.

Source check: Source checked. Not independent scientific confirmation.

Finding: AI-assisted surgical simulation evaluated pulmonary-nodule navigation in a clinical context.

Technology: LungDimensionGo uses EfficientDet for nodule detection and Mamba-UNet with SegRefiner for CT segmentation and 3D lung reconstruction to support preoperative segmentectomy planning.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LungDimensionGo uses EfficientDet for nodule detection and Mamba-UNet with SegRefiner for CT segmentation and 3D lung reconstruction to support preoperative segmentectomy planning.
Test and result
The source abstract describes retrospective (125 patients) and prospective (38 patients) cohorts and reports shorter reconstruction and operative times and fewer complications in the AI-assisted group; this single-center comparison does not establish general clinical benefit.
Limitation / still unknown
This single-institution comparison included retrospective (125) and prospective (38) patients; reported perioperative differences do not establish generalizability beyond the site or prove improved long-term outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12433101/ · https://europepmc.org/article/MED/40950911

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Multiomics Integration and Machine Learning Reveal Colony Stimulating Factor 3 Receptor as a Key Gene and Therapeutic Target in Crohn's Disease.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning feature selection across Crohn’s disease datasets converged on CSF3R, and Mendelian-randomization analysis supported a disease association. The result is reported in Mediators of Inflammation. Affiliation was not available in checked primary metadata; no organization is assigned. The study does not establish generalization beyond its reported setting or prospective real-world benefit. Affiliation was not stated in the reviewed material.

Technology: Random forest, SVM-RFE, and LASSO ranked candidate genes

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random forest, SVM-RFE, and LASSO ranked candidate genes
Test and result
Machine-learning feature selection across Crohn’s disease datasets converged on CSF3R, and Mendelian-randomization analysis supported a disease association.
Limitation / still unknown
Independent replication and prospective real-world benefit are not established in the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12411032/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC12411032/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC12411032/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AI-Driven Tai Chi mastery using deep learning framework for movement assessment and personalized training.

Source check: Source checked. Not independent scientific confirmation.

Finding: movement collection and ST-GCN/attention/LSTM model; Results validation and participant-training comparison. Reported expert-score agreement and skill-score changes are limited to the study’s constructed movement task. The bibliographic records retrieved do not state an author affiliation.

Technology: movement collection and ST-GCN/attention/LSTM model; Results validation and participant-training comparison. Reported expert-score agreement and skill-score changes are limited to the study’s constructed movement task.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
movement collection and ST-GCN/attention/LSTM model; Results validation and participant-training comparison. Reported expert-score agreement and skill-score changes are limited to the study’s constructed movement task.
Test and result
movement collection and ST-GCN/attention/LSTM model; Results validation and participant-training comparison. Reported expert-score agreement and skill-score changes are limited to the study’s constructed movement task.
Limitation / still unknown
Methods: movement collection and ST-GCN/attention/LSTM model; Results validation and participant-training comparison. Reported expert-score agreement and skill-score changes are limited to the study’s constructed movement task.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12394405/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

CEACAM6 as a machine learning derived immune biomarker for predicting neoadjuvant chemotherapy response in HR+/HER2- breast cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Methods 2.1: 956 HR+/HER2− GEO cases, five train/four test datasets; Results: 10-algorithm analysis, NeuralNet AUC .816; retrospective 106-patient cohort and paired IHC n=9. Biomarker association, no proven treatment benefit.

Technology: Machine-learning classifiers rank CEACAM6-related features and evaluate association with neoadjuvant chemotherapy response.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning classifiers rank CEACAM6-related features and evaluate association with neoadjuvant chemotherapy response.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1662004/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Multi-omics derivation of a core gene signature for predicting therapeutic response and characterizing immune dysregulation in inflammatory bowel disease.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: LASSO/random-forest selection yielded an IBD gene signature associated with subtype biology and biologic-treatment response across four datasets.

Technology: Ten-algorithm model evaluation, four treatment-response datasets, single-cell analysis and DSS-colitis experiments test different aspects of the signature.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Ten-algorithm model evaluation, four treatment-response datasets, single-cell analysis and DSS-colitis experiments test different aspects of the signature.
Limitation / still unknown
Primary Results describe the specific comparator, cohort or experimental check; limitations are stated narrowly.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1611598/full · https://europepmc.org/article/MED/40821793

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Geometry-encoded deep learning (GeoDL) framework for real-time 3D dose verification for online adaptive radiotherapy.

Source check: Source checked. Not independent scientific confirmation.

Finding: A geometry-encoded 3D U-Net estimates dose from integrated fluence maps and CT volumes; 281 prostate cases were used for training/validation and 24 separate cases for testing against treatment-planning dose. Reported agreement and speed support a retrospective dosimetric feasibility result; use remains limited to prostate VMAT/single LINAC configuration, with clinical implementation still future work.

Technology: A geometry-encoded CT/fluence-map input is processed by 3D U-Net to estimate full-volume dose.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A geometry-encoded CT/fluence-map input is processed by 3D U-Net to estimate full-volume dose.
Test and result
281 prostate cases used for training/validation and 24 separate test cases compared against TPS calculations; single site/VMAT/LINAC limits remain.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12753030/ · https://doi.org/10.1088/3049-477x/adcb4b

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The Persistent Threat of Chronic Inflammation on the Mortality Among Cervical Cancer Survivors: A Mendelian Randomization and Machine Learning Analysis Using UK Biobank and Chinese Cohort Data.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary full text supports the case-specific result: Cohort analyses linked chronic inflammation to mortality among cervical-cancer survivors using machine learning and genetic inference. The bibliographic records retrieved do not state an author affiliation.

Technology: The primary full text supports the case-specific result: Cohort analyses linked chronic inflammation to mortality among cervical-cancer survivors using machine learning and genetic inference.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary full text supports the case-specific result: Cohort analyses linked chronic inflammation to mortality among cervical-cancer survivors using machine learning and genetic inference.
Test and result
The primary full text supports the case-specific result: Cohort analyses linked chronic inflammation to mortality among cervical-cancer survivors using machine learning and genetic inference.
Limitation / still unknown
External replication and real-world benefit were not established by the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12318525/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of DNA damage response and crotonylation-related biomarkers for lung adenocarcinoma via machine learning and WGCNA.

Source check: Source checked. Not independent scientific confirmation.

Finding: WGCNA and three machine-learning feature-selection methods applied to LUAD transcriptomes from GSE27262 and GSE140797 yielded candidate DNA-damage-response/crotonylation hub genes; cell cytotoxicity assays provided follow-up checks. These are hypothesis-generating candidates, not demonstrated clinical biomarkers.

Technology: WGCNA and three machine-learning feature-selection methods identify candidate DNA-damage-response/crotonylation genes in lung-adenocarcinoma transcriptomes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
WGCNA and three machine-learning feature-selection methods identify candidate DNA-damage-response/crotonylation genes in lung-adenocarcinoma transcriptomes.
Test and result
Two transcriptomic datasets, database analyses and cell cytotoxicity checks; the candidates are not validated clinical biomarkers.
Limitation / still unknown
External replication and real-world or clinical benefit were not established by the reviewed study; author affiliation is not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s10238-025-01704-0 · https://link.springer.com/content/pdf/10.1007/s10238-025-01704-0.pdf · https://api.crossref.org/works/10.1007%2Fs10238-025-01704-0

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Generating New Coordination Compounds via Multireference Simulations, Genetic Algorithms, and Machine Learning: The Case of Co(II) and Dy(III) Molecular Magnets.

Source check: Source checked. Not independent scientific confirmation.

Finding: The original ACS Methods/Results combine multireference simulations, genetic search and ML prescreening to identify specific high-anisotropy Co(II)/Dy(III) coordination candidates. Admit narrowly as computationally supported molecular-magnet candidates; they were not experimentally synthesized/tested in this study.

Technology: Original Methods identify the model operation and how it contributes to the reported case-specific result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Original Methods identify the model operation and how it contributes to the reported case-specific result.
Test and result
Original Results establish the stated finding, cohort/materials, comparison and scope.
Limitation / still unknown
The source reports the described validation and its limitations; see the decision rationale.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pubs.acs.org/doi/10.1021/jacsau.5c00502

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting subclinical leaflet thrombosis in self-expandable prosthesis: A multimodal machine learning analysis.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: 118 TAVI patients with 6-month CT, 22 subclinical leaflet-thrombosis events; LASSO/RF/XGBoost internal stratified 5-fold validation identifies predictors including bicuspid valve morphology. Bounded internal predictive/association result only.

Technology: LASSO feature selection and random-forest/XGBoost models estimated subclinical leaflet-thrombosis risk from clinical and imaging features.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LASSO feature selection and random-forest/XGBoost models estimated subclinical leaflet-thrombosis risk from clinical and imaging features.
Limitation / still unknown
Only 22 thrombosis events were observed among 118 patients and validation was internal five-fold cross-validation; prospective clinical utility and external validation were not assessed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13244739/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prognosticating salvage stereotactic radiosurgery outcomes in relapsed primary central nervous system lymphoma: A machine learning-driven decision tree analysis.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary Methods and Results show a C5.0 decision tree trained on 209 retrospective SRS-salvage cases with a stratified 144/65 training/validation split and reported response-class sensitivities/specificities. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Methods and Results show a C5.0 decision tree trained on 209 retrospective SRS-salvage cases with a stratified 144/65 training/validation split and reported response-class sensitivities/specificities.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods and Results show a C5.0 decision tree trained on 209 retrospective SRS-salvage cases with a stratified 144/65 training/validation split and reported response-class sensitivities/specificities.
Test and result
C5.0 learned a response tree using KPS, deep-structure involvement and IELSG score; a held-out validation subset was evaluated.
Limitation / still unknown
This is single-center retrospective internal validation; it supports prediction in the studied cohort, not prospective decision impact.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12329516/ · https://europepmc.org/article/MED/40738015 · https://pmc.ncbi.nlm.nih.gov/articles/PMC12329516/#Patients

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Rapid assessment of clinical severity for salmonellosis cases via protein family domain analysis and machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: A defined pathogen-genomic result (severity-associated protein-family domains) is produced by ML on clinical Salmonella isolate data, beyond algorithm-only benchmark; candidate status and lack of biological/clinical external validation must be explicit.

Technology: Original article Methods/Results analyze clinical Salmonella genome and severity data using Pfam-domain representation and ML; reports 70 candidate domains associated with severe cases. Specific genomic candidate biomarkers are computationally prioritized; no prospective severity prediction or functional validation of domains, so remain candidate a

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Original article Methods/Results analyze clinical Salmonella genome and severity data using Pfam-domain representation and ML; reports 70 candidate domains associated with severe cases. Specific genomic candidate biomarkers are computationally prioritized; no prospective severity prediction or functional validation of domains, so remain candidate associations.
Test and result
Original article Methods/Results analyze clinical Salmonella genome and severity data using Pfam-domain representation and ML; reports 70 candidate domains associated with severe cases. Specific genomic candidate biomarkers are computationally prioritized; no prospective severity prediction or functional validation of domains, so remain candidate associations.
Limitation / still unknown
Original article Methods/Results analyze clinical Salmonella genome and severity data using Pfam-domain representation and ML; reports 70 candidate domains associated with severe cases. Specific genomic candidate biomarkers are computationally prioritized; no prospective severity prediction or functional validation of domains, so remain candidate associations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12440251/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Establishment and validation of an artificial intelligence-based system for identifying the culprit vessel in patients with ST-segment elevated myocardial infarction: the ALERT study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: An ECG AI system identified culprit vessels in STEMI cases against angiographic reference labels, with internal and external validation and cardiologist comparison.

Technology: Methods describe 698 STEMI ECGs, Res-LSTM and angiographic labels; Results report internal/external evaluation and cardiologist comparison.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe 698 STEMI ECGs, Res-LSTM and angiographic labels; Results report internal/external evaluation and cardiologist comparison.
Limitation / still unknown
Internal/external validation and cardiologist comparison establish diagnostic performance only; prospective impact on STEMI treatment or patient outcomes was not evaluated.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pmc&id=PMC13528328

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-guided single-cell multiomics uncovers GDF15-driven immunosuppressive niches in NSCLC: A translational framework for overcoming anti-PD-1 resistance

Source check: Source checked. Not independent scientific confirmation.

Finding: The full text describes harmonized survival modeling across four NSCLC immunotherapy cohorts (156 patients), single-cell analysis and external melanoma-cohort evaluation, followed by functional experiments; it distinguishes predictive association from preclinical mechanism. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: AORSF survival model

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The full text describes harmonized survival modeling across four NSCLC immunotherapy cohorts (156 patients), single-cell analysis and external melanoma-cohort evaluation, followed by functional experiments; it distinguishes predictive association from preclinical mechanism.
Test and result
The full text describes harmonized survival modeling across four NSCLC immunotherapy cohorts (156 patients), single-cell analysis and external melanoma-cohort evaluation, followed by functional experiments; it distinguishes predictive association from preclinical mechanism.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12266572/ · https://europepmc.org/article/MED/40582068 · https://pmc.ncbi.nlm.nih.gov/articles/PMC12266572/#Methods:

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning predicts onset acceleration of 38 age-associated diseases from blood and body composition biomarkers in the UK Biobank.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary source Methods/Results confirm the represented result: Neural-network Cox model predicts onset acceleration for 38 age-associated diseases in 60,396 UK Biobank participants with C-index evaluation, a distinct epidemiological result.

Technology: Neural-network Cox model predicts onset acceleration for 38 age-associated diseases in 60,396 UK Biobank participants with C-index evaluation, a distinct epidemiological result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Neural-network Cox model predicts onset acceleration for 38 age-associated diseases in 60,396 UK Biobank participants with C-index evaluation, a distinct epidemiological result.
Test and result
Neural-network Cox model predicts onset acceleration for 38 age-associated diseases in 60,396 UK Biobank participants with C-index evaluation, a distinct epidemiological result.
Limitation / still unknown
The source reports the defined cohort/experiment and its reference or test comparison; interpretation is restricted to that study design.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11357-025-01702-w

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Broadscale reconnaissance of coral reefs from citizen science and deep learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full Springer PDF: ~29,967 Great Barrier Reef images from citizen scientists; AI-alone and AI+citizen labels compared with experts. For 8,086 images, best method was category-specific; estimated coral-cover errors were small for several categories, but "other coral" accuracy was weaker. This is an empirical field-monitoring result. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit.

Technology: Deep-learning image classifiers labeled citizen-science reef images, alone and with citizen labels, for comparison with expert annotations.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep-learning image classifiers labeled citizen-science reef images, alone and with citizen labels, for comparison with expert annotations.
Limitation / still unknown
The result is bounded to the surveyed Great Barrier Reef images; category performance varied, with weaker accuracy for “other coral,” and transfer to other reef systems was not tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s10661-025-14261-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Exploring the association between circadian rhythms and osteoporosis: new diagnostic and therapeutic targets identified via machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: Four ML algorithms compared to select circadian-related osteoporosis genes; the SVM-derived five-gene diagnostic model was evaluated on a separate GEO dataset (AUC 0.904 training; 0.887 validation). Retrospective diagnostic performance only; no clinical-benefit claim. The bibliographic records retrieved do not state an author affiliation.

Technology: Methods 2.1 and 2.4-2.5: GSE56815 training (40 OP/40 controls), GSE7158 validation (12/14); four ML algorithms select diagnostic genes. Results: SVM-selected five-gene model AUC 0.904 training, 0.887 validation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods 2.1 and 2.4-2.5: GSE56815 training (40 OP/40 controls), GSE7158 validation (12/14); four ML algorithms select diagnostic genes. Results: SVM-selected five-gene model AUC 0.904 training, 0.887 validation.
Test and result
Methods 2.1 and 2.4-2.5: GSE56815 training (40 OP/40 controls), GSE7158 validation (12/14); four ML algorithms select diagnostic genes. Results: SVM-selected five-gene model AUC 0.904 training, 0.887 validation.
Limitation / still unknown
Four ML algorithms compared to select circadian-related osteoporosis genes; the SVM-derived five-gene diagnostic model was evaluated on a separate GEO dataset (AUC 0.904 training; 0.887 validation). Retrospective diagnostic performance only; no clinical-benefit claim. Methods 2.1 and 2.4-2.5: GSE56815 training (40 OP/40 controls), GSE7158 validation (12/14); four ML algorithms select diagnostic genes. Results: SVM-selected five-gene model AUC 0.904 training, 0.887 validation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/molecular-biosciences/articles/10.3389/fmolb.2025.1614221/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

CELM: An Ensemble Deep Learning Model for Early Cardiomegaly Diagnosis in Chest Radiography.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The CELM stacked CNN ensemble classified cardiomegaly in a curated, augmented four-dataset chest-X-ray benchmark, reporting 0.92 accuracy and 0.90 AUC; this is a dataset-level model result, not evidence of clinical deployment or prospective benefit.

Technology: Three CNN image classifiers feed learned probability scores to a neural stacking meta-classifier for cardiomegaly.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Three CNN image classifiers feed learned probability scores to a neural stacking meta-classifier for cardiomegaly.
Limitation / still unknown
Evaluation used curated, augmented image datasets; prospective testing in a clinical radiology workflow and patient-outcome benefit were not shown.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12249172/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Variable Selection for Multivariate Failure Time Data via Regularized Sparse-Input Neural Network.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The regularized sparse-input neural network is evaluated through simulations and applied to advanced prostate-cancer survival data, where the authors identify established clinical factors and candidate prognostic SNPs; those SNP associations are exploratory and not independently validated.

Technology: A regularized sparse-input neural network selects variables for multivariate failure-time analysis and is applied to prostate-cancer survival data.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A regularized sparse-input neural network selects variables for multivariate failure-time analysis and is applied to prostate-cancer survival data.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/bioengineering/bioengineering-12-00596/article_deploy/bioengineering-12-00596.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Diagnosis of thyroid nodules using ultrasound images based on deep learning features: online dynamic nomogram and gradient-weighted class activation mapping.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Fine-tuned ultrasound model diagnoses thyroid nodules in a retrospective patient cohort.

Technology: Fine-tuned ultrasound model diagnoses thyroid nodules in a retrospective patient cohort.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Fine-tuned ultrasound model diagnoses thyroid nodules in a retrospective patient cohort.
Limitation / still unknown
The model was assessed retrospectively on 1,501 ultrasound images; prospective performance and improved patient outcomes were not evaluated.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12209616/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development of CSOARG: a single-cell and multi-omics-based machine learning model for ovarian cancer prognosis and drug response prediction.

Source check: Source checked. Not independent scientific confirmation.

Finding: Eight-gene ovarian-cancer risk model and treatment-response associations reported; AUC for 5-year survival 0.68. The bibliographic records retrieved do not state an author affiliation.

Technology: LASSO/Cox model built from TCGA/GEO gene expression and clinical data.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LASSO/Cox model built from TCGA/GEO gene expression and clinical data.
Test and result
Eight-gene ovarian-cancer risk model and treatment-response associations reported; AUC for 5-year survival 0.68.
Limitation / still unknown
External replication and real-world benefit were not established by the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fonc.2025.1592426

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based prediction of postoperative pancreatic fistula after laparoscopic pancreaticoduodenectomy.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary full text supports the specific AI-enabled finding: A machine-learning model predicted postoperative pancreatic fistula after laparoscopic pancreaticoduodenectomy. The bibliographic records retrieved do not state an author affiliation.

Technology: The primary full text supports the specific AI-enabled finding: A machine-learning model predicted postoperative pancreatic fistula after laparoscopic pancreaticoduodenectomy.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary full text supports the specific AI-enabled finding: A machine-learning model predicted postoperative pancreatic fistula after laparoscopic pancreaticoduodenectomy.
Test and result
The primary full text supports the specific AI-enabled finding: A machine-learning model predicted postoperative pancreatic fistula after laparoscopic pancreaticoduodenectomy.
Limitation / still unknown
The primary full text supports the specific AI-enabled finding: A machine-learning model predicted postoperative pancreatic fistula after laparoscopic pancreaticoduodenectomy. The paper reports the relevant cohort/image/field dataset, comparator or reference, and validation/test evaluation; the judgment does not imply prospective clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12045007/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Innovations in Proteomic Technologies and Artificial Neural Networks: Unlocking Milk Origin Identification.

Source check: Source checked. Not independent scientific confirmation.

Finding: FNN trained on MALDI-TOF milk spectra; Methods describes train/validation/test split. Results reports species classification on held-out spectra; datasets comprise 648 bovine, 1,554 caprine and 2,392 ovine spectra. The bibliographic records retrieved do not state an author affiliation.

Technology: FNN trained on MALDI-TOF milk spectra; Methods describes train/validation/test split.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
FNN trained on MALDI-TOF milk spectra; Methods describes train/validation/test split.
Test and result
Results reports species classification on held-out spectra; datasets comprise 648 bovine, 1,554 caprine and 2,392 ovine spectra.
Limitation / still unknown
External replication and real-world benefit were not established by the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/articles/PMC12101317?pdf=render

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The use of a convolutional neural network to automate radiologic scoring of computed tomography of paranasal sinuses.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A convolutional network segmented paranasal-sinus CT scans and automated Lund–Mackay scoring against manual annotations; this is a bounded imaging-method evaluation.

Technology: A convolutional neural network segments paranasal-sinus CT images and automates Lund–Mackay radiologic scoring.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A convolutional neural network segments paranasal-sinus CT images and automates Lund–Mackay radiologic scoring.
Limitation / still unknown
The reported segmentation and scoring performance is limited to the evaluated CT scans and manual annotations; external or prospective clinical utility was not shown.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1186/s12938-025-01376-7

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning identifies clinical sepsis phenotypes that translate to the plasma proteome.

Source check: Source checked. Not independent scientific confirmation.

Finding: In 384 prospective multicenter sepsis patients, clinical-data clustering identified three severity phenotypes; plasma proteomics characterized them and supervised ML classified them. This is an observational phenotype result, not evidence of treatment benefit. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: unsupervised clustering/PCA of routine ICU clinical variables defines phenotypes; supervised Random Forest classifies phenotypes, and mass-spectrometry proteomics characterizes them.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
unsupervised clustering/PCA of routine ICU clinical variables defines phenotypes; supervised Random Forest classifies phenotypes, and mass-spectrometry proteomics characterizes them.
Test and result
384 prospective sepsis patients formed three severity phenotypes with different SOFA/mortality trajectories and corresponding plasma-proteome patterns.
Limitation / still unknown
Methods: 70/30 training/test split and 100-iteration Monte Carlo cross-validation for classifier evaluation. The observed multicenter clinical outcomes and proteomics support phenotype characterization; this does not show a treatment effect.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s15010-025-02628-3 · https://api.biorxiv.org/details/medrxiv/2025-04-11/2025-04-12/0 · https://www.medrxiv.org/content/10.1101/2025.04.11.25325574v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Inferring protein from transcript abundances using convolutional neural networks

Source check: Source checked. Not independent scientific confirmation.

Finding: Specific biological sequence-feature and protein-abundance result goes beyond a generic benchmark: learned CNN features identify interpretable motifs in two organisms. Cross-validation and limited predictive power are stated; motif causal effects remain hypotheses. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: CNN

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CNN trained/evaluated over human and Arabidopsis datasets using mRNA/protein sequence and abundance; repeated cross-validation. Predictive R² 0.30 (human), 0.32 (Arabidopsis); learned sequence motifs recover known regulatory signals, including translation initiation and UTR motifs. Local sequence-only predictions explain limited variance and do not establish causality.
Test and result
Specific biological sequence-feature and protein-abundance result goes beyond a generic benchmark: learned CNN features identify interpretable motifs in two organisms. Cross-validation and limited predictive power are stated; motif causal effects remain hypotheses.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13040-025-00434-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A weakly supervised deep learning framework for automated PD-L1 expression analysis in lung cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding: Weakly supervised multiple-instance learning estimated lung-cancer PD-L1 scores with pathologist agreement in internal and external patient cohorts. The bibliographic records retrieved do not state an author affiliation.

Technology: ResNet18 plus attention and multiple-instance learning use slide-level pathologist labels; patient-level split produced internal and external evaluation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ResNet18 plus attention and multiple-instance learning use slide-level pathologist labels; patient-level split produced internal and external evaluation.
Test and result
Pathologist-referenced TPS validation reports ICC 0.960 and external accuracy 81.7%; diagnostic assistance only, with prospective utility untested.
Limitation / still unknown
Weakly supervised multiple-instance learning estimated lung-cancer PD-L1 scores with pathologist agreement in internal and external patient cohorts. Primary Results describe the specific comparator, cohort or experimental check; limitations are stated narrowly.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1540087/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Acute mountain sickness prediction: a concerto of multidimensional phenotypic data and machine learning strategies in the framework of predictive, preventive, and personalized medicine.

Source check: Source checked. Not independent scientific confirmation.

Finding: 83 male volunteers, 66 analyzed, 12 biomarkers, MI-radial SVM/RFE; train AUC 0.97 and validation n=24 AUC 0.94. Docking is computational and does not establish mechanism. Methods/Results: 83 male volunteers, 66 analyzed, 12 biomarkers, MI-radial SVM/RFE; train AUC 0.97 and validation n=24 AUC 0.94. Testing: Methods/Results: 83 male volunteers, 66 analyzed, 12 biomarkers, MI-radial SVM/RFE; train AUC 0.97 and validation n=24 AUC 0.94. Docking is computational and does not establish mechanism.. A biomarker-based ML model predicts acute mountain sickness in a small, narrowly composed cohort, with a held-out validation cohort and reported AUC. Evidence is limited to 66 Han Chinese male volunteers after quality control. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: 83 male volunteers, 66 analyzed, 12 biomarkers, MI-radial SVM/RFE; train AUC 0.97 and validation n=24 AUC 0.94. Docking is computational and does not establish mechanism.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
83 male volunteers, 66 analyzed, 12 biomarkers, MI-radial SVM/RFE; train AUC 0.97 and validation n=24 AUC 0.94. Docking is computational and does not establish mechanism.
Test and result
83 male volunteers, 66 analyzed, 12 biomarkers, MI-radial SVM/RFE; train AUC 0.97 and validation n=24 AUC 0.94. Docking is computational and does not establish mechanism.
Limitation / still unknown
A biomarker-based ML model predicts acute mountain sickness in a small, narrowly composed cohort, with a held-out validation cohort and reported AUC. Evidence is limited to 66 Han Chinese male volunteers after quality control.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s13167-025-00404-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Automated mitochondrial oxygen consumption (mitoVO2) analysis via a bi-directional long short-term memory neural network.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Bi-directional LSTM classifies start/measurement points in mitoPO2 sequences, then modified Michaelis–Menten fit calculates mitoVO2. Dataset A (82 subjects) trains; independent Dataset B (43 subjects) is compared with three-rater consensus; 91% starts within ±5 samples and mean bias −0.057 mmHg s−1.

Technology: The cited full-text method identifies the model operation and how it is linked to the represented field result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The cited full-text method identifies the model operation and how it is linked to the represented field result.
Limitation / still unknown
The separate 43-subject dataset supports agreement with three-rater consensus in that sample; performance in other devices or clinical settings was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s10877-025-01291-1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A novel deep learning-based model for automated tooth detection and numbering in mixed and permanent dentition in occlusal photographs.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary full text supports the specific AI-enabled finding: A deep model detected and numbered teeth in occlusal photographs with reported high sensitivity. The bibliographic records retrieved do not state an author affiliation.

Technology: The primary full text supports the specific AI-enabled finding: A deep model detected and numbered teeth in occlusal photographs with reported high sensitivity.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary full text supports the specific AI-enabled finding: A deep model detected and numbered teeth in occlusal photographs with reported high sensitivity.
Test and result
The primary full text supports the specific AI-enabled finding: A deep model detected and numbered teeth in occlusal photographs with reported high sensitivity.
Limitation / still unknown
The primary full text supports the specific AI-enabled finding: A deep model detected and numbered teeth in occlusal photographs with reported high sensitivity. The paper reports the relevant cohort/image/field dataset, comparator or reference, and validation/test evaluation; the judgment does not imply prospective clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11955110/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Degrees of Uncertainty: Conformal Deep Learning for Core Body Temperature Prediction in Extreme Environments

Source check: Source checked. Not independent scientific confirmation.

Finding: The preprint compares temperature estimates and intervals with ECTemp in the reported monitoring evaluation. No evidence of improved health outcomes or prospective operational benefit is reported in the reviewed case notes.

Technology: Bidirectional LSTM with conformal prediction

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A bidirectional LSTM maps time series from wearable measurements to core-temperature estimates, with conformal prediction used to calibrate their uncertainty.
Test and result
The preprint compares temperature estimates and intervals with ECTemp in the reported monitoring evaluation.
Limitation / still unknown
No evidence of improved health outcomes or prospective operational benefit is reported in the reviewed case notes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-5705569/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and evaluation of a machine learning model for osteoporosis risk prediction in Korean women.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: An XGBoost model classified osteoporosis risk in a large Korean women's population cohort.

Technology: An XGBoost model classified osteoporosis risk in a large Korean women's population cohort.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
An XGBoost model classified osteoporosis risk in a large Korean women's population cohort.
Limitation / still unknown
The cohort-specific osteoporosis classifier does not establish prospective screening performance, external generalization or improved health outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11951505/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The interpretable machine learning model for depression associated with heavy metals via EMR mining method.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary NHANES analysis applies interpretable ML to depression and blood/urine metal data and reports model associations; the imbalanced cohort performance is modest and cannot establish causation or metal-related depression mechanisms. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: The primary NHANES analysis applies interpretable ML to depression and blood/urine metal data and reports model associations; the imbalanced cohort performance is modest and cannot establish causation or metal-related depression mechanisms.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary NHANES analysis applies interpretable ML to depression and blood/urine metal data and reports model associations; the imbalanced cohort performance is modest and cannot establish causation or metal-related depression mechanisms.
Test and result
The primary NHANES analysis applies interpretable ML to depression and blood/urine metal data and reports model associations; the imbalanced cohort performance is modest and cannot establish causation or metal-related depression mechanisms.
Limitation / still unknown
The primary NHANES analysis applies interpretable ML to depression and blood/urine metal data and reports model associations; the imbalanced cohort performance is modest and cannot establish causation or metal-related depression mechanisms.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-025-95938-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Explainable machine learning to compare the overall survival status between patients receiving mastectomy and breast conserving surgeries.

Source check: Source checked. Not independent scientific confirmation.

Finding: Nature full text reports explainable gradient-boosting survival models comparing mastectomy and breast-conserving surgery in 2,509 METABRIC patients, with SHAP-derived predictors and held-out test metrics. The bibliographic records retrieved do not state an author affiliation.

Technology: Gradient boosting models and SHAP explain predictors of overall survival by surgery group.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Gradient boosting models and SHAP explain predictors of overall survival by surgery group.
Test and result
Test accuracy and SHAP predictors differ between mastectomy and BCS groups; this is a retrospective association, not causal treatment guidance.
Limitation / still unknown
Nature full text reports explainable gradient-boosting survival models comparing mastectomy and breast-conserving surgery in 2,509 METABRIC patients, with SHAP-derived predictors and held-out test metrics. Held-out test metrics and feature-importance analysis support the bounded retrospective diagnostic/prognostic claim.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-025-91064-2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning for risk prediction of acute kidney injury in patients with diabetes mellitus combined with heart failure during hospitalization.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary study uses six ML algorithms on 1,457 ICU patients to predict AKI in a defined diabetic-heart-failure population and identifies risk factors; it is a single-center retrospective model without external validation, so no clinical benefit is established.

Technology: The primary study uses six ML algorithms on 1,457 ICU patients to predict AKI in a defined diabetic-heart-failure population and identifies risk factors; it is a single-center retrospective model without external validation, so no clinical benefit is established.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary study uses six ML algorithms on 1,457 ICU patients to predict AKI in a defined diabetic-heart-failure population and identifies risk factors; it is a single-center retrospective model without external validation, so no clinical benefit is established.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-025-87268-1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

MRI transformer deep learning and radiomics for predicting IDH wild type TERT promoter mutant gliomas.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary publisher Methods/Results support the AI-enabled result: CrossFormer-based MRI deep learning and radiomics stacking model in 1185 glioma patients; retrospective clinically specific assessment of IDH-wt/TERTp-mut prediction with survival stratification.

Technology: Methods establish the case-specific AI operation: CrossFormer-based MRI deep learning and radiomics stacking model in 1185 glioma patients; retrospective clinically specific assessment of IDH-wt/TERTp-mut prediction with survival stratification.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods establish the case-specific AI operation: CrossFormer-based MRI deep learning and radiomics stacking model in 1185 glioma patients; retrospective clinically specific assessment of IDH-wt/TERTp-mut prediction with survival stratification.
Limitation / still unknown
Full text states the cohort/data, comparison and evaluation for this bounded claim.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41698-025-00884-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial intelligence accelerates the identification of nature-derived potent LOXL2 inhibitors.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Graph-convolutional DeepVS screened natural-product libraries and prioritized Forsythoside A; CT26 cell assays measured proliferation, migration, apoptosis and LOXL2 expression.

Technology: Graph-convolutional DeepVS screened natural-product libraries and prioritized Forsythoside A; CT26 cell assays measured proliferation, migration, apoptosis and LOXL2 expression.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Graph-convolutional DeepVS screened natural-product libraries and prioritized Forsythoside A; CT26 cell assays measured proliferation, migration, apoptosis and LOXL2 expression.
Limitation / still unknown
Cell-assay findings in CT26 models do not establish in-vivo antitumor efficacy, safety or human therapeutic benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11950171/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning of Serum Cytokine and Chemokine Profiles Can Classify Inflammatory Bowel Disease Beyond Clinical Diagnosis

Source check: Source checked. Not independent scientific confirmation.

Finding: 20 serum cytokines/chemokines from 69 active IBD patients were clustered using ML; among 59 patients treated with molecular-targeted therapy, steroid-free remission rates differed across five clusters (trend p=.043). Proof-of-concept small single-center cohort; no external validation. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Unsupervised machine-learning clustering

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
20 serum cytokines/chemokines from 69 active IBD patients were clustered using ML; among 59 patients treated with molecular-targeted therapy, steroid-free remission rates differed across five clusters (trend p=.043). Proof-of-concept small single-center cohort; no external validation.
Test and result
20 serum cytokines/chemokines from 69 active IBD patients were clustered using ML; among 59 patients treated with molecular-targeted therapy, steroid-free remission rates differed across five clusters (trend p=.043). Proof-of-concept small single-center cohort; no external validation.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1016/j.gastha.2025.100667 · https://pmc.ncbi.nlm.nih.gov/articles/PMC12144436/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Accuracy of the AI-Based Smart Scope® Test as a Point-of-Care Screening and Triage Tool Compared to Colposcopy: A Pilot Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: In a prospective pilot of 130 women, Smart Scope® AI-assisted visual inspection with acetic acid (VIA) was compared with colposcopy at the same visit; positive findings triggered biopsy. Smart Scope AI accuracy was 76.53% against colposcopy (n=130). Histopathology was available for only 18 participants; against that reference, reported Smart Scope AI accuracy was 83.33%, sensitivity 83.33% and specificity 83.33%. The small histology subset leaves wide uncertainty and does not establish screening benefit in routine practice.

Technology: Cloud-based Smart Scope® AI analyzes digital cervical images for lesions during visual inspection with acetic acid.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Cloud-based AI assessed digital cervical images during VIA and returned lesion screening results; the study compared them with colposcopy and, in a subset, histopathology.
Test and result
All 130 participants received colposcopy and Smart Scope AI-assisted VIA during the same visit. Against colposcopy, AI accuracy was 76.53%; among 18 with histopathology, reported AI accuracy, sensitivity and specificity were each 83.33%.
Limitation / still unknown
Only 18 participants had histopathology, so the gold-standard comparison is imprecise. This single pilot does not establish external reproducibility or improved screening outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12022722/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 4/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Proposing a short version of the Unesp-Botucatu pig acute pain scale using a novel application of machine learning technique.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary study uses random-forest feature selection to propose a shorter pig acute-pain scale and evaluates it against painful/pain-free states; this is a specific veterinary measurement result, not evidence of improved animal outcomes.

Technology: The primary study uses random-forest feature selection to propose a shorter pig acute-pain scale and evaluates it against painful/pain-free states; this is a specific veterinary measurement result, not evidence of improved animal outcomes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary study uses random-forest feature selection to propose a shorter pig acute-pain scale and evaluates it against painful/pain-free states; this is a specific veterinary measurement result, not evidence of improved animal outcomes.
Limitation / still unknown
The primary study uses random-forest feature selection to propose a shorter pig acute-pain scale and evaluates it against painful/pain-free states; this is a specific veterinary measurement result, not evidence of improved animal outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-025-91551-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

GBCHV an advanced deep learning anatomy aware model for accurate classification of gallbladder cancer utilizing ultrasound images.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Nature full text reports GBCHV-Trans, an anatomy-aware transformer for gallbladder-cancer classification from ultrasound, with ablations and comparative diagnostic evaluation.

Technology: The transformer and image-processing pipeline perform the ultrasound classification.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The transformer and image-processing pipeline perform the ultrasound classification.
Limitation / still unknown
Classification results are bounded to the evaluated ultrasound data and comparisons; prospective clinical utility and patient outcomes were not tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-025-89232-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Enhancing Slip, Trip, and Fall Prevention: Real-World Near-Fall Detection with Advanced Machine Learning Technique

Source check: Source checked. Not independent scientific confirmation.

Finding: Models detected near-falls from real occupational IMU data, a field safety result evaluated on the Prev-Fall dataset. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Neural-network classifier

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe real-world IMU data and near-fall labels; Results report neural-network performance on recorded events.
Test and result
Models detected near-falls from real occupational IMU data, a field safety result evaluated on the Prev-Fall dataset.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11902511/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning, clinical-radiomics approach with HIM for hemorrhagic transformation prediction after thrombectomy and treatment.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A clinical-radiomics model using post-thrombectomy imaging and clinical features predicted hemorrhagic transformation after stroke treatment.

Technology: build clinical, radiomics and combined predictors from post-thrombectomy images; Results report hemorrhagic-transformation performance.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
build clinical, radiomics and combined predictors from post-thrombectomy images; Results report hemorrhagic-transformation performance.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11903287/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting postoperative cognitive dysfunction in older cardiac surgery patients: An integrated machine learning approach with a visual nomogram

Source check: Source checked. Not independent scientific confirmation.

Finding: An integrated machine-learning model predicts postoperative cognitive dysfunction in older cardiac-surgery patients. Prospective cohort includes 353 older patients; LASSO selects risk factors and the model reports POCD incidence and prediction performance. Testing: Single-center cohort and observational predictors limit transportability and causal interpretation.. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Prospective cohort includes 353 older patients; LASSO selects risk factors and the model reports POCD incidence and prediction performance.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Prospective cohort includes 353 older patients; LASSO selects risk factors and the model reports POCD incidence and prediction performance.
Test and result
Single-center cohort and observational predictors limit transportability and causal interpretation.
Limitation / still unknown
An integrated machine-learning model predicts postoperative cognitive dysfunction in older cardiac-surgery patients.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-6007903/v1.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Role of Biological Age in the Determination of Long-Term Cause-Specific Death Following Percutaneous Coronary Interventions.

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary Methods/Results establish the AI-enabled result within the stated scope and limitations. The bibliographic records retrieved do not state an author affiliation.

Technology: Primary Methods/Results establish the AI-enabled result within the stated scope and limitations.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Primary Methods/Results establish the AI-enabled result within the stated scope and limitations.
Test and result
CNN-derived ECG age gap evaluated in 535 prospective PCI patients; 124 deaths over median 2.61 years; gap associated with all-cause/CV mortality (HR about 1.07 per year). Full Methods/Results read. Association only, no causal or intervention claim.
Limitation / still unknown
Primary Methods/Results establish the AI-enabled result within the stated scope and limitations. Methods and Results establish cohort, comparator/ground truth and reported evaluation. Limitations are preserved in reason.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12132676/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 4/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Early detection of feline chronic kidney disease via 3-hydroxykynurenine and machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary author manuscript reports 3-hydroxykynurenine as a feline CKD biomarker evaluated with ML; the result is a veterinary diagnostic association and does not establish prospective clinical utility.

Technology: The primary author manuscript reports 3-hydroxykynurenine as a feline CKD biomarker evaluated with ML; the result is a veterinary diagnostic association and does not establish prospective clinical utility.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary author manuscript reports 3-hydroxykynurenine as a feline CKD biomarker evaluated with ML; the result is a veterinary diagnostic association and does not establish prospective clinical utility.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pureadmin.qub.ac.uk/ws/portalfiles/portal/632707187/Feline.pdf · https://www.nature.com/articles/s41598-025-90019-x · https://api.crossref.org/works/10.1038/s41598-025-90019-x

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Quality assurance and validity of AI-generated single best answer questions.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Generative AI produced single-best-answer medical examination questions that underwent student testing and expert review; the result concerns question quality, not general medical competence.

Technology: Generative AI produced single-best-answer medical examination questions that were assessed by student testing and expert review.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Generative AI produced single-best-answer medical examination questions that were assessed by student testing and expert review.
Limitation / still unknown
The result concerns question quality in the study’s student and expert evaluations; it does not establish improved clinical knowledge or practice.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11854382/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting 90-day risk of urinary tract infections following urostomy in bladder cancer patients using machine learning and explainability

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary study develops retrospective ML prediction of 90-day UTI risk after urostomy in a defined bladder-cancer cohort with follow-up; its single-center predictive result does not establish prospective benefit. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Machine-learning risk prediction model

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary study develops retrospective ML prediction of 90-day UTI risk after urostomy in a defined bladder-cancer cohort with follow-up; its single-center predictive result does not establish prospective benefit.
Test and result
The primary study develops retrospective ML prediction of 90-day UTI risk after urostomy in a defined bladder-cancer cohort with follow-up; its single-center predictive result does not establish prospective benefit.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-025-91075-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Enhancing resilience of distributed DC microgrids against cyber attacks using a transformer-based Kalman filter estimator.

Source check: Source checked. Not independent scientific confirmation.

Finding: The transformer/Kalman-filter estimator is tested in simulated DC microgrid attack/delay scenarios; it improves signal/state estimation and tracking versus classical estimator baselines. This is a simulation-based result for a specific microgrid control question, with no operational attack detail reproduced and no field deployment claimed. The bibliographic records retrieved do not state an author affiliation.

Technology: Original Methods examined; learned operation linked to the stated result as summarized in reason.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Original Methods examined; learned operation linked to the stated result as summarized in reason.
Test and result
Original Results/Conclusion passages examined; outcome and limitations summarized in reason.
Limitation / still unknown
The transformer/Kalman-filter estimator is tested in simulated DC microgrid attack/delay scenarios; it improves signal/state estimation and tracking versus classical estimator baselines. This is a simulation-based result for a specific microgrid control question, with no operational attack detail reproduced and no field deployment claimed. Original Methods/Results establish the cohort, reference/comparison and validation scope as summarized in reason.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-025-90959-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Finding purpose: Integrated latent profile and machine learning analyses identify purpose in life as an important predictor of high-functioning recovery after alcohol treatment.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A secondary Project MATCH analysis uses recursive partitioning and random forests to identify modestly predictive three-year alcohol-recovery profiles from end-of-treatment measures. Evidence is exploratory and internally validated only.

Technology: Recursive partitioning and random forests operate on 28 end-of-treatment biopsychosocial predictors to form/predict three-year recovery profiles and rank predictors.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Recursive partitioning and random forests operate on 28 end-of-treatment biopsychosocial predictors to form/predict three-year recovery profiles and rank predictors.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12184658/?report=xml

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Large-scale transcript variants dictate neoepitopes for cancer immunotherapy.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: AI-guided transcript-variant analysis identifies large-scale transcript variants that encode tumor neoepitopes relevant to cancer immunotherapy.

Technology: The study combines ribosome-nascent-chain transcript sequencing and predictive models, then reports experimental validation of selected neoepitopes/immunogenicity.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study combines ribosome-nascent-chain transcript sequencing and predictive models, then reports experimental validation of selected neoepitopes/immunogenicity.
Limitation / still unknown
The reported transcript-variant and neoepitope findings do not by themselves establish patient response to immunotherapy or clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11784853/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Evaluating waist-to-hip ratio in youth using frequency-modulated continuous wave radar and machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary study applies ML to radar measurements for youth waist-to-hip-ratio assessment, a specific health-measurement result; the paper calls for more participants and further evaluation before health-risk use. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: The primary study applies ML to radar measurements for youth waist-to-hip-ratio assessment, a specific health-measurement result; the paper calls for more participants and further evaluation before health-risk use.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary study applies ML to radar measurements for youth waist-to-hip-ratio assessment, a specific health-measurement result; the paper calls for more participants and further evaluation before health-risk use.
Test and result
The primary study applies ML to radar measurements for youth waist-to-hip-ratio assessment, a specific health-measurement result; the paper calls for more participants and further evaluation before health-risk use.
Limitation / still unknown
The primary study applies ML to radar measurements for youth waist-to-hip-ratio assessment, a specific health-measurement result; the paper calls for more participants and further evaluation before health-risk use.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-025-88098-x

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Risk factors and machine learning prediction models for intrahepatic cholestasis of pregnancy.

Source check: Source checked. Not independent scientific confirmation.

Finding: Retrospective single-center study includes 798 pregnant participants across normal, mild and severe groups, with train/test model comparison. The bibliographic records retrieved do not state an author affiliation.

Technology: Machine-learning models predict intrahepatic cholestasis of pregnancy severity and identify associated clinical factors.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning models predict intrahepatic cholestasis of pregnancy severity and identify associated clinical factors.
Test and result
Retrospective single-center study includes 798 pregnant participants across normal, mild and severe groups, with train/test model comparison.
Limitation / still unknown
Retrospective single-center study includes 798 pregnant participants across normal, mild and severe groups, with train/test model comparison. Single-center retrospective prediction; it does not establish prevention or treatment benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.ncbi.nlm.nih.gov/research/bionlp/RESTful/pmcoa.cgi/BioC_json/PMC11780866/unicode

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Enhancing individual glomerular filtration rate assessment: can we trust the equation? Development and validation of machine learning models to assess the trustworthiness of estimated GFR compared to

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A random forest estimates whether creatinine-based EKFC eGFR is within 30% of measured GFR; external validation across 10,107 participants produced AUC 0.675 and accuracy 0.716, with specificity 0.485 at the stated threshold. It marginally improves population-level trustworthiness and supports individual uncertainty assessment, not replacement of measured GFR.

Technology: Four ML classifiers predict eGFR agreement with measured GFR from routine features.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Four ML classifiers predict eGFR agreement with measured GFR from routine features.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12882-025-03972-0

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 4/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Using a Neural Network Architecture for the Prediction of Neurologic Outcome for Out-of-Hospital Cardiac Arrests Using Hospital Level Variables and Novel Physiologic Markers.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A neural network predicted neurologic outcomes among out-of-hospital cardiac-arrest patients who achieved return of circulation and reached hospital. Reported performance is hypothesis-generating in a small retrospective cohort.

Technology: Primary full text checked: Methods define the selected cohort and clinical/physiologic inputs; Results report held-out AUC, with small biomarker subset and sample-size limits.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Primary full text checked: Methods define the selected cohort and clinical/physiologic inputs; Results report held-out AUC, with small biomarker subset and sample-size limits.
Limitation / still unknown
This small retrospective hospital cohort supports hypothesis generation; external validation and prospective neurologic-outcome prediction were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11852285/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Upskilling or deskilling? Measurable role of an AI-supported training for radiology residents: a lesson from the pandemic.

Source check: Source checked. Not independent scientific confirmation.

Finding: Controlled pilot with eight residents and 150 CXRs found improved accuracy and error resilience with AI-supported training; limited to short-term simulated reading. The bibliographic records retrieved do not state an author affiliation.

Technology: Controlled pilot with eight residents and 150 CXRs found improved accuracy and error resilience with AI-supported training; limited to short-term simulated reading.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Controlled pilot with eight residents and 150 CXRs found improved accuracy and error resilience with AI-supported training; limited to short-term simulated reading.
Test and result
Controlled pilot with eight residents and 150 CXRs found improved accuracy and error resilience with AI-supported training; limited to short-term simulated reading.
Limitation / still unknown
Controlled pilot with eight residents and 150 CXRs found improved accuracy and error resilience with AI-supported training; limited to short-term simulated reading. Original Methods/Results validation design examined; interpretation is bounded to that design.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13244-024-01893-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Accuracy of remote, video-based supraventricular tachycardia detection in patients undergoing elective electrical cardioversion: a prospective cohort.

Source check: Source checked. Not independent scientific confirmation.

Finding: Prospective cohort in which an SVM on remote photoplethysmography detects arrhythmias during cardioversion, validated with leave-one-subject-out against ECG reference. Testing: The source reports the defined cohort/experiment and its reference or test comparison; interpretation is restricted to that study design.. The primary source Methods/Results confirm the represented result: Prospective cohort in which an SVM on remote photoplethysmography detects arrhythmias during cardioversion, validated with leave-one-subject-out against ECG reference. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Prospective cohort in which an SVM on remote photoplethysmography detects arrhythmias during cardioversion, validated with leave-one-subject-out against ECG reference.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Prospective cohort in which an SVM on remote photoplethysmography detects arrhythmias during cardioversion, validated with leave-one-subject-out against ECG reference.
Test and result
The source reports the defined cohort/experiment and its reference or test comparison; interpretation is restricted to that study design.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s10877-025-01263-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of biomarkers for the diagnosis of type 2 diabetes mellitus with metabolic associated fatty liver disease by bioinformatics analysis and experimental validation.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Integrated transcriptomic analysis and machine-learning feature selection prioritized T2DM-associated fatty-liver biomarkers, followed by expression checks in an independent/in-house cohort; findings are limited to the tested cohorts.

Technology: Machine-learning feature selection on integrated transcriptomic data prioritized biomarkers, followed by gene-expression validation in an independent/in-house cohort.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning feature selection on integrated transcriptomic data prioritized biomarkers, followed by gene-expression validation in an independent/in-house cohort.
Limitation / still unknown
Biomarker expression checks in the studied cohorts do not establish causality, prospective diagnostic performance or clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/fendo/articles/10.3389/fendo.2025.1512503/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Optimized graph neural network-multilayer perceptron fusion classifier for metastatic prostate cancer detection in Western and Asian populations.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Reactome-constrained GNN P-NETwa with downstream MLP and DeepLIFT attribution selects features and predicts metastatic prostate cancer. 18-gene panel accuracy/F1 .88/.75 internal test and .87/.85 external validation; selected genes are predictive signatures, not proven causal drivers. 1,360 development samples split 80/10/10 and two independent external datasets; retrospective public cohorts and Asian subgroup imbalance constrain use.

Technology: Reactome-constrained GNN P-NETwa with downstream MLP and DeepLIFT attribution selects features and predicts metastatic prostate cancer.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Reactome-constrained GNN P-NETwa with downstream MLP and DeepLIFT attribution selects features and predicts metastatic prostate cancer.
Limitation / still unknown
Both validations use retrospective public cohorts, with subgroup imbalance; the predictive gene panel does not establish causal drivers or prospective clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12490683/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Establishing a preoperative predictive model for gallbladder adenoma and cholesterol polyps based on machine learning: a multicentre retrospective study.

Source check: Source checked. Not independent scientific confirmation.

Finding: The multicenter model distinguishes pathology-confirmed gallbladder adenomas from cholesterol polyps and has internal/external cohort evaluation; the linked correction changes funding disclosure only, not the represented result. The bibliographic records retrieved do not state an author affiliation.

Technology: primary cohort 1,009 (706 training/303 internal) plus 139 external; 12 algorithms/110 combinations, SVM+RF AUC .972/.922; diagnosis based on postoperative pathology. Publisher correction DOI 10.1186/s12957-025-03724-2 states only funding acknowledgement correction, no result change.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
primary cohort 1,009 (706 training/303 internal) plus 139 external; 12 algorithms/110 combinations, SVM+RF AUC .972/.922; diagnosis based on postoperative pathology. Publisher correction DOI 10.1186/s12957-025-03724-2 states only funding acknowledgement correction, no result change.
Test and result
primary cohort 1,009 (706 training/303 internal) plus 139 external; 12 algorithms/110 combinations, SVM+RF AUC .972/.922; diagnosis based on postoperative pathology. Publisher correction DOI 10.1186/s12957-025-03724-2 states only funding acknowledgement correction, no result change.
Limitation / still unknown
External replication and real-world benefit were not established by the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12957-025-03671-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Determination of sexual dimorphism with CBCT images of the frontal sinus using a predictive formula and an artificial neural network.

Source check: Source checked. Not independent scientific confirmation.

Finding: On CBCT frontal-sinus measurements, an ANN estimated sex in a Brazilian forensic sample of 1,000 scans (800 model-development, 200 test) with 76% accuracy versus 73.5% for the predictive formula; sensitivity was lower, limiting scope to this sample and use.

Technology: ANN classifies sex from frontal-sinus dimensions selected using regression analysis.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ANN classifies sex from frontal-sinus dimensions selected using regression analysis.
Test and result
On 1,000 CBCT scans (800 for model construction, 200 held out for testing), an ANN estimated sex with 76% accuracy versus 73.5% for the predictive formula; sensitivity was lower.
Limitation / still unknown
Full SciELO article, Methods and Results, comparative regression/ANN metrics inspected; sample split details noted in paper.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.scielo.br/j/jaos/a/ypFJH37sB8yJvvVJQ7JtZ7w/ · https://www.scielo.br/j/jaos/a/ypFJH37sB8yJvvVJQ7JtZ7w/?format=pdf&lang=en

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Stratifying vascular disease patients into homogeneous subgroups using machine learning and FLAIR MRI biomarkers.

Source check: Source checked. Not independent scientific confirmation.

Finding: Five imaging-defined groups showed distinct vascular and cerebral characteristics in the observed atherosclerosis cohort. The bibliographic records retrieved do not state an author affiliation.

Technology: K-means grouped regional FLAIR features, with a random forest identifying markers differentiating the groups.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
K-means grouped regional FLAIR features, with a random forest identifying markers differentiating the groups.
Test and result
In 379 patients the groups differed in clinical and imaging measures, including vascular risks and white matter lesion burden.
Limitation / still unknown
Five imaging-defined groups showed distinct vascular and cerebral characteristics in the observed atherosclerosis cohort. Internal statistical comparison of reported clinical and lesion covariates supports exploratory cohort stratification. No independent replication or treatment effect was established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s44303-024-00063-x

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning model to diagnose cardiac amyloidosis from haematoxylin/eosin-stained myocardial tissue.

Source check: Source checked. Not independent scientific confirmation.

Finding: Deep learning localized and classified amyloid in HE-stained myocardial biopsy patches against pathologist/stain reference labels, a specific diagnostic validation result with single-centre limits.

Technology: deep learning scored 224×224 HE biopsy patches from 166 patients; pathological labels were established by pathologists and amyloid staining.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
deep learning scored 224×224 HE biopsy patches from 166 patients; pathological labels were established by pathologists and amyloid staining.
Test and result
patch AUC 0.965 and specimen-level AUC 1.0 in the evaluation set.
Limitation / still unknown
Retrospective single-centre 2:1 development/evaluation; no external validation, so perfect specimen AUC is limited to this small cohort.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11728699/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC11728699/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC11728699/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of sepsis-associated encephalopathy biomarkers through machine learning and bioinformatics approaches.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full primary article confirms transcriptomic ML identified 18 SAE candidate biomarkers, with a separate dataset and BV2-cell qPCR evidence; approve as candidate biology, not clinical biomarkers.

Technology: PMC Methods: three GEO datasets, WGCNA and Elastic Net/LASSO/RF/XGBoost; Results: 18 selected biomarkers, external GEO validation and in-vitro qPCR differential expression.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PMC Methods: three GEO datasets, WGCNA and Elastic Net/LASSO/RF/XGBoost; Results: 18 selected biomarkers, external GEO validation and in-vitro qPCR differential expression.
Limitation / still unknown
Candidate biomarkers and BV2-cell qPCR do not establish clinical diagnostic performance or causal disease mechanisms.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11685593/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Building Distant Metastasis Models for HNSCC Using Machine Learning Techniques

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Machine-learning models predict distant metastasis in head-and-neck squamous cell carcinoma from SEER clinical data.

Technology: The retrospective cohort includes 31,060 cases; four algorithms are compared and ROC/accuracy results reported.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The retrospective cohort includes 31,060 cases; four algorithms are compared and ROC/accuracy results reported.
Limitation / still unknown
The result is limited to the reported dataset, experimental setting and validation in this study; broader generalization and real-world benefit are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.21203/rs.3.rs-5667236/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting local control of brain metastases after stereotactic radiotherapy with clinical, radiomics and deep learning features.

Source check: Source checked. Not independent scientific confirmation.

Finding: retrospective stereotactic radiotherapy brain-metastasis cohort (129 patients); radiomic and deep-learning MRI features plus clinical covariates fed prediction models for local failure/control. Clinical+DL model accuracy 78%; clinical+radiomics 79.33%; models evaluated on held-out data. Clinical prediction result only; limited cohort and retrospective design, not treatment benefit. Methods/Results: retrospective stereotactic radiotherapy brain-metastasis cohort (129 patients); radiomic and deep-learning MRI features plus clinical covariates fed prediction models for local failure/control. Testing: Methods/Results: retrospective stereotactic radiotherapy brain-metastasis cohort (129 patients); radiomic and deep-learning MRI features plus clinical covariates fed prediction models for local failure/control. Clinical prediction result only; limited cohort and retrospective design, not treatment benefit.. ML feature models predict a concrete clinical outcome in a defined treated population and were assessed against observed local control; eligible retrospective diagnostic/prognostic validation under methodology.html, with single-center/sample limitations. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: retrospective stereotactic radiotherapy brain-metastasis cohort (129 patients); radiomic and deep-learning MRI features plus clinical covariates fed prediction models for local failure/control. Clinical+DL model accuracy 78%; clinical+radiomics 79.33%; models evaluated on held-out data. Clinical prediction result only; limited cohort and retrospective design, not treatment benefit.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
retrospective stereotactic radiotherapy brain-metastasis cohort (129 patients); radiomic and deep-learning MRI features plus clinical covariates fed prediction models for local failure/control. Clinical+DL model accuracy 78%; clinical+radiomics 79.33%; models evaluated on held-out data. Clinical prediction result only; limited cohort and retrospective design, not treatment benefit.
Test and result
retrospective stereotactic radiotherapy brain-metastasis cohort (129 patients); radiomic and deep-learning MRI features plus clinical covariates fed prediction models for local failure/control. Clinical+DL model accuracy 78%; clinical+radiomics 79.33%; models evaluated on held-out data. Clinical prediction result only; limited cohort and retrospective design, not treatment benefit.
Limitation / still unknown
ML feature models predict a concrete clinical outcome in a defined treated population and were assessed against observed local control; eligible retrospective diagnostic/prognostic validation under methodology.html, with single-center/sample limitations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13014-024-02573-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Construction of Clinical Predictive Models for Heart Failure Detection Using Six Different Machine Learning Algorithms: Identification of Key Clinical Prognostic Features.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary full text supports the case-specific result: Machine-learning models predicted heart-failure risk in patient data and report a held-out test-set evaluation. The bibliographic records retrieved do not state an author affiliation.

Technology: The primary full text supports the case-specific result: Machine-learning models predicted heart-failure risk in patient data and report a held-out test-set evaluation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary full text supports the case-specific result: Machine-learning models predicted heart-failure risk in patient data and report a held-out test-set evaluation.
Test and result
The primary full text supports the case-specific result: Machine-learning models predicted heart-failure risk in patient data and report a held-out test-set evaluation.
Limitation / still unknown
External replication and real-world benefit were not established by the reviewed study.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11693937/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction of early lung adenocarcinoma spread through air spaces by machine learning radiomics: a cross-center cohort study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Radiomics and machine-learning models using CT features predicted spread through air spaces in lung adenocarcinoma across cohorts, with cross-center evaluation; this is a preoperative pathology-prediction result limited by retrospective design.

Technology: Radiomic feature extraction and machine-learning classifiers used CT features to predict pathology-confirmed spread through air spaces.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Radiomic feature extraction and machine-learning classifiers used CT features to predict pathology-confirmed spread through air spaces.
Limitation / still unknown
The cross-center evaluation is retrospective; predictive association with pathology does not show prospective clinical utility or improved outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.21037/tlcr-24-565

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development of a Predictive Model of Occult Cancer After a Venous Thromboembolism Event Using Machine Learning: The CLOVER Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: CLOVER prospective VTE registry nested case-control (815 participants, 56 occult cancers) compares boosted-tree models; temporal split uses later cases as test. CatBoost gives the reported best performance. Testing: Small cancer-event count and temporal internal test; this is risk prediction, not validated population screening.. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: CLOVER prospective VTE registry nested case-control (815 participants, 56 occult cancers) compares boosted-tree models; temporal split uses later cases as test. CatBoost gives the reported best performance.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
CLOVER prospective VTE registry nested case-control (815 participants, 56 occult cancers) compares boosted-tree models; temporal split uses later cases as test. CatBoost gives the reported best performance.
Test and result
Small cancer-event count and temporal internal test; this is risk prediction, not validated population screening.
Limitation / still unknown
CLOVER prospective VTE registry nested case-control (815 participants, 56 occult cancers) compares boosted-tree models; temporal split uses later cases as test. CatBoost gives the reported best performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3390/medicina61010018 · https://www.mdpi.com/1648-9144/61/1/18

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A New Catalogue and Insights into the 2022 Adriatic Offshore Seismic Sequence Using a Machine Learning-Based Procedure.

Source check: Source checked. Not independent scientific confirmation.

Finding: A machine-learning phase-picking/association procedure produced a 2,780-event catalogue and new analyses of the 2022 Adriatic offshore seismic sequence. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Methods describe the ML catalogue procedure; Results report the newly assembled seismic-event catalogue and sequence analysis.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe the ML catalogue procedure; Results report the newly assembled seismic-event catalogue and sequence analysis.
Test and result
Methods describe the ML catalogue procedure; Results report the newly assembled seismic-event catalogue and sequence analysis.
Limitation / still unknown
Methods describe the ML catalogue procedure; Results report the newly assembled seismic-event catalogue and sequence analysis.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11722621/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Proteomic profiling of the local and systemic immune response to pediatric respiratory viral infections.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Proteomic profiling identifies local and systemic immune-response patterns in critically ill children with viral lower-respiratory infection and develops a diagnostic signature.

Technology: Mass-spectrometry/antibody proteomic measurements from tracheal aspirate and blood are compared across pediatric cohorts; machine learning builds a parsimonious classifier.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Mass-spectrometry/antibody proteomic measurements from tracheal aspirate and blood are compared across pediatric cohorts; machine learning builds a parsimonious classifier.
Limitation / still unknown
The signature is limited to the sampled critically ill children and infection types; independent prospective diagnostic performance was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11748518/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Structural alterations as a predictor of depression - a 7-Tesla MRI-based multidimensional approach.

Source check: Source checked. Not independent scientific confirmation.

Finding: Leave-one-out cross-validation on 41 patients and 41 controls plus permutation analysis identified MRI regions associated with classification; no external cohort was tested. The finding is bounded to the reported study and does not establish causal benefit or generalization.

Technology: SVM on 7-T structural MRI features classifies major depressive disorder

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
SVM on 7-T structural MRI features classifies major depressive disorder
Test and result
Leave-one-out cross-validation on 41 patients and 41 controls plus permutation analysis identified MRI regions associated with classification; no external cohort was tested.
Limitation / still unknown
External validation, causal effects, and sustained operational impact were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41380-024-02854-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting patients with septic shock and sepsis through analyzing whole-blood expression of NK cell-related hub genes using an advanced machine learning framework.

Source check: Source checked. Not independent scientific confirmation.

Finding: A learned six-gene SepxFindeR model distinguished sepsis from septic shock across datasets and was checked with newly recruited RT-qPCR samples. The bibliographic records retrieved do not state an author affiliation.

Technology: LDA, risk scores and LASSO ensemble combine six hub genes; four GEO cohorts and recruited clinical RT-qPCR samples test discrimination.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
LDA, risk scores and LASSO ensemble combine six hub genes; four GEO cohorts and recruited clinical RT-qPCR samples test discrimination.
Test and result
Specific sepsis-stage classification is reported; small recruited validation is preliminary and does not establish improved outcomes or deployment.
Limitation / still unknown
A learned six-gene SepxFindeR model distinguished sepsis from septic shock across datasets and was checked with newly recruited RT-qPCR samples. Primary Results describe the specific comparator, cohort or experimental check; limitations are stated narrowly.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2024.1493895/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning for predicting porosity in ultra-deep fractured vuggy reservoirs from the Shunbei oilfield in Tarim Basin, China.

Source check: Source checked. Not independent scientific confirmation.

Finding: Publisher PDF Methods and Results: seismic/log inputs trained a deep model for porosity, evaluated against blind wells and impedance inversion; prediction result is site-specific and data-limited. Testing: Publisher PDF Methods and Results: seismic/log inputs trained a deep model for porosity, evaluated against blind wells and impedance inversion; prediction result is site-specific and data-limited.. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Publisher PDF Methods and Results: seismic/log inputs trained a deep model for porosity, evaluated against blind wells and impedance inversion; prediction result is site-specific and data-limited.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher PDF Methods and Results: seismic/log inputs trained a deep model for porosity, evaluated against blind wells and impedance inversion; prediction result is site-specific and data-limited.
Test and result
Publisher PDF Methods and Results: seismic/log inputs trained a deep model for porosity, evaluated against blind wells and impedance inversion; prediction result is site-specific and data-limited.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1038/s41598-024-81051-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting place of delivery choice among childbearing women in East Africa: a comparative analysis of advanced machine learning techniques.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Using survey data from 86,009 women in East Africa, tuned SVM/CatBoost models classified facility versus home delivery and association-rule mining identified covariates; this is an observational population-prediction result, not causal evidence for the recommended interventions. The full text confirms survey design, the 86,009 analytic cohort, model comparisons and association-rule results.

Technology: Source identifies the learned method and its operation in the prediction/classification analysis; see Methods and Results. Abstract-only cases cannot fully confirm operation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Source identifies the learned method and its operation in the prediction/classification analysis; see Methods and Results. Abstract-only cases cannot fully confirm operation.
Limitation / still unknown
Using survey data from 86,009 women in East Africa, tuned SVM/CatBoost models classified facility versus home delivery and association-rule mining identified covariates; this is an observational population-prediction result, not causal evidence for the recommended interventions. The full text confirms survey design, the 86,009 analytic cohort, model comparisons and association-rule results.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/articles/PMC11631870

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Data-Driven De Novo Design of Super-Adhesive Hydrogels

Source check: Source checked. Not independent scientific confirmation.

Finding: Data-driven de novo design identifies super-adhesive hydrogels, with experimentally synthesized and tested candidate materials. This is a preprint, not a peer-reviewed publication.

Technology: Machine-learning formulation optimization

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The paper combines data-driven formulation selection with hydrogel fabrication and measured adhesion testing, reporting material performance.
Test and result
Data-driven de novo design identifies super-adhesive hydrogels, with experimentally synthesized and tested candidate materials.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-5491059/v1.pdf?c=1754566853000

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Assessing greenspace and cardiovascular health through deep-learning analysis of street-view imagery in a cohort of US children.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: In Project Viva, deep-learning segmentation of street-view imagery produced ground-level greenspace measures linked to residential buffers and cardiovascular-health scores over childhood/adolescence. A 1-SD increase in street-view trees within 500 m was cross-sectionally associated with a 1.92-point higher late-adolescent score; longitudinal and cumulative exposure associations were not found. The observational result is limited and non-causal.

Technology: A deep-learning image-segmentation algorithm estimates street-view trees/grass/other greenspace; these learned exposure measures are the explanatory variables in the cohort analysis.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A deep-learning image-segmentation algorithm estimates street-view trees/grass/other greenspace; these learned exposure measures are the explanatory variables in the cohort analysis.
Limitation / still unknown
In Project Viva, deep-learning segmentation of street-view imagery produced ground-level greenspace measures linked to residential buffers and cardiovascular-health scores over childhood/adolescence. A 1-SD increase in street-view trees within 500 m was cross-sectionally associated with a 1.92-point higher late-adolescent score; longitudinal and cumulative exposure associations were not found. The observational result is limited and non-causal.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11742899/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting efficacy in patients with locally advanced/metastatic urothelial carcinoma (mUC) treated with immunotherapy using explainable machine learning approaches: the SamUR-AI trial on behalf of the Meet-URO group.

Source check: Source checked. Not independent scientific confirmation.

Finding: The repository exposed a result-bearing thesis abstract by 20 Nov 2024 (thesis issued 11 Dec 2024); the peer-reviewed journal version appeared 30 Jan 2026. In the thesis and later 438-patient journal analysis, explainable ML had weak objective-response classification (best test F1 0.61) but moderate survival discrimination (OS C-index 0.67; PFS C-index 0.64); findings remain exploratory, not evidence of treatment benefit or validated biomarkers.

Technology: The study fits multiple classifiers and survival models to baseline clinical/laboratory variables and applies SHAP to interpret predicted ORR, PFS, and OS; the learned models produce the reported prediction results.

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study fits multiple classifiers and survival models to baseline clinical/laboratory variables and applies SHAP to interpret predicted ORR, PFS, and OS; the learned models produce the reported prediction results.
Test and result
For 438 patients, classification was weak (best test F1 0.61); extra survival trees reached OS C-index 0.67 and selected-feature survival SVM reached PFS C-index 0.64. SHAP highlighted ECOG status, treatment line/combinations, metastases and laboratory features.
Limitation / still unknown
In a retrospective 438-patient, 34-center cohort, explainable ML had weak objective-response classification (best test F1 0.61) but moderate survival discrimination (OS C-index 0.67; PFS C-index 0.64); SHAP highlighted clinical prognostic factors. These findings are exploratory and do not show treatment benefit or establish predictive biomarkers. Retrospective multicenter data from 34 Italian institutions were divided into training, cross-validation, and test sets. The article identifies class imbalance and overfitting and calls for further validation; results support only modest predictive performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://hdl.handle.net/10589/231007 · https://pmc.ncbi.nlm.nih.gov/articles/PMC13040919/ · https://politesi.polimi.it/oai/request?verb=GetRecord&metadataPrefix=datacite&identifier=oai:www.politesi.polimi.it:10589/231007 · https://pubmed.ncbi.nlm.nih.gov/41930317/

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Stacked Neural Network Model for Damage Localization.

Source check: Source checked. Not independent scientific confirmation.

Finding: A stacked neural network localized structural damage from vibration features, tested on simulated, finite-element and measured steel-beam data. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Methods describe generated and experimental vibration data and stacked networks; Results compare damage localization on simulated and measured cases.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe generated and experimental vibration data and stacked networks; Results compare damage localization on simulated and measured cases.
Test and result
Methods describe generated and experimental vibration data and stacked networks; Results compare damage localization on simulated and measured cases.
Limitation / still unknown
Methods describe generated and experimental vibration data and stacked networks; Results compare damage localization on simulated and measured cases.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11548493/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Ensemble feature selection and tabular data augmentation with generative adversarial networks to enhance cutaneous melanoma identification and interpretability.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The source compares LASSO, KNN, SVM and decision-tree classifiers using sensitivity, specificity and ROC analysis.

Technology: Ensemble feature selection and GAN-based augmentation improve cutaneous-melanoma identification from tabular/image-derived lesion features.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Ensemble feature selection and GAN-based augmentation improve cutaneous-melanoma identification from tabular/image-derived lesion features.
Limitation / still unknown
The reported classifier comparisons use the study dataset; external validation and clinical screening benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11526724/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Does the FARNet neural network algorithm accurately identify Posteroanterior cephalometric landmarks?

Source check: Source checked. Not independent scientific confirmation.

Finding: FARNet detects 47 cephalometric landmarks on 1,431 PA radiographs with a train/validation/test split; pointwise errors vary, from 0.94-mm right-gonion MRE to 3.31-mm right-condylon MRE, and performance is compared with a human expert. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: FARNet deep neural network

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
FARNet deep neural network locates anatomical landmarks on cephalograms.
Test and result
FARNet detects 47 cephalometric landmarks on 1,431 PA radiographs with a train/validation/test split; pointwise errors vary, from 0.94-mm right-gonion MRE to 3.31-mm right-condylon MRE, and performance is compared with a human expert.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12880-024-01478-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Improving Pelvic Floor Muscle Training with AI: A Novel Quality Assessment System for Pelvic Floor Dysfunction.

Source check: Source checked. Not independent scientific confirmation.

Finding: A sensor/CNN system detected and scored pelvic-floor contractions for a defined pelvic-floor-dysfunction assessment task, with separate data validation.

Technology: Methods describe contraction/non-contraction recognition and quality scoring; Results report model and strength-classification performance.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe contraction/non-contraction recognition and quality scoring; Results report model and strength-classification performance.
Test and result
Methods describe contraction/non-contraction recognition and quality scoring; Results report model and strength-classification performance.
Limitation / still unknown
Methods describe contraction/non-contraction recognition and quality scoring; Results report model and strength-classification performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11548714/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning in Spinal Endoscopy: U-Net Models for Neural Tissue Detection.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: U-Net segmentation detected neural tissue in videos from spinal endoscopy cases, with reported held-out segmentation metrics. This is a pilot technical result and does not establish surgical benefit.

Technology: Publisher full text checked: Methods use frames from 28 surgeries; Results report DSC/IoU and frame timing.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods use frames from 28 surgeries; Results report DSC/IoU and frame timing.
Limitation / still unknown
U-Net segmentation detected neural tissue in videos from spinal endoscopy cases, with reported held-out segmentation metrics. This is a pilot technical result and does not establish surgical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2306-5354/11/11/1082

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Estimation of Strawberry Canopy Volume in Unmanned Aerial Vehicle RGB Imagery Using an Object Detection-Based Convolutional Neural Network.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: An object-detection/CNN workflow estimated strawberry canopy volume from UAV RGB imagery against measured plant-volume references in a field experiment.

Technology: a 2023–24 strawberry field and measured-volume reference; Results compare CNN and geometric estimates on held-out images.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
a 2023–24 strawberry field and measured-volume reference; Results compare CNN and geometric estimates on held-out images.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11548564/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Pilot deployment of a machine-learning enhanced prediction of need for hemorrhage resuscitation after trauma - the ShockMatrix pilot study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Traumabase development cohort 28,614, NHR 11%; XGBoost F4 .76 [0.73,.78]; prospective app phase measured feasibility and compared clinician/model likelihood ratios, but article says it was not intended to assess model performance because underpowered. Methods/Results: Traumabase development cohort 28,614, NHR 11%; XGBoost F4 .76 [0.73,.78]; prospective app phase measured feasibility and compared clinician/model likelihood ratios, but article says it was not intended to assess model performance because underpowered. Testing: Methods/Results: Traumabase development cohort 28,614, NHR 11%; XGBoost F4 .76 [0.73,.78]; prospective app phase measured feasibility and compared clinician/model likelihood ratios, but article says it was not intended to assess model performance because underpowered.. XGBoost on nine prehospital predictors produced a defined NHR model and a real-time multicenter feasibility pilot; the pilot was explicitly underpowered to assess model performance. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Traumabase development cohort 28,614, NHR 11%; XGBoost F4 .76 [0.73,.78]; prospective app phase measured feasibility and compared clinician/model likelihood ratios, but article says it was not intended to assess model performance because underpowered.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Traumabase development cohort 28,614, NHR 11%; XGBoost F4 .76 [0.73,.78]; prospective app phase measured feasibility and compared clinician/model likelihood ratios, but article says it was not intended to assess model performance because underpowered.
Test and result
Traumabase development cohort 28,614, NHR 11%; XGBoost F4 .76 [0.73,.78]; prospective app phase measured feasibility and compared clinician/model likelihood ratios, but article says it was not intended to assess model performance because underpowered.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12911-024-02723-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Proof of concept study on early forecasting of antimicrobial resistance in hospitalized patients using machine learning and simple bacterial ecology data.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary proof-of-concept uses machine learning on 44,026 hospital microbiology records to estimate antimicrobial susceptibility before antibiogram availability, with stage-specific held-out performance. Evidence is retrospective at one hospital and does not show prospective treatment changes or outcomes.

Technology: The primary proof-of-concept uses machine learning on 44,026 hospital microbiology records to estimate antimicrobial susceptibility before antibiogram availability, with stage-specific held-out performance. Evidence is retrospective at one hospital and does not show prospective treatment changes or outcomes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary proof-of-concept uses machine learning on 44,026 hospital microbiology records to estimate antimicrobial susceptibility before antibiogram availability, with stage-specific held-out performance. Evidence is retrospective at one hospital and does not show prospective treatment changes or outcomes.
Limitation / still unknown
The primary proof-of-concept uses machine learning on 44,026 hospital microbiology records to estimate antimicrobial susceptibility before antibiogram availability, with stage-specific held-out performance. Evidence is retrospective at one hospital and does not show prospective treatment changes or outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/10.1038/s41598-024-71757-w

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Intra- and Postpartum Hemorrhage through Artificial Intelligence.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Prospective multicenter cohort n=203 (68 hemorrhage events) evaluates AI models for intra/postpartum hemorrhage. Naive Bayes sensitivity 96.3%, accuracy 98.6%; severe subgroup has only six cases and lower reported performance (82.4% accuracy, 89.3% sensitivity).

Technology: AI classifiers, including Naive Bayes, predicted intra- and postpartum hemorrhage from maternal measurements.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
AI classifiers, including Naive Bayes, predicted intra- and postpartum hemorrhage from maternal measurements.
Limitation / still unknown
The severe-hemorrhage subgroup contained only six cases and showed lower performance; the study does not establish performance in larger high-risk cohorts or patient-outcome benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3390/medicina60101604

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine-learning-based models for the optimization of post-cervical spinal laminoplasty outpatient follow-up schedules.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: LSTM predicts a defined 24-month PASS outcome from postoperative records with a prospective internal test set; small single-center cohort limits generalization.

Technology: 80 cervical laminoplasty patients; eight algorithms predict PASS (JOA ≥14.25) at 24 months from data through 12 months; LSTM AUROC .90±.13, prospective internal test AUROC .858±.007. No external cohort.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
80 cervical laminoplasty patients; eight algorithms predict PASS (JOA ≥14.25) at 24 months from data through 12 months; LSTM AUROC .90±.13, prospective internal test AUROC .858±.007. No external cohort.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12911-024-02693-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Efficacy of automated machine learning models and feature engineering for diagnosis of equivocal appendicitis using clinical and computed tomography findings.

Source check: Source checked. Not independent scientific confirmation.

Finding: PMC full article Methods and Results: retrospective 303 equivocal-CT adults; 213/90 train/test; AutoGluon clinical+CT AUROC 0.886 versus ridge 0.852; single-center diagnostic evidence only. Testing: PMC full article Methods and Results: retrospective 303 equivocal-CT adults; 213/90 train/test; AutoGluon clinical+CT AUROC 0.886 versus ridge 0.852; single-center diagnostic evidence only.. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: PMC full article Methods and Results: retrospective 303 equivocal-CT adults; 213/90 train/test; AutoGluon clinical+CT AUROC 0.886 versus ridge 0.852; single-center diagnostic evidence only.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PMC full article Methods and Results: retrospective 303 equivocal-CT adults; 213/90 train/test; AutoGluon clinical+CT AUROC 0.886 versus ridge 0.852; single-center diagnostic evidence only.
Test and result
PMC full article Methods and Results: retrospective 303 equivocal-CT adults; 213/90 train/test; AutoGluon clinical+CT AUROC 0.886 versus ridge 0.852; single-center diagnostic evidence only.
Limitation / still unknown
PMC full article Methods and Results: retrospective 303 equivocal-CT adults; 213/90 train/test; AutoGluon clinical+CT AUROC 0.886 versus ridge 0.852; single-center diagnostic evidence only.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1038/s41598-024-72889-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Efficiency assessment of intelligent patient-specific instrumentation in total knee arthroplasty: a prospective randomized controlled trial.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A prospective randomized 102-patient TKA trial evaluates CNN-generated patient-specific instrumentation against conventional instrumentation; the abstract describes a concrete surgical workflow comparison.

Technology: A prospective randomized 102-patient TKA trial evaluates CNN-generated patient-specific instrumentation against conventional instrumentation; the abstract describes a concrete surgical workflow comparison.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
A prospective randomized 102-patient TKA trial evaluates CNN-generated patient-specific instrumentation against conventional instrumentation; the abstract describes a concrete surgical workflow comparison.
Limitation / still unknown
The prospective randomized comparison involved 102 patients; long-term outcomes and generalization beyond the participating surgical setting were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13018-024-05010-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of immune patterns in idiopathic pulmonary fibrosis patients driven by PLA2G7-positive macrophages using an integrated machine learning survival framework.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Publisher PDF Methods and Results: three transcriptomic cohorts, ten ML methods/91 models, external cohort validation and cell/mouse analyses; retrospective biomarker association, not established human causality.

Technology: Publisher PDF Methods and Results: three transcriptomic cohorts, ten ML methods/91 models, external cohort validation and cell/mouse analyses; retrospective biomarker association, not established human causality.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher PDF Methods and Results: three transcriptomic cohorts, ten ML methods/91 models, external cohort validation and cell/mouse analyses; retrospective biomarker association, not established human causality.
Limitation / still unknown
Publisher PDF Methods and Results: three transcriptomic cohorts, ten ML methods/91 models, external cohort validation and cell/mouse analyses; retrospective biomarker association, not established human causality.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1038/s41598-024-73625-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Impact of a deep learning-based brain CT interpretation algorithm on clinical decision-making for intracranial hemorrhage in the emergency department.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: PMC full article Methods and Results: simulated ED workflow, 111 cases and 2,146 algorithm scans; DL aid increased inexperienced reader sensitivity while reducing specificity; experienced reader results not significant.

Technology: PMC full article Methods and Results: simulated ED workflow, 111 cases and 2,146 algorithm scans; DL aid increased inexperienced reader sensitivity while reducing specificity; experienced reader results not significant.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PMC full article Methods and Results: simulated ED workflow, 111 cases and 2,146 algorithm scans; DL aid increased inexperienced reader sensitivity while reducing specificity; experienced reader results not significant.
Limitation / still unknown
The 111-case simulated workflow improved sensitivity for inexperienced readers but reduced specificity; experienced-reader effects were not significant, and real-world impact was not tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1038/s41598-024-73589-0

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Interpretable machine learning models for the prediction of all-cause mortality and time to death in hemodialysis patients.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary full-text record describes a 359-patient retrospective hemodialysis cohort and evaluated ML mortality/time-to-death predictions; this supports a specific clinical prediction study, not clinical benefit.

Technology: Random forest, support-vector machine and logistic-regression models predict all-cause mortality; random-forest regression estimates time to death, with SHAP factor interpretation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random forest, support-vector machine and logistic-regression models predict all-cause mortality; random-forest regression estimates time to death, with SHAP factor interpretation.
Test and result
In 359 hemodialysis patients, the optimized full-variable random forest had mortality AUROC 0.86±0.07 (sensitivity 0.86, specificity 0.75); a 22-variable model had AUROC 0.80±0.06.
Limitation / still unknown
Retrospective single-cohort prediction does not establish improved survival or generalization to other dialysis populations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11879476/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Methodology and development of a machine learning probability calculator: Data heterogeneity limits ability to predict recurrence after arthroscopic Bankart repair.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The ML probability calculator is developed from pooled Bankart-repair literature with heterogeneity constraints and internally evaluated for recurrence; interpret as retrospective risk estimation. First publication date 2024-09-26 is corroborated by the primary bibliographic entry.

Technology: The ML probability calculator is developed from pooled Bankart-repair literature with heterogeneity constraints and internally evaluated for recurrence; interpret as retrospective risk estimation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The ML probability calculator is developed from pooled Bankart-repair literature with heterogeneity constraints and internally evaluated for recurrence; interpret as retrospective risk estimation.
Limitation / still unknown
The model was built from heterogeneous published cohorts and only internally evaluated; external prospective recurrence prediction was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11948171/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Routine ICU Surveillance after Brain Tumor Surgery: Patient Selection Using Machine Learning

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: XGBoost predicted early adverse events after brain-tumor surgery in a retrospective cohort of 1,000 patients; preoperative-only cross-validated ROC-AUC was 0.76 and combined pre/postoperative AUC 0.81. The outcome is a specific patient-selection question, with retrospective single-center validation limits.

Technology: XGBoost gradient boosting on 27 perioperative features; retrospective 1,000 consecutive adult tumor resections. Repeated 5×5 CV; nine adverse-event/intervention outcomes within 24 hours; pre-op-only AUC 0.76, combined pre/post-op AUC 0.81.

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
XGBoost gradient boosting on 27 perioperative features; retrospective 1,000 consecutive adult tumor resections. Repeated 5×5 CV; nine adverse-event/intervention outcomes within 24 hours; pre-op-only AUC 0.76, combined pre/post-op AUC 0.81.
Limitation / still unknown
The result is limited to the reported dataset, experimental setting and validation in this study; broader generalization and real-world benefit are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcm/jcm-13-05747/article_deploy/jcm-13-05747.pdf

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Deep Learning Approach for the Identification of the Molecular Subtypes of Pancreatic Ductal Adenocarcinoma Based on Whole Slide Pathology Images.

Source check: Source checked. Not independent scientific confirmation.

Finding: Deep-learning models classify PurIST molecular subtypes from H&E whole-slide and biopsy images. TCGA patient-level concordance 83.03%±6.35%, AUC .867; external frozen needle biopsy cohort reports PPV 1.0 and NPV .85. This supports feasibility, not replacement of molecular testing or clinical benefit. TCGA training/validation plus separate POG-PANGEN external biopsy cohort; small retrospective translational sample. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Deep-learning models classify PurIST molecular subtypes from H&E whole-slide and biopsy images.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep-learning models classify PurIST molecular subtypes from H&E whole-slide and biopsy images.
Test and result
TCGA patient-level concordance 83.03%±6.35%, AUC .867; external frozen needle biopsy cohort reports PPV 1.0 and NPV .85. This supports feasibility, not replacement of molecular testing or clinical benefit.
Limitation / still unknown
TCGA training/validation plus separate POG-PANGEN external biopsy cohort; small retrospective translational sample.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12179518/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Effectiveness of Machine Learning in Predicting Orthodontic Tooth Extractions: A Multi-Institutional Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Random forest prediction of orthodontic extraction versus non-extraction across two universities, with cross-institution testing, is a clinically specific retrospective assessment with a clear learned model.

Technology: Machine-learning analysis applies the stated model to the paper’s computer science dataset and evaluates the reported outcome.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis applies the stated model to the paper’s computer science dataset and evaluates the reported outcome.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/bioengineering/bioengineering-11-00888/article_deploy/bioengineering-11-00888.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and external validation of temporal fusion transformer models for continuous intraoperative blood pressure forecasting.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A temporal fusion transformer forecast continuous intraoperative blood pressure and was evaluated on separate patient/database records; sparse measurements, forward filling and treatment-response confounding limit interpretation, and no causal medication effect is established.

Technology: A temporal fusion transformer forecasts continuous intraoperative blood pressure from physiological measurements; the study evaluates transfer across separate patient/database records.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A temporal fusion transformer forecasts continuous intraoperative blood pressure from physiological measurements; the study evaluates transfer across separate patient/database records.
Limitation / still unknown
The datasets and sparse measurements with forward-filled values limit transfer to other operating settings; treatment-response confounding means forecasts do not establish causal effects of vasoactive medication.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11402414/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and validation of a machine learning model for clinical wellness visit classification in cats and dogs.

Source check: Source checked. Not independent scientific confirmation.

Finding: gradient boosting trained on 11,105 visits from 655 animals/544 practices; 400 expert-labeled visits were randomly selected from that same database, yielding specificity .94, sensitivity .86, balanced accuracy .90. The annotation sample is not independent of training; source itself calls for prospective validation. Specific veterinary visit classification result, but same-database expert sample limits independent validation; do not claim subclinical disease detection. Testing: Methods/Results: gradient boosting trained on 11,105 visits from 655 animals/544 practices; 400 expert-labeled visits were randomly selected from that same database, yielding specificity .94, sensitivity .86, balanced accuracy .90. The annotation sample is not independent of training; source itself calls for prospective validation.. Methods/Results: gradient boosting trained on 11,105 visits from 655 animals/544 practices; 400 expert-labeled visits were randomly selected from that same database, yielding specificity .94, sensitivity .86, balanced accuracy .90. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Specific veterinary visit classification result, but same-database expert sample limits independent validation; do not claim subclinical disease detection.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Specific veterinary visit classification result, but same-database expert sample limits independent validation; do not claim subclinical disease detection.
Test and result
gradient boosting trained on 11,105 visits from 655 animals/544 practices; 400 expert-labeled visits were randomly selected from that same database, yielding specificity .94, sensitivity .86, balanced accuracy .90. The annotation sample is not independent of training; source itself calls for prospective validation.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2024.1348162/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Sorghum Yield in Data-Scarce and Conflict-Affected South Sudan Using Machine Learning Techniques.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning models forecast sorghum yields in South Sudan under data-scarce and conflict-affected conditions, a specific agricultural planning result. Climate/agricultural observations from two states are used to train and compare decision trees, ANN, random forest, SVM and XGBoost, then predict 2021 yields. Testing: Forecasts are model-derived and remain geographically/time limited; no controlled field intervention is tested.. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Climate/agricultural observations from two states are used to train and compare decision trees, ANN, random forest, SVM and XGBoost, then predict 2021 yields.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Climate/agricultural observations from two states are used to train and compare decision trees, ANN, random forest, SVM and XGBoost, then predict 2021 yields.
Test and result
Forecasts are model-derived and remain geographically/time limited; no controlled field intervention is tested.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-4837792/v1.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

High-Precision Skin Disease Diagnosis through Deep Learning on Dermoscopic Images.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A CNN classified three dermoscopic skin-disease labels on held-out ISIC images with 87.64% test accuracy. This is dataset-based diagnostic validation only; the authors call for real-world clinical evaluation.

Technology: Publisher full text checked: Methods describe the image cohort and held-out split; Results report test performance and future real-world validation limitation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe the image cohort and held-out split; Results report test performance and future real-world validation limitation.
Limitation / still unknown
The result is limited to held-out ISIC images; the authors call for real-world clinical evaluation, so clinical utility is unknown.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2306-5354/11/9/867

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Genetic association and machine learning improve the prediction of type 1 diabetes risk.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary article combines genetic association/fine-mapping with a machine-learning risk model and reports specific T1D risk signals, nonlinear interactions and patient subclusters; performance and subgroup claims are cohort-specific. The result-bearing preprint preceded the 2026 journal version.

Technology: The primary article combines genetic association/fine-mapping with a machine-learning risk model and reports specific T1D risk signals, nonlinear interactions and patient subclusters; performance and subgroup claims are cohort-specific

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary article combines genetic association/fine-mapping with a machine-learning risk model and reports specific T1D risk signals, nonlinear interactions and patient subclusters; performance and subgroup claims are cohort-specific.
Test and result
The primary article combines genetic association/fine-mapping with a machine-learning risk model and reports specific T1D risk signals, nonlinear interactions and patient subclusters; performance and subgroup claims are cohort-specific. Evaluation: The primary article combines genetic association/fine-mapping with a machine-learning risk model and reports specific T1D risk signals, nonlinear interactions and patient subclusters; performance and subgroup claims are cohort-specific.
Limitation / still unknown
Risk performance and subgroup findings are cohort-specific; independent external validation and clinical utility are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41588-026-02578-y.pdf · https://pmc.ncbi.nlm.nih.gov/articles/PMC11312647/ · https://www.medrxiv.org/content/10.1101/2024.07.31.24311310v1 · https://www.nature.com/articles/s41588-026-02578-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Novel Machine-Learning Algorithm to Predict the Early Termination of Nutrition Support Team Follow-Up in Hospitalized Adults: A Retrospective Cohort Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Retrospective cohort ML algorithm predicting early termination of nutrition support follow-up, with train/test validation; clinically specific retrospective assessment. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Machine-learning analysis reported for A Novel Machine-Learning Algorithm to Predict the Early Termination of Nutrition Support Team Follow-Up in Hospitalized Adults: A Retrospective Cohort Study.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis reported for A Novel Machine-Learning Algorithm to Predict the Early Termination of Nutrition Support Team Follow-Up in Hospitalized Adults: A Retrospective Cohort Study..
Test and result
Retrospective cohort ML algorithm predicting early termination of nutrition support follow-up, with train/test validation; clinically specific retrospective assessment.
Limitation / still unknown
The result is limited to the reported study cohort, data split and comparator; external validation was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/nutrients/nutrients-16-02492/article_deploy/nutrients-16-02492.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning Artificial Intelligence Predicts Homologous Recombination Deficiency and Platinum Response From Histologic Slides.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: DeepHRD used weakly supervised CNNs on H&E slides to predict homologous-recombination deficiency and platinum response across breast/ovarian cancer cohorts, with held-out and external-cohort evaluation.

Technology: whole-slide tiling and weak labels; Results report held-out HRD prediction and external outcome associations.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
whole-slide tiling and weak labels; Results report held-out HRD prediction and external outcome associations.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11469627/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Early Detection of Pulmonary Embolism in a General Patient Population Immediately Upon Hospital Admission Using Machine Learning to Identify New, Unidentified Risk Factors: Model Development Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: A random-forest model used pre-admission records from 2,568 patients with PE and 52,598 controls to predict PE at hospital admission. It achieved an 80% geometric mean of PE and non-PE classification accuracies; clustering identified subgroups with PE prevalence above 61%, including patients with prior PE or pneumonia. These are retrospective associations, not evidence of prospective clinical impact.

Technology: A random-forest model and two imbalance-handling methods used pre-admission demographics, comorbidities and medications to identify PE at hospital admission; clustering characterized high-risk subgroups.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A random-forest model and two imbalance-handling methods used pre-admission demographics, comorbidities and medications to identify PE at hospital admission; clustering characterized high-risk subgroups.
Test and result
The primary abstract reports 2,568 PE cases and 52,598 controls, a random-forest model using pre-admission information, 80% geometric-mean classification accuracy, and risk-enriched clusters; hospital-admission PE prevalence was 4%.
Limitation / still unknown
Retrospective medical-record analysis does not establish prospective early diagnosis, external clinical generalization or improved patient outcomes; clustering identifies associations rather than causal PE risk factors.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/articles/PMC11322683

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based screening and validation of liver metastasis-specific genes in colorectal cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary study uses several ML feature-selection methods to identify colorectal liver-metastasis-associated genes and evaluates the selected genes/model in independent datasets. This is an exploratory association, not evidence that the genes cause metastasis.

Technology: The primary study uses several ML feature-selection methods to identify colorectal liver-metastasis-associated genes and evaluates the selected genes/model in independent datasets. This is an exploratory association, not evidence that the genes cause metastasis.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary study uses several ML feature-selection methods to identify colorectal liver-metastasis-associated genes and evaluates the selected genes/model in independent datasets. This is an exploratory association, not evidence that the genes cause metastasis.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/10.1038/s41598-024-68706-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Toward Machine-Learning-Accelerated Design of All-Dielectric Magnetophotonic Nanostructures

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The full primary article uses neural-network and polynomial-regression models trained on 12,170 simulated magnetophotonic grating configurations to predict transverse magneto-optical Kerr response and resonance. Validation against 3,043 held-out full-wave simulations and a previously unseen grating design supports accelerated design of an all-dielectric biosensing nanostructure; predicted performance remains simulation-based and within the training parameter range.

Technology: Neural network and polynomial regression models map grating geometry and wavelength to simulated TMOKE amplitude/resonance for rapid design.

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Neural network and polynomial regression models map grating geometry and wavelength to simulated TMOKE amplitude/resonance for rapid design.
Limitation / still unknown
The result is limited to the reported dataset, experimental setting and validation in this study; broader generalization and real-world benefit are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pubs.acs.org/aamick/article/16/32/42828/138535

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deciphering KDM8 dysregulation and CpG methylation in hepatocellular carcinoma using multi-omics and machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Multi-omics analysis with machine-learning selection identifies KDM8 dysregulation and CpG methylation patterns associated with hepatocellular carcinoma.

Technology: The study combines HCC expression/methylation datasets and reports candidate KDM8-related methylation and diagnostic/prognostic analyses.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study combines HCC expression/methylation datasets and reports candidate KDM8-related methylation and diagnostic/prognostic analyses.
Limitation / still unknown
The multi-omics associations do not establish that KDM8 dysregulation or methylation causes hepatocellular carcinoma or has clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11370911/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Intercomparison of deep learning models in predicting streamflow patterns: insight from CMIP6.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary hydrology study compares deep-learning models against observed discharge in a specified basin and uses the best-performing model to project streamflow under CMIP6 scenarios. The projection is scenario-dependent and unverified against future observations. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: The primary hydrology study compares deep-learning models against observed discharge in a specified basin and uses the best-performing model to project streamflow under CMIP6 scenarios. The projection is scenario-dependent and unverified against future observations.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary hydrology study compares deep-learning models against observed discharge in a specified basin and uses the best-performing model to project streamflow under CMIP6 scenarios. The projection is scenario-dependent and unverified against future observations.
Test and result
The primary hydrology study compares deep-learning models against observed discharge in a specified basin and uses the best-performing model to project streamflow under CMIP6 scenarios. The projection is scenario-dependent and unverified against future observations.
Limitation / still unknown
The primary hydrology study compares deep-learning models against observed discharge in a specified basin and uses the best-performing model to project streamflow under CMIP6 scenarios. The projection is scenario-dependent and unverified against future observations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/10.1038/s41598-024-63989-7

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting inferior vena cava filter complications using machine learning

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary Methods/Results use 14,476 patients in VQI; 70/30 train/test split predicts one-year IVC-filter complications. XGBoost AUROC 0.93 versus logistic regression 0.63. Retrospective registry validation and dataset-specific risk prediction only. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: XGBoost and logistic regression

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Primary Methods/Results use 14,476 patients in VQI; 70/30 train/test split predicts one-year IVC-filter complications. XGBoost AUROC 0.93 versus logistic regression 0.63. Retrospective registry validation and dataset-specific risk prediction only.
Test and result
Primary Methods/Results use 14,476 patients in VQI; 70/30 train/test split predicts one-year IVC-filter complications. XGBoost AUROC 0.93 versus logistic regression 0.63. Retrospective registry validation and dataset-specific risk prediction only.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11523346/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Clinical utility of a rapid two-dimensional balanced steady-state free precession sequence with deep learning reconstruction

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The checked primary abstract reports this specific result: Unverified Haiku assessment. Prospective clinical evaluation of accelerated cine MRI with deep learning reconstruction in 108 participants with reader ratings and volumetry agreement; the DL reconstruction is the enabling AI and the clinical utility evaluation is suited. The abstract identifies the model or learned operation and an observed evaluation; the conclusion is limited to the reported study setting. Original PMC full-text Methods and Results were checked; date 2024-07-28 reconciled to the primary record.

Technology: Publisher/PMC Methods establish the learned model and its operation in the specific study.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher/PMC Methods establish the learned model and its operation in the specific study.
Limitation / still unknown
The result is limited to the reported dataset, experimental setting and validation in this study; broader generalization and real-world benefit are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11367510/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Comparison of the Efficacy of Artificial Intelligence-Powered Software in Crown Design: An In Vitro Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: PMC11806298 Methods/Results: 33 real posterior crown scans; standard was an experienced technician’s chairside-adjusted and rescanned crown; AI Automate and Dentbird compared with novice/experienced CAD by time and 3D RMS; AI markedly cut design time, but Dentbird occlusal morphology was less accurate. Catalog title search found no same study. Testing: PMC11806298 Methods/Results: 33 real posterior crown scans; standard was an experienced technician’s chairside-adjusted and rescanned crown; AI Automate and Dentbird compared with novice/experienced CAD by time and 3D RMS; AI markedly cut design time, but Dentbird occlusal morphology was less accurate. Catalog title search found no same study.. In 33 patient-derived crown cases, AI CAD software shortened design time relative to technician CAD; a clinician-adjusted crown served as reference, and one AI product had worse occlusal morphology. This is a specific dental restoration design result, not evidence of patient benefit. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: PMC11806298 Methods/Results: 33 real posterior crown scans; standard was an experienced technician’s chairside-adjusted and rescanned crown; AI Automate and Dentbird compared with novice/experienced CAD by time and 3D RMS; AI markedly cut design time, but Dentbird occlusal morphology was less accurate. Catalog title search found no same study.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PMC11806298 Methods/Results: 33 real posterior crown scans; standard was an experienced technician’s chairside-adjusted and rescanned crown; AI Automate and Dentbird compared with novice/experienced CAD by time and 3D RMS; AI markedly cut design time, but Dentbird occlusal morphology was less accurate. Catalog title search found no same study.
Test and result
PMC11806298 Methods/Results: 33 real posterior crown scans; standard was an experienced technician’s chairside-adjusted and rescanned crown; AI Automate and Dentbird compared with novice/experienced CAD by time and 3D RMS; AI markedly cut design time, but Dentbird occlusal morphology was less accurate. Catalog title search found no same study.
Limitation / still unknown
In 33 patient-derived crown cases, AI CAD software shortened design time relative to technician CAD; a clinician-adjusted crown served as reference, and one AI product had worse occlusal morphology. This is a specific dental restoration design result, not evidence of patient benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11806298/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Expert gaze as a usability indicator of medical AI decision support systems: a preliminary study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Dentists could toggle an AI caries detector while examining bitewings; eye tracking found AI use on roughly one quarter of inspection time and increased attention to interface elements.

Technology: The AI caries-detection support was the study intervention; eye tracking measured expert interaction with its optional output.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The AI caries-detection support was the study intervention; eye tracking measured expert interaction with its optional output.
Limitation / still unknown
This preliminary dentist usability study measured gaze and interface attention, not diagnostic accuracy or patient outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41746-024-01192-8

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development of a deep learning model for cancer diagnosis by inspecting cell-free DNA end-motifs.

Source check: Source checked. Not independent scientific confirmation.

Finding: Across six datasets and an independent lung-cancer test set, EMIT AUROC was compared with a baseline; this remains retrospective biomarker validation without clinical outcome testing. The finding is bounded to the reported study and does not establish causal benefit or generalization.

Technology: EMIT transformer learns cfDNA end-motif representations for cancer discrimination

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
EMIT transformer learns cfDNA end-motif representations for cancer discrimination
Test and result
Across six datasets and an independent lung-cancer test set, EMIT AUROC was compared with a baseline; this remains retrospective biomarker validation without clinical outcome testing.
Limitation / still unknown
External validation, causal effects, and sustained operational impact were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41698-024-00635-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning allows robust classification of visceral fat in women with obesity using common laboratory metrics

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary clinical study tests whether learned classifiers can group women with obesity by DXA-defined visceral-fat categories using common laboratory metrics. It reports limited, cohort-specific discrimination and does not establish a replacement for direct measurement. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Logistic regression and six supervised classifiers

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary clinical study tests whether learned classifiers can group women with obesity by DXA-defined visceral-fat categories using common laboratory metrics. It reports limited, cohort-specific discrimination and does not establish a replacement for direct measurement.
Test and result
The primary clinical study tests whether learned classifiers can group women with obesity by DXA-defined visceral-fat categories using common laboratory metrics. It reports limited, cohort-specific discrimination and does not establish a replacement for direct measurement.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/10.1038/s41598-024-68269-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and interpretation of a multimodal predictive model for prognosis of gastrointestinal stromal tumor.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The source reports a bounded machine-learning result using the named model or method on the stated study data; interpretation is limited to that evaluation. The record is limited to the reported cohort, dataset or experiment and does not establish wider clinical or operational benefit.

Technology: Machine-learning analysis applies the stated model to the paper’s medicine dataset and evaluates the reported outcome.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis applies the stated model to the paper’s medicine dataset and evaluates the reported outcome.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41698-024-00636-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction of Major Adverse Cardiac Events After Transcatheter Aortic Valve Implantation: A Machine Learning Approach with GRACE Score.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Machine learning with GRACE score predicts major adverse cardiac events after transcatheter aortic valve implantation.

Technology: TAVI patients are evaluated with a clinical risk model and comparison to GRACE-based prediction.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
TAVI patients are evaluated with a clinical risk model and comparison to GRACE-based prediction.
Limitation / still unknown
The model reports post-TAVI event prediction; prospective external performance and changes in patient outcomes were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11249994/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Improvements of 177 Lu SPECT images from sparsely acquired projections by reconstruction with deep-learning-generated synthetic projections.

Source check: Source checked. Not independent scientific confirmation.

Finding: Deep-learning-generated synthetic projections reconstructed sparse 177Lu SPECT acquisitions; 16 sequential patients were used to assess image quality and dosimetric agreement against full projections. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Methods describe projection generation and test cohort; Results compare reconstructions and quantitative imaging/dose measures.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe projection generation and test cohort; Results compare reconstructions and quantitative imaging/dose measures.
Test and result
Methods describe projection generation and test cohort; Results compare reconstructions and quantitative imaging/dose measures.
Limitation / still unknown
Methods describe projection generation and test cohort; Results compare reconstructions and quantitative imaging/dose measures.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11213840/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Fluorescence excitation-scanning hyperspectral imaging with scalable 2D-3D deep learning framework for colorectal cancer detection.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary study applies deep learning to colorectal-cancer hyperspectral images and reports an independent new-patient test with high classification performance. This is a specific diagnostic-validation result on a small research dataset, not a deployed screening system.

Technology: A scalable two-dimensional/three-dimensional deep-learning framework classified colorectal cancer from fluorescence hyperspectral images.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A scalable two-dimensional/three-dimensional deep-learning framework classified colorectal cancer from fluorescence hyperspectral images.
Limitation / still unknown
The independent new-patient test was conducted on a small research dataset; prospective screening performance and clinical deployment benefit were not assessed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/10.1038/s41598-024-64917-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and validation of a machine learning predictive model for perioperative myocardial injury in cardiac surgery with cardiopulmonary bypass.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Eleven ML approaches predict perioperative myocardial injury after bypass surgery, with external validation across three centers; this is a specific retrospective clinical prediction result.

Technology: Machine-learning analysis applies the stated model to the paper’s medicine dataset and evaluates the reported outcome.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis applies the stated model to the paper’s medicine dataset and evaluates the reported outcome.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13019-024-02856-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial intelligence-enhanced electrocardiography derived body mass index as a predictor of future cardiometabolic disease

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary publisher Methods/Results support the AI-enabled result: AI-ECG model predicts BMI from ECGs; the delta-BMI residual is associated with future cardiometabolic disease in BIDMC and UK Biobank cohorts after adjustment for measured BMI. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: AI-ECG model

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
AI-ECG model predicts BMI from ECGs; the delta-BMI residual is associated with future cardiometabolic disease in BIDMC and UK Biobank cohorts after adjustment for measured BMI.
Test and result
The primary publisher Methods/Results support the AI-enabled result: AI-ECG model predicts BMI from ECGs; the delta-BMI residual is associated with future cardiometabolic disease in BIDMC and UK Biobank cohorts after adjustment for measured BMI.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41746-024-01170-0

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning model for the prediction of all-cause mortality among long term care people in China: a prospective cohort study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary prospective-cohort study uses an ANN survival model on 42,353 disabled long-term-care participants and validates mortality-risk strata on a later cohort subset. The reported risk groups are limited to this Chinese pilot-city population and do not establish care benefit.

Technology: The primary prospective-cohort study uses an ANN survival model on 42,353 disabled long-term-care participants and validates mortality-risk strata on a later cohort subset. The reported risk groups are limited to this Chinese pilot-city population and do not establish care benefit.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary prospective-cohort study uses an ANN survival model on 42,353 disabled long-term-care participants and validates mortality-risk strata on a later cohort subset. The reported risk groups are limited to this Chinese pilot-city population and do not establish care benefit.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/10.1038/s41598-024-65601-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning-Enhanced Extraction of Biomarkers for High-Grade Serous Ovarian Cancer from Proteomics Data

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary article reports ML-derived HGSOC serum protein panels evaluated in independent public cohorts and outperforming standard markers in those datasets; the authors note that larger clinical assay validation remains needed.

Technology: The primary article reports ML-derived HGSOC serum protein panels evaluated in independent public cohorts and outperforming standard markers in those datasets; the authors note that larger clinical assay validation remains needed.

Publication: Publication status unspecified in annotation · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The primary article reports ML-derived HGSOC serum protein panels evaluated in independent public cohorts and outperforming standard markers in those datasets; the authors note that larger clinical assay validation remains needed.
Limitation / still unknown
The result is limited to the reported dataset, experimental setting and validation in this study; broader generalization and real-world benefit are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11199488/

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Socially interactive agents for robotic neurorehabilitation training: conceptualization and proof-of-concept study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The integrated learned-agent system passed an 18-volunteer usability study; this supports feasibility, not rehabilitation efficacy.

Technology: Learned attention, pain and stress classifiers feed the robotic-training interface; the work evaluates the integrated agent system.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned attention, pain and stress classifiers feed the robotic-training interface; the work evaluates the integrated agent system.
Limitation / still unknown
The 18-volunteer study supports usability feasibility only; rehabilitation efficacy was not measured.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/frai.2024.1441955

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Towards Predicting Smoking Events for Just-in-time Interventions.

Source check: Source checked. Not independent scientific confirmation.

Finding: In 28 participants observed in naturalistic settings, ECG/GPS machine-learning models detected and predicted smoking events with modest accuracy; the paper supports a limited proof-of-concept for individualized just-in-time prediction. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Methods describe two-week naturalistic data and LSTM/BiLSTM plus conventional models; Results report held-out/cross-validation detection and prediction accuracy and dependence on participant data quantity/quality.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe two-week naturalistic data and LSTM/BiLSTM plus conventional models; Results report held-out/cross-validation detection and prediction accuracy and dependence on participant data quantity/quality.
Test and result
Methods describe two-week naturalistic data and LSTM/BiLSTM plus conventional models; Results report held-out/cross-validation detection and prediction accuracy and dependence on participant data quantity/quality.
Limitation / still unknown
Methods describe two-week naturalistic data and LSTM/BiLSTM plus conventional models; Results report held-out/cross-validation detection and prediction accuracy and dependence on participant data quantity/quality.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11141818/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning Approach for Pitting Corrosion Detection in Gas Pipelines

Source check: Source checked. Not independent scientific confirmation.

Finding: A CNN detected pitting corrosion in gas-pipeline imagery using a labeled corrosion-image dataset and optimized model evaluation. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Convolutional neural network

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe CNN design and training; Results report held-out corrosion detection performance.
Test and result
A CNN detected pitting corrosion in gas-pipeline imagery using a labeled corrosion-image dataset and optimized model evaluation.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11175124/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Detection of atmospheric radon concentration anomalies and their potential for earthquake prediction using Random Forest analysis

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Random Forest residuals identified historical radon anomalies before the Kobe and Tohoku-oki earthquakes at two observation sites. This supports a retrospective geophysical association only; it does not validate prospective earthquake prediction.

Technology: Random Forest residuals identified historical radon anomalies before the Kobe and Tohoku-oki earthquakes at two observation sites. This supports a retrospective geophysical association only; it does not validate prospective earthquake prediction.

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random Forest residuals identified historical radon anomalies before the Kobe and Tohoku-oki earthquakes at two observation sites. This supports a retrospective geophysical association only; it does not validate prospective earthquake prediction.
Limitation / still unknown
The result is limited to the reported dataset, experimental setting and validation in this study; broader generalization and real-world benefit are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/10.1038/s41598-024-61887-6

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Using explainable machine learning and fitbit data to investigate predictors of adolescent obesity.

Source check: Source checked. Not independent scientific confirmation.

Finding: Explainable boosting machine on ABCD Fitbit subset; cross-validation/fine-tuned AUC and interaction associations reported. Testing: Explainable boosting machine on ABCD Fitbit subset; cross-validation/fine-tuned AUC and interaction associations reported.. An explainable boosting machine learns associations between ABCD Fitbit-derived behavior/heart-rate features and adolescent obesity; cross-validation and a fine-tuned model achieved AUC ≥0.726 and ≥0.737, respectively, and identified interactions. This is cross-sectional prediction/association, not causal or clinical evidence. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: Explainable boosting machine on ABCD Fitbit subset; cross-validation/fine-tuned AUC and interaction associations reported.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Explainable boosting machine on ABCD Fitbit subset; cross-validation/fine-tuned AUC and interaction associations reported.
Test and result
Explainable boosting machine on ABCD Fitbit subset; cross-validation/fine-tuned AUC and interaction associations reported.
Limitation / still unknown
An explainable boosting machine learns associations between ABCD Fitbit-derived behavior/heart-rate features and adolescent obesity; cross-validation and a fine-tuned model achieved AUC ≥0.726 and ≥0.737, respectively, and identified interactions. This is cross-sectional prediction/association, not causal or clinical evidence.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-024-60811-2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Impact of Various Non-Contrast-Enhanced MRA Techniques on Lumen Visibility in Vascular Flow Models with a Surpass Evolve Flow Diverter

Source check: Source checked. Not independent scientific confirmation.

Finding: In a silicone flow-loop phantom with two flow-diverter stents, deep-learning reconstruction of silent MRA produced higher in-stent/background and in-stent/vessel signal ratios than TOF imaging; severe stent-tip artifacts seen on TOF were absent on silent MRA. The study did not involve patients.

Technology: GE deep-learning denoising and deranging reconstruction applied to silent MRA images; it increased in-stent-to-background and in-stent-to-vessel signal ratios compared with TOF images.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
GE deep-learning denoising and deranging reconstruction applied to silent MRA images; it increased in-stent-to-background and in-stent-to-vessel signal ratios compared with TOF images.
Test and result
Two Surpass Evolve stents were placed in silicone flow-loop models; TOF and silent MRA were compared, with deep-learning reconstruction applied to silent MRA. In-stent/background and in-stent/vessel signal ratios were higher with silent MRA and DL reconstruction than with TOF.
Limitation / still unknown
This is a phantom comparison of two stents and tested configurations; no patient imaging, clinical patency outcome or prospective clinical performance was evaluated.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/articles/PMC11171646

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Enhancing the potential of phenomic and genomic prediction in winter wheat breeding using high-throughput phenotyping and deep learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: Deep neural phenomic/genomic prediction was evaluated in multi-location winter-wheat breeding trials with forward prediction across years/locations for grain yield and quality traits; this is a crop-selection prediction result, not discovery of a new cultivar. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Deep neural model combines UAV phenomics and genomic markers to predict breeding traits.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep neural model combines UAV phenomics and genomic markers to predict breeding traits.
Test and result
Deep neural phenomic/genomic prediction was evaluated in multi-location winter-wheat breeding trials with forward prediction across years/locations for grain yield and quality traits; this is a crop-selection prediction result, not discovery of a new cultivar.
Limitation / still unknown
Multi-location/year forward-prediction design and measured grain/yield-quality outcomes are described.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2024.1410249/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial Intelligence-Powered Optimization and Milk Permeate Upcycling for Innovative Sesame Milk with Enhanced Probiotic Viability and Sensory Appeal.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: ANN-guided sesame-milk extraction was checked against an actual total-solids yield, and the study measured that permeate-containing milk supported lactic-acid-bacteria growth and had sensory/nutritional properties.

Technology: Artificial neural networks optimized sesame-milk extraction conditions against total-solids yield, followed by experimental confirmation.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Artificial neural networks optimized sesame-milk extraction conditions against total-solids yield, followed by experimental confirmation.
Limitation / still unknown
The reported optimization and product assays are confined to the tested sesame-milk formulation and conditions; health effects in consumers were not studied.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11170702/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Towards a machine-learning assisted non-invasive classification of dengue severity using wearable PPG data: a prospective clinical study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Random forest, MLP/SVM and time-frequency CNN models classify dengue severity from wearable continuous PPG segments; five-fold cross-validation evaluates models. Prospective Vietnamese clinical dataset: 132 adults recruited, 127 after signal-quality filtering (1,223 hours), and 15 children in secondary acute-vs-follow-up analysis. Models classified clinical severity; best adult RF accuracy about .69 baseline and time-frequency CNN improved metrics by >8%, final accuracy .78; pediatric distinction accuracy .89. This supports physiological signal association and feasibility, not a validated triage device. Prospective observational cohort, but internal 5-fold CV on small single-site Vietnamese population; paper notes homogeneous cohort and need for broader validation. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Random forest, MLP/SVM and time-frequency CNN models classify dengue severity from wearable continuous PPG segments; five-fold cross-validation evaluates models.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random forest, MLP/SVM and time-frequency CNN models classify dengue severity from wearable continuous PPG segments; five-fold cross-validation evaluates models.
Test and result
Prospective Vietnamese clinical dataset: 132 adults recruited, 127 after signal-quality filtering (1,223 hours), and 15 children in secondary acute-vs-follow-up analysis. Models classified clinical severity; best adult RF accuracy about .69 baseline and time-frequency CNN improved metrics by >8%, final accuracy .78; pediatric distinction accuracy .89. This supports physiological signal association and feasibility, not a validated triage device.
Limitation / still unknown
Prospective observational cohort, but internal 5-fold CV on small single-site Vietnamese population; paper notes homogeneous cohort and need for broader validation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11167237/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Vision-language foundation model for echocardiogram interpretation.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary study trains EchoCLIP on over one million echocardiogram/text pairs and reports internal and external clinical interpretation results, including cardiac-function and device identification; it is validation evidence, not proof of clinical benefit.

Technology: The primary study trains EchoCLIP on over one million echocardiogram/text pairs and reports internal and external clinical interpretation results, including cardiac-function and device identification; it is validation evidence, not proof of clinical benefit.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary study trains EchoCLIP on over one million echocardiogram/text pairs and reports internal and external clinical interpretation results, including cardiac-function and device identification; it is validation evidence, not proof of clinical benefit.
Limitation / still unknown
The primary study trains EchoCLIP on over one million echocardiogram/text pairs and reports internal and external clinical interpretation results, including cardiac-function and device identification; it is validation evidence, not proof of clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41591-024-02959-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

AI-algorithm training and validation for identification of endometrial CD138+ cells in infertility-associated conditions; polycystic ovary syndrome (PCOS) and recurrent implantation failure (RIF).

Source check: Source checked. Not independent scientific confirmation.

Finding: The pipeline was compared with three pathologists on 193 samples across control, PCOS and RIF groups; cycle-phase cell-percentage findings remain limited to this cohort. The finding is bounded to the reported study and does not establish causal benefit or generalization.

Technology: Two-CNN pipeline detects endometrial CD138-positive cells in histology images

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Two-CNN pipeline detects endometrial CD138-positive cells in histology images
Test and result
The pipeline was compared with three pathologists on 193 samples across control, PCOS and RIF groups; cycle-phase cell-percentage findings remain limited to this cohort.
Limitation / still unknown
External validation, causal effects, and sustained operational impact were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11140811/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning-driven imaging of cell division and cell growth across an entire eukaryotic life cycle

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Preprint v1 Methods identify Cellpose CNN segmentation and RIFE deep frame interpolation; Results validate tracking against manual annotations (n=680) and report cell growth/cell-cycle exit findings.

Technology: Preprint v1 Methods identify Cellpose CNN segmentation and RIFE deep frame interpolation; Results validate tracking against manual annotations (n=680) and report cell growth/cell-cycle exit findings.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Preprint v1 Methods identify Cellpose CNN segmentation and RIFE deep frame interpolation; Results validate tracking against manual annotations (n=680) and report cell growth/cell-cycle exit findings.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11071524/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prospective and External Validation of Prognostic Machine Learning Models for Short- and Long-Term Mortality Among Acutely Admitted Patients Based on Blood Tests.

Source check: Source checked. Not independent scientific confirmation.

Finding: Models using routine admission biomarkers predict short- and long-term mortality across three cohorts from two Danish hospitals; the preprint reports a specific clinical prediction result, not evidence of prospective impact on care.

Technology: Machine-learning prognostic models using routine admission biomarkers

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Uses routine admission biomarkers to predict short- and long-term mortality across three hospital cohorts.
Test and result
The preprint reports prediction of short- and long-term mortality from admission biomarkers across three cohorts drawn from two Danish hospitals.
Limitation / still unknown
the preprint reports a specific clinical prediction result, not evidence of prospective impact on care.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-4277483/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and validation of a smartwatch algorithm for differentiating physical activity intensity in health monitoring.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary study trained SVM/regression models on smartwatch signals and measured energy expenditure, then evaluated activity-intensity estimates in adult and child cohorts against indirect calorimetry. This is a specific measurement-validation result, not evidence of health benefit.

Technology: Support-vector and regression models mapped smartwatch signals to activity intensity and energy expenditure, evaluated against indirect calorimetry.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Support-vector and regression models mapped smartwatch signals to activity intensity and energy expenditure, evaluated against indirect calorimetry.
Limitation / still unknown
Agreement with indirect calorimetry supports activity-intensity estimates in the studied adult and child cohorts, not health benefit or performance across other devices/populations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/10.1038/s41598-024-59602-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting response to non-selective beta-blockers with liver-spleen stiffness and heart rate in patients with liver cirrhosis and high-risk varices

Source check: Source checked. Not independent scientific confirmation.

Finding: A prospective study used LASSO logistic regression to select three-month percentage changes in spleen stiffness, liver stiffness and heart rate for predicting 12-month NSBB response in cirrhosis with high-risk varices. The combined model reached AUROC 0.96 in derivation and 0.91 in a separate geographic validation cohort. This supports a bounded prediction result, not proof of improved care or HVPG-confirmed response. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: LASSO

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Tenfold-CV LASSO logistic regression selects three-month percentage reductions in spleen stiffness, liver stiffness and heart rate; a combined logistic model estimates clinical NSBB response probability.
Test and result
A prospective study used LASSO logistic regression to select three-month percentage changes in spleen stiffness, liver stiffness and heart rate for predicting 12-month NSBB response in cirrhosis with high-risk varices. The combined model reached AUROC 0.96 in derivation and 0.91 in a separate geographic validation cohort. This supports a bounded prediction result, not proof of improved care or HVPG-confirmed response.
Limitation / still unknown
No prospective clinical benefit or broader external validity is established by the documented study. Organization affiliation was not available in retrieved Crossref metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12003444/ · https://pubmed.ncbi.nlm.nih.gov/38664292/ · https://link.springer.com/article/10.1007/s12072-024-10649-7

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Body weight prediction of Belgian Blue crossbred using random forest.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The March 2024 primary article uses random forest on 26 Belgian Blue × Friesian Holstein crossbreds measured at 11 ages. It reports chest girth as the most important body-weight predictor (24.49%) and random-forest R²=0.86 versus stepwise-regression R²=0.83. This is a small, single-cohort prediction study with no independent validation.

Technology: Random forest analyzes morphometric measurements to estimate cattle body weight and rank predictors.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Random forest analyzes morphometric measurements to estimate cattle body weight and rank predictors.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.ejmanager.com/mnstemps/39/39-1691030977.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep-learning enabled online mass spectrometry of the reaction product from a single catalyst nanoparticle

Source check: Source checked. Not independent scientific confirmation.

Finding: A constrained denoising autoencoder enables online mass-spectrometric detection of CO-oxidation products from a single Pd nanoparticle, experimentally reducing the required catalyst area by roughly three orders of magnitude.

Technology: A constrained denoising autoencoder processes online mass-spectrometry signals from CO oxidation on a single palladium nanoparticle.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A constrained denoising autoencoder processes online mass-spectrometry signals from CO oxidation on a single palladium nanoparticle.
Test and result
The preprint reports detection of reaction products from a single catalyst nanoparticle while reducing the required catalyst area by roughly three orders of magnitude.
Limitation / still unknown
The result is demonstrated in the reported single-particle experimental setup; broader catalyst systems and independent reproduction are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-3797128/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

An advanced nomogram model using deep learning radiomics and clinical data for predicting occult lymph node metastasis in lung adenocarcinoma.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The full text confirms a two-hospital cohort of 473 stage-IA lung adenocarcinoma cases, a fixed 404/69 train/test split, ResNet50 deep-transfer features and logistic models; the test-set DTL-radiomics and nomogram AUCs were 0.885 and 0.898.

Technology: ResNet50 transfer learning extracted CT radiomic features that logistic models combined with clinical variables to predict occult lymph-node metastasis.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ResNet50 transfer learning extracted CT radiomic features that logistic models combined with clinical variables to predict occult lymph-node metastasis.
Limitation / still unknown
The fixed 404/69 split comes from two hospitals and stage-IA cases; prospective utility or benefit from changing surgical management was not tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10998193/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predictors of flares in recent-onset psoriatic arthritis. Results of a multivariable model based on machine learning

Source check: Source checked. Not independent scientific confirmation.

Finding: A prospective multicenter PsA cohort and cross-validated ML models identify patient/disease factors associated with subsequent flares; this is a specific clinical prediction result, while its associations remain observational.

Technology: Cross-validated machine-learning models for psoriatic-arthritis flare factors

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Uses baseline clinical features in a prospective multicenter cohort to identify factors associated with subsequent PsA flares.
Test and result
Cross-validated models in a prospective multicenter PsA cohort identified baseline factors associated with later flares; the reported associations are observational.
Limitation / still unknown
The reported evaluation does not establish prospective benefit or performance beyond the studied setting.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-4165437/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Validation of machine learning models for estimation of left ventricular ejection fraction on point-of-care ultrasound: insights on features that impact performance.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine learning estimated left-ventricular ejection fraction from point-of-care ultrasound videos and was compared with five echocardiographic/expert reference standards; source notes many scans were not analyzable. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Methods describe prospectively acquired POCUS data, model architecture and five reference standards; Results report agreement and performance limits by image quality.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe prospectively acquired POCUS data, model architecture and five reference standards; Results report agreement and performance limits by image quality.
Test and result
Methods describe prospectively acquired POCUS data, model architecture and five reference standards; Results report agreement and performance limits by image quality.
Limitation / still unknown
Methods describe prospectively acquired POCUS data, model architecture and five reference standards; Results report agreement and performance limits by image quality.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10976698/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning in risk prediction of continuous renal replacement therapy after coronary artery bypass grafting surgery in patients.

Source check: Source checked. Not independent scientific confirmation.

Finding: Full Springer PDF: 190 CABG patients, 72 required CRRT; seven models with 90/10 random split and 10-fold cross-validation. Gaussian Naive Bayes selected for highest AUC; top predictors included troponin T, CK-MB, albumin, LDL, NYHA, creatinine and age. Single-center internal model only. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit.

Technology: Full Springer PDF: 190 CABG patients, 72 required CRRT; seven models with 90/10 random split and 10-fold cross-validation. Gaussian Naive Bayes selected for highest AUC; top predictors included troponin T, CK-MB, albumin, LDL, NYHA, creatinine and age. Single-center internal model only.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full Springer PDF: 190 CABG patients, 72 required CRRT; seven models with 90/10 random split and 10-fold cross-validation. Gaussian Naive Bayes selected for highest AUC; top predictors included troponin T, CK-MB, albumin, LDL, NYHA, creatinine and age. Single-center internal model only
Test and result
Full Springer PDF: 190 CABG patients, 72 required CRRT; seven models with 90/10 random split and 10-fold cross-validation. Gaussian Naive Bayes selected for highest AUC; top predictors included troponin T, CK-MB, albumin, LDL, NYHA, creatinine and age. Single-center internal model only
Limitation / still unknown
Full Springer PDF: 190 CABG patients, 72 required CRRT; seven models with 90/10 random split and 10-fold cross-validation. Gaussian Naive Bayes selected for highest AUC; top predictors included troponin T, CK-MB, albumin, LDL, NYHA, creatinine and age. Single-center internal model only
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s10157-024-02472-z · https://link.springer.com/content/pdf/10.1007/s10157-024-02472-z.pdf · https://api.crossref.org/works/10.1007%2Fs10157-024-02472-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Strength of spatial correlation between gray matter connectivity and patterns of proto-oncogene and neural network construction gene expression is associated with diffuse glioma survival.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Imaging transcriptomics linked gray-matter connectivity and gene-expression patterns with diffuse-glioma survival in a 669-patient cohort.

Technology: correlate resting-state gray-matter connectivity maps with gene-expression patterns; Results associate those patterns with survival in diffuse glioma.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
correlate resting-state gray-matter connectivity maps with gene-expression patterns; Results associate those patterns with survival in diffuse glioma.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11004301/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Application of machine learning for identification of heterotic groups in sunflower through combined approach of phenotyping, genotyping and protein profiling.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary study applies unsupervised ML to combined phenotype, marker-genotype and protein profiles from 109 sunflower genotypes to identify heterotic groups. It supports a specific breeding-analysis result, but does not show that the groupings improved hybrid yield. Testing: The primary study applies unsupervised ML to combined phenotype, marker-genotype and protein profiles from 109 sunflower genotypes to identify heterotic groups. It supports a specific breeding-analysis result, but does not show that the groupings improved hybrid yield.. Primary full-text PDF/PMC report identifies three clustering methods applied to 109 sunflower genotypes, forms 12 heterotic groups and evaluates selected crosses; the evidence supports a breeding-analysis result, not proven yield gains. Source affiliation was not stated in the reviewed case checks and was unavailable in the local bibliographic cache.

Technology: The primary study applies unsupervised ML to combined phenotype, marker-genotype and protein profiles from 109 sunflower genotypes to identify heterotic groups. It supports a specific breeding-analysis result, but does not show that the groupings improved hybrid yield.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary study applies unsupervised ML to combined phenotype, marker-genotype and protein profiles from 109 sunflower genotypes to identify heterotic groups. It supports a specific breeding-analysis result, but does not show that the groupings improved hybrid yield.
Test and result
The primary study applies unsupervised ML to combined phenotype, marker-genotype and protein profiles from 109 sunflower genotypes to identify heterotic groups. It supports a specific breeding-analysis result, but does not show that the groupings improved hybrid yield.
Limitation / still unknown
Generalizability beyond the reported study population, data, and protocol was not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/10.1038/s41598-024-58049-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning to Estimate Cardiovascular Risk From Chest Radiographs : A Risk Prediction Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Author institutional copy of the primary paper provides Methods/Results: PLCO development plus two-hospital external outpatient validation; CXR model risk strata associate with subsequent cardiovascular outcomes.

Technology: A deep-learning model mapped chest radiographs to cardiovascular-risk estimates and strata.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A deep-learning model mapped chest radiographs to cardiovascular-risk estimates and strata.
Limitation / still unknown
The PLCO development and two-hospital validation are retrospective prognostic analyses; they do not establish that using the model prevents cardiovascular events.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://cris.maastrichtuniversity.nl/ws/files/194734536/Weiss-2023-Deep-Learning-to-Estimate-Cardiovascular.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Classification of cognitive ability of healthy older individuals using resting-state functional connectivity magnetic resonance imaging and an extreme learning machine.

Source check: Source checked. Not independent scientific confirmation.

Finding: An extreme learning machine on resting-state functional-connectivity features classified excellent versus poor cognitive-test performance in healthy older adults with mean accuracy 86.67%, relative to conventional correlation analysis. This is retrospective cohort classification and does not establish early disease diagnosis. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Extreme learning machine uses resting-state functional-connectivity scores to classify cognitive-test group.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Extreme learning machine uses resting-state functional-connectivity scores to classify cognitive-test group.
Test and result
An extreme learning machine on resting-state functional-connectivity features classified excellent versus poor cognitive-test performance in healthy older adults with mean accuracy 86.67%, relative to conventional correlation analysis. This is retrospective cohort classification and does not establish early disease diagnosis.
Limitation / still unknown
98 older adults plus 90 youth reference dataset; classification accuracy compared to conventional correlation analysis.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12880-024-01250-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Sequential Contrastive and Deep Learning Models to Identify Selective Butyrylcholinesterase Inhibitors

Source check: Source checked. Not independent scientific confirmation.

Finding: PMC full Methods/Results: contrastive/deep-learning virtual screen of 5.37m compounds; 7 of 20 tested compounds are selective BChE inhibitors. Two assay replicates; in-vitro hit evidence only, no in-vivo/clinical result. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Sequential contrastive and deep-learning models

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
PMC full Methods/Results: contrastive/deep-learning virtual screen of 5.37m compounds; 7 of 20 tested compounds are selective BChE inhibitors. Two assay replicates; in-vitro hit evidence only, no in-vivo/clinical result.
Test and result
PMC full Methods/Results: contrastive/deep-learning virtual screen of 5.37m compounds; 7 of 20 tested compounds are selective BChE inhibitors. Two assay replicates; in-vitro hit evidence only, no in-vivo/clinical result.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11331448/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Genistin: A Novel Estrogen Analogue Targeting ERβ to Alleviate Thrombocytopenia.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Publisher full text/PDF Methods and Results show RF screening over 777 descriptors, followed by cell and mouse experiments on genistin; biological effects and ERβ-related mechanism are supported preclinically.

Technology: ML narrowed candidate search; genistin promoted megakaryocyte/platelet outcomes in vitro and improved radiation thrombocytopenia in mice. This is preclinical evidence, not clinical treatment efficacy.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ML narrowed candidate search; genistin promoted megakaryocyte/platelet outcomes in vitro and improved radiation thrombocytopenia in mice. This is preclinical evidence, not clinical treatment efficacy.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.ijbs.com/v20p2236.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial intelligence (AI)-assisted chest computer tomography (CT) insights: a study on intensive care unit (ICU) admittance trends in 78 coronavirus disease 2019 (COVID-19) patients.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: AI-assisted chest CT analysis linked imaging findings to ICU admission trends in COVID-19 patients.

Technology: AI-assisted chest CT analysis linked imaging findings to ICU admission trends in COVID-19 patients.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
AI-assisted chest CT analysis linked imaging findings to ICU admission trends in COVID-19 patients.
Limitation / still unknown
The analysis concerns 78 hospitalized COVID-19 patients; imaging associations with ICU admission trends do not establish causal factors or prospective triage benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10944742/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting extremely low body weight from 12-lead electrocardiograms using a deep neural network

Source check: Source checked. Not independent scientific confirmation.

Finding: A 1D CNN learns from 12-lead ECG waveforms to identify BMI ≤12.6 kg/m² in a retrospective inpatient cohort; splitting by patient, the test set of 187 ECG/BMI pairs yielded AUC 0.807 (95% CI 0.745–0.869), sensitivity 0.702 and specificity 0.796. Scope is this mostly female single-center cohort. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: One-dimensional convolutional neural network

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CNN processes ECG waveforms; patient-level split; independent test subset reported with AUC, sensitivity and specificity.
Test and result
A 1D CNN learns from 12-lead ECG waveforms to identify BMI ≤12.6 kg/m² in a retrospective inpatient cohort; splitting by patient, the test set of 187 ECG/BMI pairs yielded AUC 0.807 (95% CI 0.745–0.869), sensitivity 0.702 and specificity 0.796. Scope is this mostly female single-center cohort.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-024-55453-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Cell-cell communication network-based interpretable machine learning predicts cancer patient response to immune checkpoint inhibitors.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Interpretable machine learning on patient cell-cell communication networks predicted immune-checkpoint-inhibitor response across independent bulk and single-cell cohorts.

Technology: Materials and Methods describe network construction and cohort analyses; Results report response prediction across nine bulk and two single-cell cohorts.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Materials and Methods describe network construction and cohort analyses; Results report response prediction across nine bulk and two single-cell cohorts.
Limitation / still unknown
Independent bulk and single-cell cohorts support predictive associations, not causal cell-cell mechanisms or prospective treatment-selection benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pmc&id=PMC10830106

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Structure of the space of folding protein sequences defined by large language models.

Source check: Source checked. Not independent scientific confirmation.

Finding: The paper uses ESMFold-predicted structures to define an effective foldability energy, then statistical-mechanics sampling to find that natural protein-G sequences occupy a wider low-energy basin than sampled artificial sequences. This is a model-dependent computational result for the studied protein, not experimental confirmation or a general result for all proteins. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: ESMFold structure predictions define the effective foldability energy used in the sequence-space analysis.

Family ESM / ESMFold: profile · models page

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ESMFold structure predictions define the effective foldability energy used in the sequence-space analysis.
Test and result
Natural protein-G sequences occupy wider low-energy basins than the sampled artificial sequences under the paper’s model and sampling.
Limitation / still unknown
Computational sampling, stationarity checks and predicted-structure confidence are reported; this remains a model-dependent result without experimental validation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://air.unimi.it/retrieve/5d331ae7-9b34-47a5-8f1e-f7bd8dd02003/Zambon_2024_Phys._Biol._21_026002.pdf · https://doi.org/10.1088/1478-3975/ad205c

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Comparison between model-based RSA and an AI-based CT-RSA: an accuracy study of 30 patients.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: AI-based CT-RSA measurement agreed with the established migration reference within its precision in 30 hip replacements.

Technology: AI segmentation/landmark measurement compared with model-based RSA in 30 THA patients; agreement and repeatability were assessed against reference precision.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
AI segmentation/landmark measurement compared with model-based RSA in 30 THA patients; agreement and repeatability were assessed against reference precision.
Limitation / still unknown
Agreement was assessed in 30 hip replacements; broader accuracy, longer-term migration outcomes and clinical benefit remain unestablished.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.2340/17453674.2024.35749

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Evaluating a radiotherapy deep learning synthetic CT algorithm for PET-MR attenuation correction in the pelvis.

Source check: Source checked. Not independent scientific confirmation.

Finding: A deep-learning synthetic CT algorithm improved pelvic PET attenuation correction against CT reference images, with near-zero whole-image SUV difference; it did not improve thresholded tumor delineation. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Methods describe synthetic CT generation and PET data analysis; Results report SUV and tumor-volume comparisons against CT attenuation correction.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe synthetic CT generation and PET data analysis; Results report SUV and tumor-volume comparisons against CT attenuation correction.
Test and result
Methods describe synthetic CT generation and PET data analysis; Results report SUV and tumor-volume comparisons against CT attenuation correction.
Limitation / still unknown
Methods describe synthetic CT generation and PET data analysis; Results report SUV and tumor-volume comparisons against CT attenuation correction.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11266329/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and prognostic validation of a three-level NHG-like deep learning-based model for histological grading of breast cancer

Source check: Source checked. Not independent scientific confirmation.

Finding: Learned WSI grader yields a clinically specific prognostic stratification and is assessed on internal and fully independent external cohort; satisfies validation for defined result. Do not infer improved clinical outcomes. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Deep-learning histology-image grader

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
routine histology WSIs from two cohorts (SöS-BC-4 N=2,421 training/internal test; SCAN-B-Lund N=1,262 external test); DL assigns three-level NHG-like grade. Prognostic recurrence-free survival analysis: grade-3 vs grade-1 HR 2.27 (95% CI 1.07–4.82, p=.033) internal after adjustment; independent cohort assessed. Specific prognostic grading result, not prospective benefit.
Test and result
Learned WSI grader yields a clinically specific prognostic stratification and is assessed on internal and fully independent external cohort; satisfies validation for defined result. Do not infer improved clinical outcomes.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13058-024-01770-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

An Automatic Framework for Nasal Esthetic Assessment by ResNet Convolutional Neural Network.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full Springer PDF: automated nasal landmark/symmetry framework was compared against otolaryngologist ratings; source reports close association and visualization of asymmetry, but no independent validation or clinical-outcome benefit. Bounded measurement result. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit.

Technology: Full Springer PDF: automated nasal landmark/symmetry framework was compared against otolaryngologist ratings; source reports close association and visualization of asymmetry, but no independent validation or clinical-outcome benefit. Bounded measurement result.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Full Springer PDF: automated nasal landmark/symmetry framework was compared against otolaryngologist ratings; source reports close association and visualization of asymmetry, but no independent validation or clinical-outcome benefit. Bounded measurement result.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s10278-024-00973-7 · https://link.springer.com/content/pdf/10.1007/s10278-024-00973-7.pdf · https://api.crossref.org/works/10.1007%2Fs10278-024-00973-7

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Material Design of Porous Hydroxyapatite Ceramics via Inverse Analysis of an Estimation Model for Bone-Forming Ability Based on Machine Learning and Experimental Validation of Biological Hard Tissue Responses.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full MDPI text reviewed in Chrome: two machine-learning models estimate bioceramic properties from fabrication settings and bone-formation rate from material properties/in-vivo conditions. Inverse design candidates were fabricated, and a 12-week porcine-tibia implantation comparison found measured bone-formation rates within the model error range. Approve as a machine-learning-guided biomaterials result with limited sample/data and animal-model scope.

Technology: GMR Models 1 and 2 learn fabrication-to-material-property and material/in-vivo-condition-to-bone-formation mappings; inverse analysis proposes target-specific material conditions.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
GMR Models 1 and 2 learn fabrication-to-material-property and material/in-vivo-condition-to-bone-formation mappings; inverse analysis proposes target-specific material conditions.
Limitation / still unknown
Full MDPI text reviewed in Chrome: two machine-learning models estimate bioceramic properties from fabrication settings and bone-formation rate from material properties/in-vivo conditions. Inverse design candidates were fabricated, and a 12-week porcine-tibia implantation comparison found measured bone-formation rates within the model error range. Approve as a machine-learning-guided biomaterials result with limited sample/data and animal-model scope.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/1996-1944/17/3/571

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Deep Learning Model with a Self-Attention Mechanism for Leg Joint Angle Estimation across Varied Locomotion Modes.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A self-attention model estimated ankle and knee joint-angle trajectories across locomotion modes using held-out subject testing.

Technology: baseline and attention models and leave-one-subject-out testing; Results report joint-angle error and fine-tuning comparisons.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
baseline and attention models and leave-one-subject-out testing; Results report joint-angle error and fine-tuning comparisons.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10781404/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

MLP-RL-CRD: diagnosis of cardiovascular risk in athletes using a reinforcement learning-based multilayer perceptron

Source check: Source checked. Not independent scientific confirmation.

Finding: The full author-posted article identifies a reinforcement-learning multilayer perceptron with mutual-learning artificial-bee-colony initialization, trained on 26,002 athletes and assessed against expert ECG labels; the case is a retrospective athlete-screening result, limited to the study dataset and not evidence of clinical benefit. Compared with conventional/deep baselines and reports F-measure 87.4%, geometric mean 89.6%; no prospective/external cohort is reported, so scope is limited.

Technology: MLP with reward-based sequential classification and mutual-learning ABC initialization maps 26,002 athletes’ clinical variables to expert ECG risk labels.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
MLP with reward-based sequential classification and mutual-learning ABC initialization maps 26,002 athletes’ clinical variables to expert ECG risk labels.
Test and result
Reports the athlete-cohort risk-classification result; full paper identifies dataset and ECG-derived labels.
Limitation / still unknown
Compared with conventional/deep baselines and reports F-measure 87.4%, geometric mean 89.6%; no prospective/external cohort is reported, so scope is limited. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchgate.net/publication/376419248_MLP-RL-CRD_diagnosis_of_cardiovascular_risk_in_athletes_using_a_reinforcement_learning-based_multilayer_perceptron · https://doi.org/10.1088/1361-6579/ad1459

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Phenotyping Refractory Cardiogenic Shock Patients Receiving Venous-arterial Extracorporeal Membrane Oxygenation with Machine Learning Algorithms

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Clustering of a prospective 210-patient VA-ECMO cohort identified three cardiogenic-shock phenotypes with distinct in-hospital mortality; this exploratory result has single-center scope.

Technology: Unsupervised clustering of a prospective VA-ECMO cohort identified three cardiogenic-shock phenotypes with different in-hospital mortality.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Unsupervised clustering of a prospective VA-ECMO cohort identified three cardiogenic-shock phenotypes with different in-hospital mortality.
Limitation / still unknown
This is an exploratory single-center cohort result; phenotype stability and predictive value in other VA-ECMO populations were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-3767608/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A novel deep learning model for a computed tomography diagnosis of coronary plaque erosion

Source check: Source checked. Not independent scientific confirmation.

Finding: A momentum-distillation composite transformer learns CTA features to classify plaque erosion against OCT labels; it was evaluated on non-overlapping patient splits (316 development, 79 test patients), with test AUC 0.899 (95% CI 0.841–0.957). PPV was 71.1%, so this is retrospective diagnostic performance, not clinical deployment. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Momentum-distillation transformer

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Transformer-based CTA classifier uses OCT reference; patient-level nonoverlapping test set of 79; test AUC 0.899.
Test and result
A momentum-distillation composite transformer learns CTA features to classify plaque erosion against OCT labels; it was evaluated on non-overlapping patient splits (316 development, 79 test patients), with test AUC 0.899 (95% CI 0.841–0.957). PPV was 71.1%, so this is retrospective diagnostic performance, not clinical deployment.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-023-50483-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Automatic Seizure Detection Based on Stockwell Transform and Transformer.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The S-transform plus Transformer method performs a specific retrospective seizure-detection evaluation on the CHB-MIT EEG database. The source reports segment-level and event-level metrics, including 96.15% accuracy, 96.57% event sensitivity, 0.38 false detections/hour and 20.62 s latency. Event testing uses different seizures from the same patients, so this supports patient-specific database performance, not generalization to unseen patients or clinical benefit.

Technology: 2. Methods; 3. Dataset and experimental results; 5. Conclusions: S-transform time-frequency features feed Transformer channel attention/classification; CHB-MIT event-level testing uses held-out seizures within the same patients.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
2. Methods; 3. Dataset and experimental results; 5. Conclusions: S-transform time-frequency features feed Transformer channel attention/classification; CHB-MIT event-level testing uses held-out seizures within the same patients.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/sensors/sensors-24-00077/article_deploy/sensors-24-00077.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based gait analysis to predict clinical frailty scale in elderly patients with heart failure

Source check: Source checked. Not independent scientific confirmation.

Finding: Smartphone pose estimation plus LightGBM enabled automated frailty scoring in older heart-failure patients, and scores were associated with subsequent mortality. Prospective derivation/validation cohorts and mortality follow-up; observational association only, not intervention benefit.

Technology: smartphone pose estimation extracted gait features and LightGBM predicted clinical frailty score in seven-centre prospective cohort.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
smartphone pose estimation extracted gait features and LightGBM predicted clinical frailty score in seven-centre prospective cohort.
Test and result
derivation n=194 and validation n=223; predicted frailty also associated with later mortality.
Limitation / still unknown
Prospective derivation/validation cohorts and mortality follow-up; observational association only, not intervention benefit. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10944685/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC10944685/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC10944685/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting pressure gradient using artificial intelligence for transcatheter aortic valve replacement.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary article evaluates an AI model to predict post-TAVR pressure gradient using clinical/echocardiographic variables and reports cohort performance against measured gradients. Bounded retrospective clinical prediction; no outcome benefit claim.

Technology: Primary article evaluates an AI model to predict post-TAVR pressure gradient using clinical/echocardiographic variables and reports cohort performance against measured gradients. Bounded retrospective clinical prediction; no outcome benefit claim.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Primary article evaluates an AI model to predict post-TAVR pressure gradient using clinical/echocardiographic variables and reports cohort performance against measured gradients. Bounded retrospective clinical prediction; no outcome benefit claim.
Limitation / still unknown
Primary article evaluates an AI model to predict post-TAVR pressure gradient using clinical/echocardiographic variables and reports cohort performance against measured gradients. Bounded retrospective clinical prediction; no outcome benefit claim.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1016/j.xjtc.2023.11.011 · https://pmc.ncbi.nlm.nih.gov/articles/PMC10859647/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Diagnostic plasma small extracellular vesicles miRNA signatures for pancreatic cancer using machine learning methods.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The full text reports LASSO, random-forest and SVM-RFE selection of plasma sEV-miRNA signatures in 106 participants, with training/validation cohort comparisons and follow-up marker experiments.

Technology: The full text reports LASSO, random-forest and SVM-RFE selection of plasma sEV-miRNA signatures in 106 participants, with training/validation cohort comparisons and follow-up marker experiments.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The full text reports LASSO, random-forest and SVM-RFE selection of plasma sEV-miRNA signatures in 106 participants, with training/validation cohort comparisons and follow-up marker experiments.
Limitation / still unknown
The 106-participant signature and follow-up marker experiments do not establish population screening performance or causal roles for the selected miRNAs.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10730862/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Diagnostic Accuracy and Performance Analysis of a Scanner-Integrated Artificial Intelligence Model for the Detection of Intracranial Hemorrhages in a Traumatology Emergency Department.

Source check: Source checked. Not independent scientific confirmation.

Finding: Retrospective evaluation of a scanner-integrated AI model for intracranial hemorrhage on 435 consecutive CT scans with radiologist reference, giving a clinically specific diagnostic accuracy assessment.

Technology: A scanner-integrated AI model analyzes CT images to detect intracranial hemorrhage, with radiologist readings as reference.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A scanner-integrated AI model analyzes CT images to detect intracranial hemorrhage, with radiologist readings as reference.
Test and result
Retrospective evaluation included 435 consecutive CT scans and measured diagnostic accuracy against radiologist assessment.
Limitation / still unknown
The study is a retrospective, single-department accuracy assessment; external performance and patient outcomes are unknown.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/bioengineering/bioengineering-10-01362/article_deploy/bioengineering-10-01362.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Beta-Lactam Target Non-Attainment in ICU Patients at Treatment Initiation: Development and External Validation of Three Novel (Machine Learning) Models.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Clinical ML models predicting beta-lactam target non-attainment in ICU patients, with external validation across two multicenter cohorts.

Technology: Clinical ML models predicting beta-lactam target non-attainment in ICU patients, with external validation across two multicenter cohorts.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Clinical ML models predicting beta-lactam target non-attainment in ICU patients, with external validation across two multicenter cohorts.
Limitation / still unknown
External cohorts support model validation, but prospective use to guide dosing and improved infection outcomes were not tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/antibiotics/antibiotics-12-01674/article_deploy/antibiotics-12-01674.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting sepsis within 24 hours at triage stage in emergency department using machine learning

Source check: Source checked. Not independent scientific confirmation.

Finding: XGBoost predicts sepsis within 24 hours from triage data in MIMIC-IV, with a held-out test set and cross-validation; this is a specific retrospective clinical prediction study, without external or prospective model validation. This is a preprint, not a peer-reviewed publication.

Technology: XGBoost

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
XGBoost predicts sepsis within 24 hours from triage data in MIMIC-IV, with a held-out test set and cross-validation; this is a specific retrospective clinical prediction study, without external or prospective model validation.
Test and result
XGBoost predicts sepsis within 24 hours from triage data in MIMIC-IV, with a held-out test set and cross-validation; this is a specific retrospective clinical prediction study, without external or prospective model validation.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-3623290/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

ADHD/CD-NET: automated EEG-based characterization of ADHD and CD using explainable deep neural network technique.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary source Methods/Results confirm the represented result: A CNN on EEG wavelet correlation matrices distinguishes ADHD, ADHD+CD and CD, evaluated by 10-fold cross-validation on a private set and an external public dataset; a clinically specific retrospective assessment with Grad-CAM explanations.

Technology: A CNN on EEG wavelet correlation matrices distinguishes ADHD, ADHD+CD and CD, evaluated by 10-fold cross-validation on a private set and an external public dataset; a clinically specific retrospective assessment with Grad-CAM explanations.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A CNN on EEG wavelet correlation matrices distinguishes ADHD, ADHD+CD and CD, evaluated by 10-fold cross-validation on a private set and an external public dataset; a clinically specific retrospective assessment with Grad-CAM explanations.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11571-023-10028-2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction of the Fatigue Strength of Steel Based on Interpretable Machine Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: GBT predicts steel fatigue strength from 437 NIMS records; 5-fold CV and 80/20 holdout. Composition-only R²=.92/MAPE 8.15%; adding heat treatment yields .98/3.14%; atomic features .98/3.05%. Feature-selection/SHAP/SR identify TT, Mo and related inputs.

Technology: GBT predicts steel fatigue strength from 437 NIMS records; 5-fold CV and 80/20 holdout. Composition-only R²=.92/MAPE 8.15%; adding heat treatment yields .98/3.14%; atomic features .98/3.05%. Feature-selection/SHAP/SR identify TT, Mo and related inputs.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
GBT predicts steel fatigue strength from 437 NIMS records; 5-fold CV and 80/20 holdout. Composition-only R²=.92/MAPE 8.15%; adding heat treatment yields .98/3.14%; atomic features .98/3.05%. Feature-selection/SHAP/SR identify TT, Mo and related inputs.
Limitation / still unknown
Retrospective database prediction, not newly run fatigue tests; performance depends on the NIMS sample and split.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3390/ma16237354 · https://www.mdpi.com/1996-1944/16/23/7354

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Endoscopy-based IBD identification by a quantized deep learning pipeline.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Endoscopy images with clinical diagnoses are used for model training/testing and comparison with prior detection/classification performance.

Technology: A quantized deep-learning pipeline identifies inflammatory bowel disease from endoscopic images.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A quantized deep-learning pipeline identifies inflammatory bowel disease from endoscopic images.
Limitation / still unknown
The result is limited to the endoscopy datasets and diagnoses evaluated; prospective performance or improved IBD outcomes were not shown.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10675910/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

An explainable machine learning-based phenomapping strategy for adaptive predictive enrichment in randomized clinical trials

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary publisher Methods/Results support the AI-enabled result: Simulated group-sequential analyses of IRIS and SPRINT trials show ML phenomapping can enrich adaptive enrolment, with trial-specific evaluation. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Machine-learning phenomapping

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Simulated group-sequential analyses of IRIS and SPRINT trials show ML phenomapping can enrich adaptive enrolment, with trial-specific evaluation.
Test and result
The primary publisher Methods/Results support the AI-enabled result: Simulated group-sequential analyses of IRIS and SPRINT trials show ML phenomapping can enrich adaptive enrolment, with trial-specific evaluation.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41746-023-00963-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

"A net for everyone": fully personalized and unsupervised neural networks trained with longitudinal data from a single patient.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A per-patient Wasserstein-GAN trained on two longitudinal MRI scans maps tumor change in glioblastoma; the study reports AUC 0.87 and evaluates two datasets, with a modified RANO classification accuracy of 66%. This is a small, retrospective technical clinical result.

Technology: Patient-specific unsupervised Wasserstein-GAN estimates change between two MRI timepoints.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Patient-specific unsupervised Wasserstein-GAN estimates change between two MRI timepoints.
Limitation / still unknown
This small retrospective technical study used per-patient longitudinal MRI data; generalization to other patients and prospective treatment benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12880-023-01128-w

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Predicts Acute Kidney Injury in Hospitalized Patients with Sickle Cell Disease.

Source check: Source checked. Not independent scientific confirmation.

Finding: XGBoost applied to minute-level bedside vital-sign streams predicted acute kidney injury up to 12 hours before onset in hospitalized sickle-cell patients. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: The classifier uses five minute-sampled physiologic streams from bedside monitors.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The classifier uses five minute-sampled physiologic streams from bedside monitors.
Test and result
In 1,178 filtered encounters across five hospitals, AUROC was 0.91 at 12 hours and 0.82 at 48 hours before AKI.
Limitation / still unknown
Retrospective regional-network evaluation with limited AKI cases; abstract and full paper note generalizability limits and no intervention outcome.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://karger.com/ajn/article-pdf/55/1/18/4163536/000534864.pdf · https://karger.com/ajn/article-pdf/55/1/18/4163536/000534864.pdf#methods · https://karger.com/ajn/article-pdf/55/1/18/4163536/000534864.pdf#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A transformer-based multi-task deep learning model for simultaneous infiltrated brain area identification and segmentation of gliomas.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A multitask transformer identified and segmented infiltrated brain regions in glioma imaging.

Technology: A multitask transformer identified and segmented infiltrated brain regions in glioma imaging.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A multitask transformer identified and segmented infiltrated brain regions in glioma imaging.
Limitation / still unknown
The segmentation result is limited to the evaluated glioma imaging data; prospective workflow or patient-outcome benefit was not tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10612240/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Integration of machine learning to identify diagnostic genes in leukocytes for acute myocardial infarction patients.

Source check: Source checked. Not independent scientific confirmation.

Finding: Integrated ML selected leukocyte diagnostic genes from four AMI datasets; the abstract reports a separate 20 AMI/20 stable-coronary-disease clinical validation, supporting a specific diagnostic result. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Machine-learning analysis reported for Integration of machine learning to identify diagnostic genes in leukocytes for acute myocardial infarction patients.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis reported for Integration of machine learning to identify diagnostic genes in leukocytes for acute myocardial infarction patients..
Test and result
Integrated ML selected leukocyte diagnostic genes from four AMI datasets; the abstract reports a separate 20 AMI/20 stable-coronary-disease clinical validation, supporting a specific diagnostic result.
Limitation / still unknown
The result is limited to the reported study cohort, data split and comparator; external validation was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12967-023-04573-x

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

MS-Net: a novel lightweight and precise model for plant disease identification

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: MS-Net is evaluated on apple-leaf images including outdoor field-condition images and reports disease-identification performance beyond PlantVillage; the result is a specific plant-disease recognition application, with dataset and deployment limits.

Technology: Lightweight MobileNetV3-based architecture classifies plant diseases from leaf images.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Lightweight MobileNetV3-based architecture classifies plant diseases from leaf images.
Limitation / still unknown
Full article reports benchmark plus outdoor-image evaluation and comparisons; field accuracy claim remains dataset-specific. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2023.1276728/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting the Performance of Functional Materials Composed of Polymeric Multicomponent Systems Using Artificial Intelligence-Formulations of Cleansing Foams as an Example

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Machine-learning models predict cleansing performance across 537 experimentally prepared foam formulations; measured eyeliner-removal outcomes support the formulation-performance result, with predictive performance assessed on held-out samples.

Technology: Machine-learning formulation-performance models

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Predicts cleansing performance from composition data across 537 prepared foam formulations and compares predictions with measured eyeliner-removal outcomes.
Limitation / still unknown
The reported result is limited to the study data, samples and conditions; broader application was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/polymers/polymers-15-04216/article_deploy/polymers-15-04216.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A deep-learning pipeline to diagnose pediatric intussusception and assess severity during ultrasound scanning: a multicenter retrospective-prospective study.

Source check: Source checked. Not independent scientific confirmation.

Finding: A deep-learning pipeline trained on 14,085 ultrasound images from 8,736 patients was evaluated on internal images and external videos. It achieved average AUCs of 0.972 for detecting three signs and 0.956 for classifying 184 videos. In a prospective 242-volunteer reader study, AI assistance raised junior sonographers’ average AUC from 0.857 to 0.966 and reduced median scan time from 9.46 to 3.66 minutes, similar to senior sonographers; patient outcomes were not measured.

Technology: A real-time deep-learning pipeline localized ultrasound signs and classified ileocolic intussusception and severity; prospective AI assistance was compared across sonographer experience levels.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
A real-time deep-learning pipeline localized ultrasound signs and classified ileocolic intussusception and severity; prospective AI assistance was compared across sonographer experience levels.
Test and result
Primary abstract reports training on 14,085 images from 8,736 patients, internal image and external video evaluation, plus a prospective 242-volunteer sonographer study. Junior readers’ average AUC rose from 0.857 to 0.966 and median scan time fell from 9.46 to 3.66 minutes with AI assistance.
Limitation / still unknown
The reader study was a prospective pilot; it did not measure changes in treatment decisions, patient outcomes or sustained clinical workflow performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/articles/PMC10541898

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Evaluation of tumour infiltrating lymphocytes in luminal breast cancer using artificial intelligence.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary study applies supervised CNN segmentation/quantification to tumor-infiltrating lymphocytes in luminal breast cancer and reports outcome associations in discovery/test and external cohorts. Preserve the mixed finding: tTIL results were prognostic while AI-derived sTIL proportion was not. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Publisher PDF, original Methods/Results: Nottingham discovery/test cohorts (1,572/659) and external cohort (318); model-derived spatial TIL measures and reported survival associations.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher PDF, original Methods/Results: Nottingham discovery/test cohorts (1,572/659) and external cohort (318); model-derived spatial TIL measures and reported survival associations.
Test and result
Publisher PDF, original Methods/Results: Nottingham discovery/test cohorts (1,572/659) and external cohort (318); model-derived spatial TIL measures and reported survival associations.
Limitation / still unknown
Publisher PDF, original Methods/Results: Nottingham discovery/test cohorts (1,572/659) and external cohort (318); model-derived spatial TIL measures and reported survival associations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41416-023-02451-3.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

In Vitro Antioxidant and In Vivo Antigenotoxic Features of a Series of 61 Essential Oils and Quantitative Composition-Activity Relationships Modeled through Machine Learning Algorithms.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Experimental antioxidant assays on 61 essential oils with ML-derived composition-activity models identifying key constituents; distinct field result with AI as enabling analysis.

Technology: Machine-learning analysis applies the stated model to the paper’s computer science dataset and evaluates the reported outcome.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis applies the stated model to the paper’s computer science dataset and evaluates the reported outcome.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/antioxidants/antioxidants-12-01815/article_deploy/antioxidants-12-01815.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

VIDIIA Hunter diagnostic platform: a low-cost, smartphone connected, artificial intelligence-assisted COVID-19 rapid diagnostics approved for medical use in the UK.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The VH6 system uses automated image acquisition and an embedded deep-learning image classifier to interpret RT-LAMP COVID-19 assays. Against RT-qPCR, final validation of 400 NHS samples yielded 98% sensitivity overall (100% in high viral-load samples) and 100% specificity; clinical and variant tests provide measured diagnostic performance.

Technology: An embedded deep-learning image classifier interpreted RT-LAMP assay images after automated image capture.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
An embedded deep-learning image classifier interpreted RT-LAMP assay images after automated image capture.
Limitation / still unknown
The 400-sample NHS validation supports diagnostic performance against RT-qPCR; sustained operational outcomes or improved patient outcomes were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/molecular-biosciences/articles/10.3389/fmolb.2023.1144001/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction of post-stroke cognitive impairment after acute ischemic stroke using machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: ML models predict 3–6-month post-stroke cognitive impairment in a defined 951-patient ischemic-stroke cohort; the abstract supports a specific retrospective clinical prediction result.

Technology: ML models predict 3–6-month post-stroke cognitive impairment in a defined 951-patient ischemic-stroke cohort; the abstract supports a specific retrospective clinical prediction result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ML models predict 3–6-month post-stroke cognitive impairment in a defined 951-patient ischemic-stroke cohort; the abstract supports a specific retrospective clinical prediction result.
Limitation / still unknown
The abstract describes prediction within a 951-patient cohort; prospective external validation and clinical impact were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13195-023-01289-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of growth years for Puerariae Thomsonii Radix based on hyperspectral imaging technology and deep learning algorithm

Source check: Source checked. Not independent scientific confirmation.

Finding: CNN and VGG16 classify hyperspectral images by Puerariae Thomsonii Radix growth year; the paper reports held-out testing, including outer-surface 1-year-versus-other F1 scores of 93.51% and 92.90%, supporting a narrow medicinal-material quality-control result. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: CNN and VGG16

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CNN/VGG16 learned image classifiers; test set; best F1 reported for growth-year classification and 1-year eligibility task.
Test and result
CNN and VGG16 classify hyperspectral images by Puerariae Thomsonii Radix growth year; the paper reports held-out testing, including outer-surface 1-year-versus-other F1 scores of 93.51% and 92.90%, supporting a narrow medicinal-material quality-control result.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-023-40863-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

An AI-powered patient triage platform for future viral outbreaks using COVID-19 as a disease model

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine learning combining clinical, comorbidity and metabolomics data predicted hospitalization outcomes in SARS-CoV-2 patients, with a defined clinical cohort and measured outcome evaluation.

Technology: Results report prediction of survival/mortality, length of stay and ventilation; Materials and Methods describe study participants, clinical/metabolomics data and machine-learning analyses.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Results report prediction of survival/mortality, length of stay and ventilation; Materials and Methods describe study participants, clinical/metabolomics data and machine-learning analyses.
Test and result
Results report prediction of survival/mortality, length of stay and ventilation; Materials and Methods describe study participants, clinical/metabolomics data and machine-learning analyses.
Limitation / still unknown
Results report prediction of survival/mortality, length of stay and ventilation; Materials and Methods describe study participants, clinical/metabolomics data and machine-learning analyses. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10463861/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The Prognostic Value of ASPHD1 and ZBTB12 in Colorectal Cancer: A Machine Learning-Based Integrated Bioinformatics Approach.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Publisher full text read in Chrome, Methods and Results: transcriptomic ML analysis of colorectal-cancer samples nominated ASPHD1/ZBTB12 prognostic markers; follow-up qPCR in 64 samples and whole-exome validation in 15 samples. Specific gene-marker associations with experimental validation, but small validation subsets and observational evidence do not establish clinical utility.

Technology: Transcriptomic machine-learning analysis ranked colorectal-cancer prognostic markers, which were checked with qPCR and whole-exome data.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Transcriptomic machine-learning analysis ranked colorectal-cancer prognostic markers, which were checked with qPCR and whole-exome data.
Limitation / still unknown
Publisher full text read in Chrome, Methods and Results: transcriptomic ML analysis of colorectal-cancer samples nominated ASPHD1/ZBTB12 prognostic markers; follow-up qPCR in 64 samples and whole-exome validation in 15 samples. Specific gene-marker associations with experimental validation, but small validation subsets and observational evidence do not establish clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2072-6694/15/17/4300

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Ultrasound characteristics of follicular and parafollicular thyroid neoplasms: diagnostic performance of artificial neural network

Source check: Source checked. Not independent scientific confirmation.

Finding: Specific histology-linked sonographic differences and ANN discrimination evaluated against pathological diagnoses constitute a defined diagnostic result. Accuracy varies widely; do not claim established clinical readiness. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Three-layer artificial neural network

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
institutional retrospective cohort, 148 thyroidectomy patients (37 each medullary, papillary, follicular carcinoma and follicular adenoma); seven demographic/morphologic features input to 3-layer ANN; leave-one-out CV plus 75/20/5 split. Contour/structure/calcification differ by pathology; pairwise ANN accuracy 0.59–0.98 (wide CI). Histology is reference; small balanced sample and retrospective design limit clinical generalization.
Test and result
Specific histology-linked sonographic differences and ANN discrimination evaluated against pathological diagnoses constitute a defined diagnostic result. Accuracy varies widely; do not claim established clinical readiness.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13044-023-00168-2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Circularly polarized printed dual port MIMO antenna with polarization diversity optimized by machine learning approach for 5G NR n77/n78 frequency band applications

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Scientific Reports full text identifies the DOI/title and first publication date (2023-08-26); Methods describe CATBoost/ANN optimization of defected-ground shifts from simulation-derived axial-ratio data (with an 80:20/90:10 split), and Results report fabricated-prototype VNA measurements, measured 3.17–4.55 GHz resonance and 2.66–3.92 GHz circular-polarization bands, gain/radiation measurements, and polarization diversity. The finding is a bounded experimentally checked antenna design, not a claim of deployment or network performance.

Technology: CATBoost and ANN map right/left shifts of the defected ground structure to simulated axial-ratio outcomes and select an optimized geometry; this operation is directly tied to the antenna design.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
CATBoost and ANN map right/left shifts of the defected ground structure to simulated axial-ratio outcomes and select an optimized geometry; this operation is directly tied to the antenna design.
Limitation / still unknown
The result is limited to the reported dataset, experimental setting and validation in this study; broader generalization and real-world benefit are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/article/MED/37634021 · https://pmc.ncbi.nlm.nih.gov/articles/PMC10460397/ · https://www.nature.com/articles/s41598-023-41302-2 · https://pubmed.ncbi.nlm.nih.gov/37634021/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning-Based Evaluation of Ultrasound Images for Benign Skin Tumors.

Source check: Source checked. Not independent scientific confirmation.

Finding: This retrospective, single-center diagnostic study uses a residual-plus-attention CNN to classify three histopathology-confirmed benign skin tumors from 698 ultrasound images. The source reports patient-grouped five-fold evaluation, 95.87% mean accuracy, 90.10% sensitivity and 96.23% specificity, with class-activation-map comparison to physician findings. The result is limited to this cohort and needs external validation.

Technology: 2. Materials and Methods, 2.4.1; 3. Results: 250 patients were assigned by patient to five groups, with pathology labels; the combined CNN is evaluated by five-fold cross-validation and CAM comparison.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
2. Materials and Methods, 2.4.1; 3. Results: 250 patients were assigned by patient to five groups, with pathology labels; the combined CNN is evaluated by five-fold cross-validation and CAM comparison.
Test and result
2. Materials and Methods, 2.4.1; 3. Results: 250 patients were assigned by patient to five groups, with pathology labels; the combined CNN is evaluated by five-fold cross-validation and CAM comparison.
Limitation / still unknown
2. Materials and Methods, 2.4.1; 3. Results: 250 patients were assigned by patient to five groups, with pathology labels; the combined CNN is evaluated by five-fold cross-validation and CAM comparison.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/sensors/sensors-23-07374/article_deploy/sensors-23-07374.pdf · https://api.crossref.org/works/10.3390/s23177374

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and validation of convolutional neural network-based model to predict the risk of sentinel or non-sentinel lymph node metastasis in patients with breast cancer: a machine learning study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A CNN using learned MRI representations predicted pathology-referenced lymph-node metastasis in the retrospective cohort; second-stage sentinel-dissection input and hypothetical surgery decisions do not demonstrate clinical benefit.

Technology: A convolutional neural network learned MRI representations to predict sentinel and non-sentinel lymph-node metastasis, evaluated against retrospective nodal references.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A convolutional neural network learned MRI representations to predict sentinel and non-sentinel lymph-node metastasis, evaluated against retrospective nodal references.
Limitation / still unknown
A CNN using learned MRI representations predicted pathology-referenced lymph-node metastasis in the retrospective cohort; second-stage sentinel-dissection input and hypothetical surgery decisions do not demonstrate clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10474371/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Classification of Alzheimer's disease stages from magnetic resonance images using deep learning

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary PeerJ article/PDF Methods and Results report CNN/transfer-learning image classification on ADNI/OASIS MRI with test performance by stage; scope is retrospective benchmark classification.

Technology: Deep CNN family predicts Alzheimer stage labels from MRI. Results are strongest for AD/control and weaker for adjacent MCI groups; no clinical deployment claim.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Deep CNN family predicts Alzheimer stage labels from MRI. Results are strongest for AD/control and weaker for adjacent MCI groups; no clinical deployment claim.
Test and result
Deep CNN family predicts Alzheimer stage labels from MRI. Results are strongest for AD/control and weaker for adjacent MCI groups; no clinical deployment claim.
Limitation / still unknown
Primary PeerJ article/PDF Methods and Results report CNN/transfer-learning image classification on ADNI/OASIS MRI with test performance by stage; scope is retrospective benchmark classification. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://peerj.com/articles/cs-1490.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Deep Learning-Based Algorithm for Identifying Precipitation Clouds Using Fengyun-4A Satellite Observation Data.

Source check: Source checked. Not independent scientific confirmation.

Finding: PCINet classified precipitation clouds from Fengyun-4A observations across day/night conditions using image-to-image segmentation. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Methods describe multi-scale CNN and skip constraints; Results compare cloud segmentation against five baselines across daytime, nighttime and twilight.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe multi-scale CNN and skip constraints; Results compare cloud segmentation against five baselines across daytime, nighttime and twilight.
Test and result
Methods describe multi-scale CNN and skip constraints; Results compare cloud segmentation against five baselines across daytime, nighttime and twilight.
Limitation / still unknown
Methods describe multi-scale CNN and skip constraints; Results compare cloud segmentation against five baselines across daytime, nighttime and twilight.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10422346/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Welding Spark Detection on Construction Sites Using Contour Detection with Automatic Parameter Tuning and Deep-Learning-Based Filters

Source check: Source checked. Not independent scientific confirmation.

Finding: A computer-vision/deep-learning detector identified welding sparks in construction-site video for a specific fire-safety task. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: YOLO-based deep-learning filters

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe contour detection and YOLO filters; Results compare welding-spark detection performance on real-site imagery.
Test and result
A computer-vision/deep-learning detector identified welding sparks in construction-site video for a specific fire-safety task.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10422306/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification and validation of the diagnostic signature associated with immune microenvironment of acute kidney injury based on ferroptosis-related genes through integrated bioinformatics analysis and machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Machine learning selected acute-kidney-injury diagnostic markers that were validated across datasets and by PCR.

Technology: Machine learning selected acute-kidney-injury diagnostic markers that were validated across datasets and by PCR.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine learning selected acute-kidney-injury diagnostic markers that were validated across datasets and by PCR.
Limitation / still unknown
Dataset and PCR checks support candidate-marker associations, not causal ferroptosis mechanisms or validated clinical diagnostic utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10413118/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Using Hypothesis-Led Machine Learning and Hierarchical Cluster Analysis to Identify Disease Pathways Prior to Dementia: Longitudinal Cohort Study

Source check: Source checked. Not independent scientific confirmation.

Finding: Hypothesis-led machine learning and hierarchical clustering identified diagnostic and symptom sequences preceding dementia in a matched longitudinal claims cohort, with held-out analysis. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Hypothesis-led machine-learning analysis

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe nationwide cohort and matched train/test split; Results report temporal disease/symptom patterns before dementia.
Test and result
Hypothesis-led machine learning and hierarchical clustering identified diagnostic and symptom sequences preceding dementia in a matched longitudinal claims cohort, with held-out analysis.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10413246/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Ovarian Recurrence risk Assessment using Clinical and serum protein LEvels (ORACLE): A Machine Learning Risk Score to Predict Survival in High Grade Ovarian Cancer

Source check: Source checked. Not independent scientific confirmation.

Finding: The ORACLE ML risk score combines serum proteins and clinical factors to predict ovarian-cancer recurrence and is evaluated in an independent prospectively collected cohort; the abstract supports a disease-specific prognostic result.

Technology: The ORACLE machine-learning risk score combines serum-protein and clinical predictors to estimate ovarian-cancer recurrence and survival.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The ORACLE machine-learning risk score combines serum-protein and clinical predictors to estimate ovarian-cancer recurrence and survival.
Test and result
The preprint reports evaluation in an independent prospectively collected cohort for disease-specific recurrence/survival prediction.
Limitation / still unknown
The preprint and described cohort evidence do not establish clinical utility; external reproducibility and prospective impact are unknown.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-3112324/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Continuous and early prediction of future moderate and severe Acute Kidney Injury in critically ill patients: Development and multi-centric, multi-national external validation of a machine-learning model.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: An hourly machine-learning model predicted future stage 2/3 acute kidney injury across 16,760 ICU patients from 145 centers in three countries, with multicenter external validation reported.

Technology: An hourly machine-learning model predicted future stage 2/3 acute kidney injury across 16,760 ICU patients from 145 centers in three countries, with multicenter external validation reported.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
An hourly machine-learning model predicted future stage 2/3 acute kidney injury across 16,760 ICU patients from 145 centers in three countries, with multicenter external validation reported.
Limitation / still unknown
Multicenter external validation supports prediction across the reported ICU data; prospective clinical deployment and improved AKI outcomes were not demonstrated.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1371/journal.pone.0287398

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Ovarian recurrence risk assessment using machine learning, clinical information, and serum protein levels to predict survival in high grade ovarian cancer

Source check: Source checked. Not independent scientific confirmation.

Finding: The July 25, 2023 Research Square v1 reports ORACLE, an elastic-net score combining clinical factors with serum proteins to stratify time-to-recurrence risk after complete response to chemotherapy for high-grade ovarian cancer. In a discovery cohort of 71 patients, the score separated recurrence outcomes; a prospectively collected independent cohort of 33 patients showed validation TTR HR 4.71 (95% CI 1.75–12.7, p=0.002). The peer-reviewed journal version appeared November 27, 2023 and reports the same study. Validation was single-institution and does not establish clinical utility.

Technology: ORACLE elastic-net risk score

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Elastic net combines age, treatment/surgical factors, stage, CA125 and four serum proteins into the ORACLE score; the score is tuned against survival/TTR.
Test and result
ORACLE combined age, treatment/surgical factors, stage, CA125 and serum proteins. The discovery cohort included 71 patients; an independently prospectively collected 33-patient cohort had TTR HR 4.71 (95% CI 1.75–12.7, p=0.002) for the reported risk groups.
Limitation / still unknown
The validation cohort was prospectively collected at the same institution; wider external validity and clinical utility remain unestablished.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://assets-eu.researchsquare.com/files/rs-3112324/v1/67b40c92-f203-4ef6-b534-f8184d9ab326.pdf · https://www.nature.com/articles/s41598-023-47983-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial Intelligence to Predict the BRAF V595E Mutation in Canine Urinary Bladder Urothelial Carcinomas

Source check: Source checked. Not independent scientific confirmation.

Finding: U-Net analysis of H&E slides predicts PCR-confirmed canine BRAF V595E mutation in a held-out set of 84 cases; sensitivity/specificity are modest and materially depend on tissue quality, so the finding is a bounded veterinary diagnostic proof of concept. PCR is a reference standard in a case-held-out validation set; performance is limited (57/96 agreement; 58% sensitivity, 63% specificity) and quality-dependent, with no external site validation.

Technology: Visiopharm decision-forest tissue segmentation and U-Net tumor segmentation/mutation prediction operate on H&E whole-slide images, trained against ddPCR and compared with PCR in held-out cases.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Visiopharm decision-forest tissue segmentation and U-Net tumor segmentation/mutation prediction operate on H&E whole-slide images, trained against ddPCR and compared with PCR in held-out cases.
Test and result
The result is prediction of a molecular tumor mutation from routine histology images, a specific diagnostic application; BRAF mutation itself was known before this paper.
Limitation / still unknown
PCR is a reference standard in a case-held-out validation set; performance is limited (57/96 agreement; 58% sensitivity, 63% specificity) and quality-dependent, with no external site validation. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/research/bionlp/RESTful/pmcoa.cgi/BioC_json/PMC10416820/unicode · https://doi.org/10.3390/ani13152404

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Using Ensemble OCT-Derived Features beyond Intensity Features for Enhanced Stargardt Atrophy Prediction with Deep Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Ensemble deep network using OCT-derived layer features predicts Stargardt atrophy with Dice gains over mean-intensity baselines, a distinct clinical imaging result.

Technology: Machine-learning analysis applies the stated model to the paper’s computer science dataset and evaluates the reported outcome.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis applies the stated model to the paper’s computer science dataset and evaluates the reported outcome.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/applsci/applsci-13-08555/article_deploy/applsci-13-08555.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development of a Soft Sensor Using Machine Learning Algorithms for Predicting the Water Quality of an Onsite Wastewater Treatment System.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: SVR and Cubist soft sensors use in-line water measurements to estimate offline COD, TSS and E. coli in a field-tested treatment system. Two years of site data (56 points) support COD/TSS estimates; E. coli performance is explicitly poor and transfer to other sites is untested.

Technology: SVR and Cubist soft sensors use in-line water measurements to estimate offline COD, TSS and E. coli in a field-tested treatment system.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
SVR and Cubist soft sensors use in-line water measurements to estimate offline COD, TSS and E. coli in a field-tested treatment system.
Limitation / still unknown
The field soft sensor was tested at one site with 56 samples; E. coli estimates performed poorly and transfer to other sites remains untested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10515708/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Machine Learning Model for the Accurate Prediction of 1-Year Survival in TAVI Patients: A Retrospective Observational Cohort Study

Source check: Source checked. Not independent scientific confirmation.

Finding: An earlier PMU Science Get Together abstract reported ML prediction of one-year mortality in the same 629-patient TAVI cohort using the same model family and reject option (NPV 98%, PPV 89% after abstaining on 145/447); the later journal article reports revised metrics (NPV 96%, PPV 92%). The retrospective, single-center result remains limited by selective abstention. Machine-learning models predicted one-year survival after TAVI, with 10-fold cross-validation and a separate test sample; a random-forest reject-option subgroup yielded 96% accuracy. This retrospective cohort result is limited by selective abstention and single-center data.

Technology: Five ML approaches including RF/SVM trained in learning samples using 10-fold CV, then evaluated in held-out test data. 629 transfemoral TAVI patients; 565 had one-year outcome. With reject option, model predicts for a subset (~81% of original cohort) with reported accuracy 96%, NPV 96%, PPV 92%.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Five ML approaches including RF/SVM trained in learning samples using 10-fold CV, then evaluated in held-out test data. 629 transfemoral TAVI patients; 565 had one-year outcome. With reject option, model predicts for a subset (~81% of original cohort) with reported accuracy 96%, NPV 96%, PPV 92%.
Test and result
Five ML approaches including RF/SVM trained in learning samples using 10-fold CV, then evaluated in held-out test data. 629 transfemoral TAVI patients; 565 had one-year outcome. With reject option, model predicts for a subset (~81% of original cohort) with reported accuracy 96%, NPV 96%, PPV 92%.
Limitation / still unknown
Machine-learning models predicted one-year survival after TAVI, with 10-fold cross-validation and a separate test sample; a random-forest reject-option subgroup yielded 96% accuracy. This retrospective cohort result is limited by selective abstention and single-center data. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcm/jcm-12-05481/article_deploy/jcm-12-05481.pdf · https://www.pmu.ac.at/fileadmin/CONTENT/BEREICHE/Forschungsmanagement/Dokumente/FM_Dokumente/SGT/Abstractband___Wissenschaftsbericht/SGT23_Abstractband_Digital.pdf · https://www.pmu.ac.at/en/research-innovation/research-management/news-events/science-get-together-sgt/vsgt23 · https://api.crossref.org/works/10.3390/jcm12175481

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Feasibility of AI-assisted compressed sensing protocols in knee MR imaging: a prospective multi-reader study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A prospective multi-reader study compared AI-assisted compressed-sensing knee MRI against parallel imaging in 130 participants (150 examinations); radiologists found equivalent detection performance for meniscal, cruciate-ligament and cartilage abnormalities with shorter scans. This is a specific clinical diagnostic-protocol result, not patient-outcome evidence.

Technology: The AI-assisted compressed-sensing reconstruction integrates a trained CNN module into compressed sensing for accelerated knee MRI.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The AI-assisted compressed-sensing reconstruction integrates a trained CNN module into compressed sensing for accelerated knee MRI.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s00330-023-09823-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Dysbiosis signatures of gut microbiota and the progression of type 2 diabetes: a machine learning approach in a Mexican cohort

Source check: Source checked. Not independent scientific confirmation.

Finding: In a cohort of 410 treatment-naïve Mexican participants, random-forest microbiome models classified T2D and prediabetes versus normoglycemia and identified candidate taxa associated with each group. This is cross-sectional association, not causal evidence of disease progression or a validated diagnostic. Cohort stratification and comparative model performance reported; cross-sectional cohort and association-only interpretation retained.

Technology: Six ML models compare 16S rRNA microbiome profiles; random forest ranks taxa for glycemic-group classification.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Six ML models compare 16S rRNA microbiome profiles; random forest ranks taxa for glycemic-group classification.
Test and result
In a cohort of 410 treatment-naïve Mexican participants, random-forest microbiome models classified T2D and prediabetes versus normoglycemia and identified candidate taxa associated with each group. This is cross-sectional association, not causal evidence of disease progression or a validated diagnostic.
Limitation / still unknown
Cohort stratification and comparative model performance reported; cross-sectional cohort and association-only interpretation retained. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2023.1170459/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of Adolescent Menarche Status Using Biplanar X-ray Images: A Deep Learning-Based Method.

Source check: Source checked. Not independent scientific confirmation.

Finding: Deep-learning ROI detection plus classifier identifies adolescent menarche status from EOS radiographs, evaluated on a retrospective development set and a separate clinical validation set (AUC 0.959 for females).

Technology: Deep-learning region-of-interest detection and classification estimate adolescent menarche status from biplanar EOS radiographs.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep-learning region-of-interest detection and classification estimate adolescent menarche status from biplanar EOS radiographs.
Test and result
A retrospective development set and separate clinical validation set were used; the reported female validation AUC was 0.959.
Limitation / still unknown
The result is an imaging classification in the studied cohorts and does not establish broader clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/bioengineering/bioengineering-10-00769/article_deploy/bioengineering-10-00769.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Modality-based attention and dual-stream multiple instance convolutional neural network for predicting microvascular invasion of hepatocellular carcinoma.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Original full text, Methods/Results: modality attention plus dual-stream multiple-instance CNN used MRI to predict pathology-confirmed MVI in 283 HCC cases; 4:1 split and five-fold CV, reported AUC 0.7422. Retrospective diagnostic result; no clinical-outcome benefit claimed.

Technology: Methods identify the ML/deep-learning operation used for the specific result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods identify the ML/deep-learning operation used for the specific result.
Limitation / still unknown
The 283-case retrospective study reports internal split/CV performance (AUC 0.7422); external prospective utility and improved outcomes were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.1195110/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Computer-aided diagnosis using embedded ensemble deep learning for multiclass drug-resistant tuberculosis classification.

Source check: Source checked. Not independent scientific confirmation.

Finding: An ensemble of three CNNs classifies drug-resistant TB on 5,039 chest X-rays with an 80/20 split and a further 760-image check, reaching 93.4% test accuracy, plus a usability study with 33 staff. This is a retrospective diagnostic validation of a specific clinical classifier. The reviewed source supports this bounded result; external generalization, causal effects, and sustained operational impact remain unestablished where the source reports retrospective or internal evaluation.

Technology: Machine-learning analysis reported for Computer-aided diagnosis using embedded ensemble deep learning for multiclass drug-resistant tuberculosis classification.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis reported for Computer-aided diagnosis using embedded ensemble deep learning for multiclass drug-resistant tuberculosis classification..
Test and result
An ensemble of three CNNs classifies drug-resistant TB on 5,039 chest X-rays with an 80/20 split and a further 760-image check, reaching 93.4% test accuracy, plus a usability study with 33 staff. This is a retrospective diagnostic validation of a specific clinical classifier.
Limitation / still unknown
The result is limited to the reported study cohort, data split and comparator; external validation was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2023.1122222/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A new RNN based machine learning model to forecast COVID-19 incidence, enhanced by the use of mobility data from the bike-sharing service in Madrid

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full article evaluates an RNN/LSTM COVID-19 incidence forecast using Madrid bike-sharing mobility data; time-series training/validation and comparisons report increased forecast accuracy with mobility inputs. Bounded historical epidemiological forecasting, not causal evidence that mobility changes infections.

Technology: Full article evaluates an RNN/LSTM COVID-19 incidence forecast using Madrid bike-sharing mobility data; time-series training/validation and comparisons report increased forecast accuracy with mobility inputs. Bounded historical epidemiological forecasting, not causal evidence that mobility changes infections.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full article evaluates an RNN/LSTM COVID-19 incidence forecast using Madrid bike-sharing mobility data; time-series training/validation and comparisons report increased forecast accuracy with mobility inputs. Bounded historical epidemiological forecasting, not causal evidence that mobility changes infections.
Limitation / still unknown
The result is limited to the reported dataset, experimental setting and validation in this study; broader generalization and real-world benefit are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1016/j.heliyon.2023.e17625 · https://pmc.ncbi.nlm.nih.gov/articles/PMC10290181/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Migraine with aura detection and subtype classification using machine learning algorithms and morphometric magnetic resonance imaging data

Source check: Source checked. Not independent scientific confirmation.

Finding: MRI morphometry plus machine learning distinguished migraine with aura and its subtypes in a disease-specific imaging cohort; the small sample limits generalization. Methods analyze MRI morphometry from 78 subjects with ML; Results distinguish migraine-with-aura groups, with limited sample size.

Technology: Methods analyze MRI morphometry from 78 subjects with ML; Results distinguish migraine-with-aura groups, with limited sample size.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods analyze MRI morphometry from 78 subjects with ML; Results distinguish migraine-with-aura groups, with limited sample size.
Test and result
Methods analyze MRI morphometry from 78 subjects with ML; Results distinguish migraine-with-aura groups, with limited sample size.
Limitation / still unknown
Methods analyze MRI morphometry from 78 subjects with ML; Results distinguish migraine-with-aura groups, with limited sample size. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10333052/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Study on the detection of water status of tomato ( Solanum lycopersicum L.) by multimodal deep learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A multimodal deep-learning model fuses RGB, NIR and depth images to classify five tomato water-status levels under controlled irrigation; held-out testing and single-modality comparisons are reported. This establishes a bounded crop-stress sensing result, not field irrigation benefit.

Technology: Multimodal fusion network combines RGB, NIR and depth imagery for tomato water-status classification.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Multimodal fusion network combines RGB, NIR and depth imagery for tomato water-status classification.
Limitation / still unknown
Methods/Results report irrigation classes, held-out evaluation and comparisons with single-modal models.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2023.1094142/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of high-risk factors associated with mortality at 1-, 3-, and 5-year intervals in gastric cancer patients undergoing radical surgery and immunotherapy: an 8-year multicenter retrospective analysis.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Machine-learning analysis of multicenter cohorts identified mortality risk factors after gastric-cancer surgery and immunotherapy, with external validation; the retrospective result does not establish treatment effects.

Technology: Machine-learning analysis of multicenter clinical cohorts identified mortality-associated factors after gastric-cancer surgery and immunotherapy, with external validation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis of multicenter clinical cohorts identified mortality-associated factors after gastric-cancer surgery and immunotherapy, with external validation.
Limitation / still unknown
The retrospective multicenter validation supports risk associations in the studied treatment cohorts; it does not establish causal treatment effects or prospective clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10264693/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Damage Detection of Unwashed Eggs through Video and Deep Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary Methods and Results support a bounded food-quality application; validation is limited to the reported samples and conditions.

Technology: YOLOv5-based video detection with tracking associates broken-egg detections across moving production-line frames.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
YOLOv5-based video detection with tracking associates broken-egg detections across moving production-line frames.
Test and result
Primary Results report the tested video method and practical production-line detection; model metrics are in the original Results.
Limitation / still unknown
Video and egg-line experiments evaluate detection under dynamic conditions; no broad commercial generalization is claimed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2304-8158/12/11/2179/pdf?version=1685353425

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Standardization of Body Composition Status in Patients with Advanced Urothelial Tumors: The Role of a CT-Based AI-Powered Software for the Assessment of Sarcopenia and Patient Outcome Correlation.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Quantib CT software automatically segmented L3 muscle in 97 retrospectively studied advanced urothelial-tumor patients; body-composition changes associated with chemotherapy response/survival. Single-center observational correlation, not a causal treatment effect.

Technology: Quantib CT software automatically segmented L3 muscle in 97 retrospectively studied advanced urothelial-tumor patients; body-composition changes associated with chemotherapy response/survival. Single-center observational correlati

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Quantib CT software automatically segmented L3 muscle in 97 retrospectively studied advanced urothelial-tumor patients; body-composition changes associated with chemotherapy response/survival. Single-center observational correlation, not a causal treatment effect.
Limitation / still unknown
This is a single-center retrospective association; body-composition changes do not show that treatment caused the observed response or survival differences.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2072-6694/15/11/2968

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Comprehensive machine learning-generated classifier identifies pro-metastatic characteristics and predicts individual treatment in pancreatic cancer: A multicenter cohort study based on super-enhancer

Source check: Source checked. Not independent scientific confirmation.

Finding: PubMed/PMC primary full-text record gives the exact DOI/title and article text. Methods describe 1,705 patients across 14 cohorts and 188 ML integrations for SEMet; Results link classification to metastatic phenotype and treatment response, with mouse validation. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: SEMet super-enhancer machine-learning classifier

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Super-enhancer-related ML classifier built from ChIP-seq features and multicohort outcomes; validation includes independent cohorts and animal experiments. Retrospective classifier evidence does not establish prospective treatment benefit.
Test and result
PubMed/PMC primary full-text record gives the exact DOI/title and article text. Methods describe 1,705 patients across 14 cohorts and 188 ML integrations for SEMet; Results link classification to metastatic phenotype and treatment response, with mouse validation.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10283048/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Accounting for uncertainty in training data to improve machine learning performance in predicting new disease activity in early multiple sclerosis.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A probabilistic random forest accounted for uncertain training labels when predicting new disease activity in early multiple sclerosis trial data.

Technology: compare handling of uncertain MS activity labels with nested cross-validation; Results report performance on confirmed-label cases.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
compare handling of uncertain MS activity labels with nested cross-validation; Results report performance on confirmed-label cases.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10251494/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting molecular mechanisms of hereditary diseases by using their tissue-selective manifestation

Source check: Source checked. Not independent scientific confirmation.

Finding: An interpretable tissue-feature ML framework predicts tissue-associated disease genes across 18,927 protein-coding genes and eight tissues and analyzes tissue-selective hereditary disease mechanisms; prediction is supported by held-out disease-gene evaluation and patient-diagnostic analyses. Full Molecular Systems Biology article, Methods and Results, held-out evaluation and diagnostic case analyses inspected.

Technology: Interpretable ML combines tissue expression, eQTL, PPI and gene-function features to score tissue-associated disease genes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Interpretable ML combines tissue expression, eQTL, PPI and gene-function features to score tissue-associated disease genes.
Test and result
An interpretable tissue-feature ML framework predicts tissue-associated disease genes across 18,927 protein-coding genes and eight tissues and analyzes tissue-selective hereditary disease mechanisms; prediction is supported by held-out disease-gene evaluation and patient-diagnostic analyses.
Limitation / still unknown
Full Molecular Systems Biology article, Methods and Results, held-out evaluation and diagnostic case analyses inspected. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.15252/msb.202211407

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction model of in-hospital mortality in intensive care unit patients with cardiac arrest: a retrospective analysis of MIMIC -IV database based on machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: LASSO and XGBoost feature screening support a 70/30 MIMIC-IV model comparison for ICU cardiac-arrest in-hospital mortality (1,722 patients); models were assessed for discrimination/calibration/clinical utility and compared with NEWS2. The selected LASSO nomogram is a retrospective single-database risk model, not clinical use evidence.

Technology: MIMIC-IV clinical features are screened with LASSO/XGBoost; final LASSO risk model.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
MIMIC-IV clinical features are screened with LASSO/XGBoost; final LASSO risk model.
Limitation / still unknown
Methods and Results: random 70/30 patient split, validation set n=516; compared AUC/calibration/DCA and NEWS2.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12871-023-02138-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial intelligence-assisted ultrasound image analysis to discriminate early breast cancer in Chinese population: a retrospective, multicentre, cohort study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The AI-assisted ultrasound system combined learned image representations from paired views to distinguish lesions classified as benign on initial ultrasound that were later diagnosed as malignant at pathology or follow-up. This retrospective, selected cohort result includes external hospital assessment but does not show outcome benefit or new biology.

Technology: A deep model extracted paired ultrasound-image features for lesion classification.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A deep model extracted paired ultrasound-image features for lesion classification.
Limitation / still unknown
Retrospective selection of ultrasound-benign lesions and model selection limit generalization; no treatment or patient-outcome benefit was tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10220307/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

hist2RNA : An Efficient Deep Learning Architecture to Predict Gene Expression from Breast Cancer Histopathology Images.

Source check: Source checked. Not independent scientific confirmation.

Finding: hist2RNA inferred breast-cancer gene-expression patterns from histopathology images, with held-out and external tissue-microarray evaluation. External replication and prospective utility were not established in the reviewed material.

Technology: The study uses a deep neural network for the stated hist2rna : an efficient deep learning architecture to predict gene expression from breast cancer histopathology images task.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study uses a deep neural network for the stated hist2rna : an efficient deep learning architecture to predict gene expression from breast cancer histopathology images task.
Test and result
hist2RNA inferred breast-cancer gene-expression patterns from histopathology images, with held-out and external tissue-microarray evaluation.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10177559/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and validation of a deep learning-based model to distinguish acetabular fractures on pelvic anteroposterior radiographs.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A deep-learning model distinguishes acetabular fractures on pelvic AP radiographs and is validated against clinicians and separate data; the paper reports diagnostic comparison on a specific fracture question. This is retrospective diagnostic validation, not prospective clinical benefit.

Technology: DenseNet-169-based model classifies acetabular fracture images.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
DenseNet-169-based model classifies acetabular fracture images.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2023.1146910/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Detection of the rotator cuff tears using a novel convolutional neural network from magnetic resonance image (MRI)

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full article evaluates a CNN on MRI for rotator-cuff tear detection, comparing 2D and 3D imaging and clinician interpretations. The specific tear-detection result is evaluated on patient MRI data; conclusions limited to that dataset and reader comparison.

Technology: Full article evaluates a CNN on MRI for rotator-cuff tear detection, comparing 2D and 3D imaging and clinician interpretations. The specific tear-detection result is evaluated on patient MRI data; conclusions limited to that dataset and reader comparison.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full article evaluates a CNN on MRI for rotator-cuff tear detection, comparing 2D and 3D imaging and clinician interpretations. The specific tear-detection result is evaluated on patient MRI data; conclusions limited to that dataset and reader comparison.
Limitation / still unknown
Full article evaluates a CNN on MRI for rotator-cuff tear detection, comparing 2D and 3D imaging and clinician interpretations. The specific tear-detection result is evaluated on patient MRI data; conclusions limited to that dataset and reader comparison. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10189183/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Accuracy of Artificial Intelligence-Based Automated Quantitative Coronary Angiography Compared to Intravascular Ultrasound: Retrospective Cohort Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: In 47 patients with 54 significant coronary lesions, AI-QCA measurements showed moderate-to-strong correlation with IVUS for proximal reference area (r=0.57), distal reference area (r=0.80) and minimal luminal area (r=0.52); correlation was weaker for percent area stenosis (r=0.29) and lesion length (r=0.33). Correcting proximal/distal margin mismatch strengthened reference-area correlations to r=0.70 and r=0.83, respectively.

Technology: Fully automated artificial-intelligence quantitative coronary angiography (AI-QCA) measured coronary lesion dimensions for comparison with IVUS; adjusting lesion margins reduced geographic mismatch.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Automated image analysis measured reference vessel areas, minimal luminal area, percent area stenosis and lesion length from coronary angiograms.
Limitation / still unknown
Retrospective single-center comparison in 47 patients and 54 lesions; AI-QCA correlated less well for percent stenosis and lesion length and showed geographic-margin mismatch, especially distally.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.2196/45299 · https://europepmc.org/article/MED/37099368 · https://europepmc.org/articles/PMC10173041 · https://www.ncbi.nlm.nih.gov/research/bionlp/RESTful/pmcoa.cgi/BioC_json/PMC10173041/unicode

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Fecal-Microbial-Extracellular-Vesicles-Based Metabolomics Machine Learning Framework and Biomarker Discovery for Predicting Colorectal Cancer Patients.

Source check: Source checked. Not independent scientific confirmation.

Finding: Fecal microbial extracellular-vesicle metabolomics in 36 CRC cases and 40 controls identifies five metabolite features; SVM 10-fold CV AUC .985 (35/36 cases, 39/40 controls). External replication and prospective utility were not established in the reviewed material.

Technology: Fecal microbial extracellular-vesicle metabolomics in 36 CRC cases and 40 controls identifies five metabolite features; SVM 10-fold CV AUC .985 (35/36 cases, 39/40 controls).

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Fecal microbial extracellular-vesicle metabolomics in 36 CRC cases and 40 controls identifies five metabolite features; SVM 10-fold CV AUC .985 (35/36 cases, 39/40 controls).
Test and result
Fecal microbial extracellular-vesicle metabolomics in 36 CRC cases and 40 controls identifies five metabolite features; SVM 10-fold CV AUC .985 (35/36 cases, 39/40 controls).
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3390/metabo13050589

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction of Water Resistance of Magnesium Oxychloride Cement Concrete Based upon Hybrid-BP Neural Network

Source check: Source checked. Not independent scientific confirmation.

Finding: PSO/GWO/WOA-optimized BP networks predict water resistance of magnesium oxychloride concrete from 144 samples, split 100/22/22; reported test R² up to .99 and RMSE .73 for compressive strength and .62 for softening coefficient. This is a measured-material property prediction. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: PSO/GWO/WOA-optimized BP neural networks

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PSO/GWO/WOA-optimized BP networks predict water resistance of magnesium oxychloride concrete from 144 samples, split 100/22/22; reported test R² up to .99 and RMSE .73 for compressive strength and .62 for softening coefficient. This is a measured-material property prediction.
Test and result
PSO/GWO/WOA-optimized BP networks predict water resistance of magnesium oxychloride concrete from 144 samples, split 100/22/22; reported test R² up to .99 and RMSE .73 for compressive strength and .62 for softening coefficient. This is a measured-material property prediction.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3390/ma16093371 · https://www.mdpi.com/1996-1944/16/9/3371

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Prediction Model for Osteoporosis Risk Using a Machine-Learning Approach and Its Validation in a Large Cohort.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: XGBoost selected among seven models using 8,421 Ansan/Anseong participants and 80/20 train/test; selected model was applied to HEXA cohort. AUC reached .890 internally; fracture association in HEXA is indirect because that cohort lacked measured BMD.

Technology: Primary Methods/Results: XGBoost selected among seven models using 8,421 Ansan/Anseong participants and 80/20 train/test; selected model was applied to HEXA cohort. AUC reached .890 internally; fracture association in HEXA is indirect because that cohort lacked measured BMD.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
XGBoost selected among seven models using 8,421 Ansan/Anseong participants and 80/20 train/test; selected model was applied to HEXA cohort. AUC reached .890 internally; fracture association in HEXA is indirect because that cohort lacked measured BMD.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.jkms.org/DOIx.php?id=10.3346%2Fjkms.2023.38.e162

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of patients with dilated phase of hypertrophic cardiomyopathy using a convolutional neural network applied to multiple, dual, and single lead electrocardiograms

Source check: Source checked. Not independent scientific confirmation.

Finding: A CNN applied to ECGs identified dilated-phase hypertrophic cardiomyopathy in a retrospective single-center cohort; in the held-out test data the eight-lead AUC was 0.929 and V5 single-lead AUC 0.953. This is a specific diagnostic-validation result, not evidence of clinical deployment. PMC10160501 Methods/Results: Shinken cohort, 17,378 analyzed subjects, 54 dHCM cases, CNN, five-run internal test evaluation and lead comparisons. Small positive cohort and no external validation limit scope. Catalog DCM ECG paper is a different disease/model/cohort, not same result.

Technology: PMC10160501 Methods/Results: Shinken cohort, 17,378 analyzed subjects, 54 dHCM cases, CNN, five-run internal test evaluation and lead comparisons. Small positive cohort and no external validation limit scope. Catalog DCM ECG paper is a different disease/model/cohort, not same result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PMC10160501 Methods/Results: Shinken cohort, 17,378 analyzed subjects, 54 dHCM cases, CNN, five-run internal test evaluation and lead comparisons. Small positive cohort and no external validation limit scope. Catalog DCM ECG paper is a different disease/model/cohort, not same result.
Test and result
PMC10160501 Methods/Results: Shinken cohort, 17,378 analyzed subjects, 54 dHCM cases, CNN, five-run internal test evaluation and lead comparisons. Small positive cohort and no external validation limit scope. Catalog DCM ECG paper is a different disease/model/cohort, not same result.
Limitation / still unknown
PMC10160501 Methods/Results: Shinken cohort, 17,378 analyzed subjects, 54 dHCM cases, CNN, five-run internal test evaluation and lead comparisons. Small positive cohort and no external validation limit scope. Catalog DCM ECG paper is a different disease/model/cohort, not same result. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10160501/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A graph neural network-based interpretable framework reveals a novel DNA fragility-associated chromatin structural unit

Source check: Source checked. Not independent scientific confirmation.

Finding: A GNN and GNNExplainer applied to Hi-C maps identify a DNA-fragility-associated chromatin interaction network and relate its neck interactions to double-strand breaks; this is a specific biological result. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Graph neural network and GNNExplainer

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A GNN and GNNExplainer applied to Hi-C maps identify a DNA-fragility-associated chromatin interaction network and relate its neck interactions to double-strand breaks; this is a specific biological result.
Test and result
A GNN and GNNExplainer applied to Hi-C maps identify a DNA-fragility-associated chromatin interaction network and relate its neck interactions to double-strand breaks; this is a specific biological result.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13059-023-02916-x

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A vision transformer for decoding surgeon activity from surgical videos.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary article reports a vision-transformer system that decodes surgical steps, actions and skill-related quality from videos across three hospitals; this is a specific clinical measurement study, with no claim of improved surgical outcomes.

Technology: The primary article reports a vision-transformer system that decodes surgical steps, actions and skill-related quality from videos across three hospitals; this is a specific clinical measurement study, with no claim of improved su

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The primary article reports a vision-transformer system that decodes surgical steps, actions and skill-related quality from videos across three hospitals; this is a specific clinical measurement study, with no claim of improved surgical outcomes.
Limitation / still unknown
The study measures activity/skill labels from surgical video; it does not show that using the measurements improves operative performance or patient outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41551-023-01010-8.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Factors contributing to preventing operating room "never events": a machine learning analysis.

Source check: Source checked. Not independent scientific confirmation.

Finding: A field-specific patient-safety result quantifies factors associated with actual never events; ML role and observed-data validation explicit. Association and modest event sample bound the claim.

Technology: 9,234 safety-standard observations and 101 actual root-cause analyses of major operating-room never events; three RF models, AUC 0.81–0.85, identify department-specific factors and associations with event probability. Retrospective case/root-cause data; findings are associations and do not establish causal prevention effects.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
9,234 safety-standard observations and 101 actual root-cause analyses of major operating-room never events; three RF models, AUC 0.81–0.85, identify department-specific factors and associations with event probability. Retrospective case/root-cause data; findings are associations and do not establish causal prevention effects.
Test and result
9,234 safety-standard observations and 101 actual root-cause analyses of major operating-room never events; three RF models, AUC 0.81–0.85, identify department-specific factors and associations with event probability. Retrospective case/root-cause data; findings are associations and do not establish causal prevention effects.
Limitation / still unknown
Methods/Results: 9,234 safety-standard observations and 101 actual root-cause analyses of major operating-room never events; three RF models, AUC 0.81–0.85, identify department-specific factors and associations with event probability. Retrospective case/root-cause data; findings are associations and do not establish causal prevention effects.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13037-023-00356-x

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Cardiovascular disease risk assessment using a deep-learning-based retinal biomarker: a comparison with existing risk scores.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A deep-learning retinal biomarker classified cardiovascular risk against established risk tools in two population cohorts; the source does not show treatment benefit.

Technology: deep-learning retinal biomarker Reti-CVD was compared with established risk-score groups in UK Biobank and Singapore SEED.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
deep-learning retinal biomarker Reti-CVD was compared with established risk-score groups in UK Biobank and Singapore SEED.
Limitation / still unknown
The risk classification was compared with existing risk scores in two cohorts; no treatment impact or patient benefit was evaluated.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10232236/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC10232236/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC10232236/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Treatment response assessment of glioma tumors using deep learning based on magnetic resonance imaging

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: U-Net segmentation and transfer-learned classification are applied to MRI before and after glioma radiotherapy, with the segmented volumes used to assess treatment response in 49 patients; this is a specific clinical imaging result.

Technology: U-Net segmentation and transfer-learned MRI classifier

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Segments glioma volumes on pre- and post-radiotherapy MRI and classifies treatment response in the reported 49-patient cohort.
Limitation / still unknown
The reported evaluation does not establish prospective benefit or performance beyond the studied setting.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-2733829/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Hole Depth Prediction in a Femtosecond Laser Drilling Process Using Deep Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: Publisher PDF Methods/Results: 200 laser-drilling data points, 140 train/60 test; DL predicts hole depth from process inputs; validation MAE chosen across architectures and complexity showed overfit. Exact test-error tables inspected in PDF. External replication and prospective utility were not established in the reviewed material.

Technology: Publisher PDF Methods/Results: 200 laser-drilling data points, 140 train/60 test; DL predicts hole depth from process inputs; validation MAE chosen across architectures and complexity showed overfit. Exact test-error tables inspected in PDF.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher PDF Methods/Results: 200 laser-drilling data points, 140 train/60 test; DL predicts hole depth from process inputs; validation MAE chosen across architectures and complexity showed overfit. Exact test-error tables inspected in PDF.
Test and result
Publisher PDF Methods/Results: 200 laser-drilling data points, 140 train/60 test; DL predicts hole depth from process inputs; validation MAE chosen across architectures and complexity showed overfit. Exact test-error tables inspected in PDF.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2072-666X/14/4/743/pdf?version=1679972257

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Simultaneous Sleep Stage and Sleep Disorder Detection from Multimodal Sensors Using Deep Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A source-described retrospective sleep-diagnostic study fuses EEG, ECG and EMG CNN outputs with a shallow neural network to recognize six sleep stages and eight disorder labels in the PhysioNet CAP Sleep database. The paper reports 94.34% sleep-stage and 99.09% disorder accuracy. Scope is a closed-set sample split: the same patient groups appear in train and test, so it does not establish performance on unseen patients.

Technology: 2. Materials and Methods, 2.5–2.6; 5. Conclusions: MML-DMS uses modality-specific CNNs and a fusion NN on 108 CAP Sleep participants; it reports six-stage/eight-disorder results and explicitly limits generalization to new patients.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
2. Materials and Methods, 2.5–2.6; 5. Conclusions: MML-DMS uses modality-specific CNNs and a fusion NN on 108 CAP Sleep participants; it reports six-stage/eight-disorder results and explicitly limits generalization to new patients.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/sensors/sensors-23-03468/article_deploy/sensors-23-03468.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Novel artificial intelligence-enabled deep learning system to enhance adenoma detection: a prospective randomized controlled study

Source check: Source checked. Not independent scientific confirmation.

Finding: A prospective randomized study of 674 white-light colonoscopies found higher adenoma detection with the DEEP2 computer-aided detection system (37% vs 27%; 122/330 vs 93/344) and more adenomas per colonoscopy, with about four false alerts per AI exam. The result is a bounded single-center clinical detection outcome. PMC12850877 Methods/Results: randomized 1:1 comparison, 1,033 enrolled and 674 analyzed after protocol exclusions; pathology-confirmed lesions; ADR and APC; unblinded study, single center, false alerts recorded by endoscopists. Existing catalog 2020 CADe trial is a different paper/system and cohort; not same result.

Technology: PMC12850877 Methods/Results: randomized 1:1 comparison, 1,033 enrolled and 674 analyzed after protocol exclusions; pathology-confirmed lesions; ADR and APC; unblinded study, single center, false alerts recorded by endoscopists. Existing catalog 2020 CADe trial is a different paper/system and cohort; not same result.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
PMC12850877 Methods/Results: randomized 1:1 comparison, 1,033 enrolled and 674 analyzed after protocol exclusions; pathology-confirmed lesions; ADR and APC; unblinded study, single center, false alerts recorded by endoscopists. Existing catalog 2020 CADe trial is a different paper/system and cohort; not same result.
Test and result
PMC12850877 Methods/Results: randomized 1:1 comparison, 1,033 enrolled and 674 analyzed after protocol exclusions; pathology-confirmed lesions; ADR and APC; unblinded study, single center, false alerts recorded by endoscopists. Existing catalog 2020 CADe trial is a different paper/system and cohort; not same result.
Limitation / still unknown
PMC12850877 Methods/Results: randomized 1:1 comparison, 1,033 enrolled and 674 analyzed after protocol exclusions; pathology-confirmed lesions; ADR and APC; unblinded study, single center, false alerts recorded by endoscopists. Existing catalog 2020 CADe trial is a different paper/system and cohort; not same result. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12850877/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Renalase Identified by Machine Learning Methods As A Novel Independent Predictor Of Mortality In Hospitalized Patients With COVID-19

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Among 437 hospitalized patients with COVID-19, XGBoost ranked low serum renalase among predictors of 180-day mortality. This single-cohort analysis had no external validation and does not establish renalase as a causal factor; it had also been discussed in earlier literature.

Technology: XGBoost ranked clinical and laboratory variables, including renalase, for 180-day mortality prediction.

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
XGBoost ranked clinical and laboratory variables, including renalase, for 180-day mortality prediction.
Limitation / still unknown
Single cohort of 437 hospitalized patients, without external validation; the association does not establish renalase as causal and renalase was already discussed in prior literature.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-2492699/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identifying subtypes of chronic kidney disease with machine learning: development, internal validation and prognostic validation using linked electronic health records in 350,067 individuals

Source check: Source checked. Not independent scientific confirmation.

Finding: Seven unsupervised clustering methods on linked EHR features identified five CKD profiles; XGBoost predicted subtype labels. Across incident and prevalent CKD cohorts (350,067 and 195,422), five profiles (early-onset, late-onset, cancer, metabolic, cardiometabolic) differed in 5-year mortality, admissions and incident chronic disease. Cardiometabolic incident subtype had highest 5-year mortality/admission risks; this is observational prognostic stratification, not causal etiology. Population-based UK EHR development; internal clustering/method consistency and prognostic validation across linked outcomes. No independent external geography validation, a limitation stated by authors. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Unsupervised clustering and XGBoost

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Seven unsupervised clustering methods on linked EHR features identified five CKD profiles; XGBoost predicted subtype labels.
Test and result
Seven unsupervised clustering methods on linked EHR features identified five CKD profiles; XGBoost predicted subtype labels. Across incident and prevalent CKD cohorts (350,067 and 195,422), five profiles (early-onset, late-onset, cancer, metabolic, cardiometabolic) differed in 5-year mortality, admissions and incident chronic disease. Cardiometabolic incident subtype had highest 5-year mortality/admission risks; this is observational prognostic stratification, not causal etiology. Population-based UK EHR development; internal clustering/method consistency and prognostic validation across linked outcomes. No independent external geography validation, a limitation stated by authors.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9989643/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A New Artificial Intelligence Approach Using Extreme Learning Machine as the Potentially Effective Model to Predict and Analyze the Diagnosis of Anemia.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: An ELM classifies laboratory-confirmed anemia types in a held-out patient set; evidence supports diagnostic performance, not clinical benefit or external generalization.

Technology: ELM maps CBC features to four anemia classes, compared with laboratory reference diagnoses.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ELM maps CBC features to four anemia classes, compared with laboratory reference diagnoses.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/healthcare/healthcare-11-00697/article_deploy/healthcare-11-00697.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction of Early Visual Outcome of Small-Incision Lenticule Extraction (SMILE) Based on Deep Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: ResNet50 and XGBoost models predicted early visual recovery from SMILE surgical videos, with prospective cohorts used for testing. The article reports 318 eyes but inconsistent patient totals (159 in the abstract versus 164 across results cohorts), and the eye-level split may leave paired-eye dependence.

Technology: ResNet50 classifies scanning images from SMILE surgical videos; PCA-compressed image probabilities feed an XGBoost video-level visual-recovery classifier; Grad-CAM shows image regions.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ResNet50 classifies scanning images from SMILE surgical videos; PCA-compressed image probabilities feed an XGBoost video-level visual-recovery classifier; Grad-CAM shows image regions.
Limitation / still unknown
ResNet50 and XGBoost models predicted early visual recovery from SMILE surgical videos, with prospective cohorts used for testing. The article reports 318 eyes but inconsistent patient totals (159 in the abstract versus 164 across results cohorts), and the eye-level split may leave paired-eye dependence. The result is limited to the reported cohort, data or experimental conditions and does not establish broader clinical or field benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10011351/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Impact Ionization Coefficient Prediction of a Lateral Power Device Using Deep Neural Network

Source check: Source checked. Not independent scientific confirmation.

Finding: DNN predicts lateral power-device impact-ionization coefficient as a function of 2D device variables, evaluated against TCAD simulations. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Deep neural network

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
DNN predicts lateral power-device impact-ionization coefficient as a function of 2D device variables, evaluated against TCAD simulations.
Test and result
DNN predicts lateral power-device impact-ionization coefficient as a function of 2D device variables, evaluated against TCAD simulations.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3390/mi14030522 · https://www.mdpi.com/2072-666X/14/3/522

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine-learning based strategy identifies a robust protein biomarker panel for Alzheimer’s disease in cerebrospinal fluid

Source check: Source checked. Not independent scientific confirmation.

Finding: The preprint evaluates the panel across five Alzheimer’s and neurodegeneration cohorts, including mild-cognitive-impairment discrimination. This supports a candidate biomarker and diagnostic result, not clinical utility or a validated treatment decision.

Technology: SVM-RFECV selects a 12-protein cerebrospinal-fluid panel for Alzheimer’s disease classification.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
SVM-RFECV selects a 12-protein cerebrospinal-fluid panel for Alzheimer’s disease classification.
Test and result
The preprint evaluates the panel across five Alzheimer’s and neurodegeneration cohorts, including mild-cognitive-impairment discrimination.
Limitation / still unknown
This supports a candidate biomarker and diagnostic result, not clinical utility or a validated treatment decision.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-2603826/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Construction of a Novel Diagnostic Model Based on Ferroptosis-Related Genes for Hepatocellular Carcinoma Using Machine and Deep Learning Methods.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: SVM-RFE/LASSO selection of ferroptosis-related genes and validation across HCC expression datasets yielded a disease-specific diagnostic signature.

Technology: ML feature selection ranked candidate ferroptosis genes for the HCC classifier.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ML feature selection ranked candidate ferroptosis genes for the HCC classifier.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9981290/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Effect of daily new cases of COVID-19 on public sentiment and concern: Deep learning-based sentiment classification and semantic network analysis

Source check: Source checked. Not independent scientific confirmation.

Finding: The study used ERNIE to classify manually annotated TikTok comments across five Shanghai lockdown stages and found stage-specific themes and concerns in 1,220 comments; the classifier reached 92.6% accuracy/F1 on its reported held-out evaluation. This is a specific public-health communication result from the study sample, not a causal estimate of video impact. Two annotators reached Holsti intercoder reliability 1.0; the classifier reports accuracy/precision/recall/F1 of 92.6% on its evaluation. The observation sample is one official account during Shanghai lockdown.

Technology: Authors trained ERNIE on 244 manually annotated comments (110 positive, 134 negative), then applied it to the remaining 80%; semantic-network analysis followed the classifications.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Authors trained ERNIE on 244 manually annotated comments (110 positive, 134 negative), then applied it to the remaining 80%; semantic-network analysis followed the classifications.
Test and result
From 79 official-account videos and 1,220 comments across five CERC stages, the analysis reports stage-specific public concerns; it does not establish that videos caused those attitudes.
Limitation / still unknown
Two annotators reached Holsti intercoder reliability 1.0; the classifier reports accuracy/precision/recall/F1 of 92.6% on its evaluation. The observation sample is one official account during Shanghai lockdown. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/content/pdf/10.1007/s10389-023-01833-4.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting the HER2 status in oesophageal cancer from tissue microarrays using convolutional neural networks.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A convolutional neural network classified esophageal-cancer HER2 status from tissue-microarray samples; the study evaluated it on an independent 307-sample test set drawn from 1,602 samples.

Technology: A convolutional neural network analyzed tissue-microarray images to classify HER2 status.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A convolutional neural network analyzed tissue-microarray images to classify HER2 status.
Limitation / still unknown
The result is limited to the study’s tissue-microarray samples and independent 307-sample test set; it does not establish clinical deployment or treatment benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1038/s41416-023-02143-y

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Algorithm for Efficient Design of Separated Buffer Super-Junction IGBT

Source check: Source checked. Not independent scientific confirmation.

Finding: ML surrogate models TCAD electrical characteristics for separated-buffer super-junction IGBT design and supports a selected optimized buffer design; results are simulation-based. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Machine-learning TCAD surrogate model

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ML surrogate models TCAD electrical characteristics for separated-buffer super-junction IGBT design and supports a selected optimized buffer design; results are simulation-based.
Test and result
ML surrogate models TCAD electrical characteristics for separated-buffer super-junction IGBT design and supports a selected optimized buffer design; results are simulation-based.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3390/mi14020334 · https://www.mdpi.com/2072-666X/14/2/334

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Energy Level Prediction of Organic Semiconductors for Photodetectors and Mining of a Photovoltaic Database to Search for New Building Units

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning models trained on literature-derived molecular descriptors estimate organic-semiconductor energy levels and screen a photovoltaic database for candidate building units; the candidates are computationally prioritized and not synthesized or measured here.

Technology: Machine-learning molecular-property models

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Uses literature-derived molecular descriptors to estimate organic-semiconductor energy levels and screen a photovoltaic database for candidate building units.
Test and result
The study evaluates descriptor-based energy-level predictions and database screening; candidate building units were computationally prioritized and not experimentally synthesized or measured.
Limitation / still unknown
the candidates are computationally prioritized and not synthesized or measured here.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/molecules/molecules-28-01240/article_deploy/molecules-28-01240.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Hybrid Generic Framework for Heart Problem Diagnosis Based on a Machine Learning Paradigm.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A two-stage hybrid ML/DL framework was evaluated on named heart-disease datasets and a separate 299-patient heart-failure survival dataset. This is a retrospective clinical prediction result, not evidence of clinical benefit or deployment.

Technology: The paper applies multiple ML/DL classifiers in a two-stage feature-selection and voting framework to clinical heart-disease/survival labels.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The paper applies multiple ML/DL classifiers in a two-stage feature-selection and voting framework to clinical heart-disease/survival labels.
Limitation / still unknown
A two-stage hybrid ML/DL framework was evaluated on named heart-disease datasets and a separate 299-patient heart-failure survival dataset. This is a retrospective clinical prediction result, not evidence of clinical benefit or deployment.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.ncbi.nlm.nih.gov/research/bionlp/RESTful/pmcoa.cgi/BioC_json/PMC9921250/unicode · https://europepmc.org/article/MED/36772430 · https://doi.org/10.3390/s23031392

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Interpretable machine learning analysis to identify risk factors for diabetes using the anonymous living census data of Japan

Source check: Source checked. Not independent scientific confirmation.

Finding: Interpretable ML on Japanese living-conditions census data identifies diabetes-associated factors, including several reported as new associations. These are cross-sectional associations, not causal risk factors; there is no external validation. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Logistic regression, random forest and XGBoost

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods/Results compare logistic, linear, LDA, RF and XGBoost models and SHAP on the Japanese survey; reported associations include household, insurance/pension and health-awareness variables, with 11 replicated factors. No external validation or causal design.
Test and result
Interpretable ML on Japanese living-conditions census data identifies diabetes-associated factors, including several reported as new associations. These are cross-sectional associations, not causal risk factors; there is no external validation.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s12553-023-00730-w

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning based identification of bone scintigraphies containing metastatic bone disease foci.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A deep-learning model detected metastatic bone-disease foci in bone scintigraphy and was evaluated on a 998-image external-center cohort.

Technology: training/validation and external test cohorts plus an in-silico clinical trial; Results report model performance and reader-relevant evaluation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
training/validation and external test cohorts plus an in-silico clinical trial; Results report model performance and reader-relevant evaluation.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9875407/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Diagnostic performance of artificial intelligence algorithms for detection of pulmonary involvement by COVID-19 based on portable radiography.

Source check: Source checked. Not independent scientific confirmation.

Finding: The checked primary abstract reports this specific result: Unverified Haiku assessment. A prospective observational study in which AI algorithms detect COVID-19 pulmonary involvement on portable radiography, with AUC comparisons against each other and against specialist assessment in a defined cohort. The AI role and evaluation fit the claim. The abstract identifies the model or learned operation and an observed evaluation; the conclusion is limited to the reported study setting. Original PMC full-text Methods and Results were checked; date 2022-12-30 reconciled to the primary record.

Technology: Publisher/PMC Methods establish the learned model and its operation in the specific study.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher/PMC Methods establish the learned model and its operation in the specific study.
Test and result
Publisher/PMC Results report the represented result and its cohort-specific scope.
Limitation / still unknown
Validation design and reported performance/comparison were checked in the full text and are stated narrowly.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9801183/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Electrolyte-Gated Graphene Field Effect Transistor-Based Ca 2+ Detection Aided by Machine Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Random-forest regression from graphene sensor transfer curves predicted spiked Ca²⁺ concentrations; on a 3,600-row test split, error was within one log unit for 95.11% of predictions and four sensor curves tracked nominal concentrations with Pearson r>0.999. This is a standard-solution proof of concept.

Technology: Methods §2.5 and Results §§3.4–3.5: regressors map sensor bias/transfer curves to log Ca²⁺ concentration.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods §2.5 and Results §§3.4–3.5: regressors map sensor bias/transfer curves to log Ca²⁺ concentration.
Limitation / still unknown
The calcium estimates were evaluated on spiked standard solutions and four sensor curves, not clinical samples or a deployed measurement workflow.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/1424-8220/23/1/353 · https://europepmc.org/article/MED/36616952

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The company we keep. Using hemodialysis social network data to classify patients' kidney transplant attitudes with machine learning algorithms.

Source check: Source checked. Not independent scientific confirmation.

Finding: Among 110 hemodialysis patients in two clinics, logistic-regression ML using clinic social-network data classified transplant attitudes more accurately than sociodemographic/clinical features alone (accuracy 65% vs 61%; combined data 74%). This is a small cross-sectional behavioral result. External replication and prospective utility were not established in the reviewed material.

Technology: Logistic-regression classifiers compare social-network, demographic and clinical features.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Logistic-regression classifiers compare social-network, demographic and clinical features.
Test and result
Among 110 hemodialysis patients in two clinics, logistic-regression ML using clinic social-network data classified transplant attitudes more accurately than sociodemographic/clinical features alone (accuracy 65% vs 61%; combined data 74%). This is a small cross-sectional behavioral result.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12882-022-03049-2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A novel deep neural network model based Xception and genetic algorithm for detection of COVID-19 from X-ray images.

Source check: Source checked. Not independent scientific confirmation.

Finding: The study optimized an Xception CNN with a genetic algorithm for COVID-19 classification on chest radiographs and reported held-out diagnostic performance on 277 images from a public 1,251-image collection. The study-specific task and test-set comparison support a retrospective diagnostic-validation result; they do not establish external clinical performance or patient benefit.

Technology: A genetic algorithm searches Xception CNN architecture/hyperparameters for classification of COVID-19, viral pneumonia, bacterial pneumonia and normal X-rays.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A genetic algorithm searches Xception CNN architecture/hyperparameters for classification of COVID-19, viral pneumonia, bacterial pneumonia and normal X-rays.
Test and result
For the public 1,251-image dataset (974 train/277 test), the paper reports accuracy .996 for binary, .989 for three-class and .924 for four-class classification, compared with cited prior models.
Limitation / still unknown
The 277-image held-out set includes 60 COVID-19 images and is compared with image labels/model baselines. This is retrospective validation on a public dataset, without external-site or prospective clinical evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9790088/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning-Based Automated Detection and Quantification of Geographic Atrophy and Hypertransmission Defects Using Spectral Domain Optical Coherence Tomography.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Automated deep-learning models used 900 OCT volumes/100,266 scans from 341 AMD patients to detect/segment geographic atrophy. B-scan detection accuracy 91%, area pixel accuracy 94%; en-face detection 82% and measurement 96%, with device-specific differences.

Technology: §§2.1–2.12: B-scan/en-face ML segmentation trained against expert-reviewed masks; automated detection and area quantification are the enabling operation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
§§2.1–2.12: B-scan/en-face ML segmentation trained against expert-reviewed masks; automated detection and area quantification are the enabling operation.
Limitation / still unknown
Performance was evaluated retrospectively on 341 patients and varied by OCT device; no prospective clinical-impact evaluation was reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2075-4426/13/1/37 · https://europepmc.org/article/MED/36675697

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Discovery of novel CSF biomarkers to predict progression in dementia using machine learning

Source check: Source checked. Not independent scientific confirmation.

Finding: The authors report group classification and candidate prioritization, while stating that no candidate is a strong individual predictor. The proteins are exploratory biomarkers; the source does not establish individual-level prognostic accuracy or clinical utility.

Technology: Random forests analyze cerebrospinal-fluid proteomics to classify fast versus slow dementia progression and prioritize TNFRSF4 and TGF-β1 candidates.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random forests analyze cerebrospinal-fluid proteomics to classify fast versus slow dementia progression and prioritize TNFRSF4 and TGF-β1 candidates.
Test and result
The authors report group classification and candidate prioritization, while stating that no candidate is a strong individual predictor.
Limitation / still unknown
The proteins are exploratory biomarkers; the source does not establish individual-level prognostic accuracy or clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-2382373/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

3D CT-Inclusive Deep-Learning Model to Predict Mortality, ICU Admittance, and Intubation in COVID-19 Patients.

Source check: Source checked. Not independent scientific confirmation.

Finding: Full Springer PDF: 3D CT model trained/evaluated on Italian hospital cohorts for mortality, ICU and intubation; external validation reported. CT-only mortality sensitivity was low (about 28%); adding clinical features raised performance, so the result is narrow and cohort-specific. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit.

Technology: Full Springer PDF: 3D CT model trained/evaluated on Italian hospital cohorts for mortality, ICU and intubation; external validation reported. CT-only mortality sensitivity was low (about 28%); adding clinical features raised performance, so the result is narrow and cohort-specific.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full Springer PDF: 3D CT model trained/evaluated on Italian hospital cohorts for mortality, ICU and intubation; external validation reported. CT-only mortality sensitivity was low (about 28%); adding clinical features raised performance, so the result is narrow and cohort-specific
Test and result
Full Springer PDF: 3D CT model trained/evaluated on Italian hospital cohorts for mortality, ICU and intubation; external validation reported. CT-only mortality sensitivity was low (about 28%); adding clinical features raised performance, so the result is narrow and cohort-specific
Limitation / still unknown
Full Springer PDF: 3D CT model trained/evaluated on Italian hospital cohorts for mortality, ICU and intubation; external validation reported. CT-only mortality sensitivity was low (about 28%); adding clinical features raised performance, so the result is narrow and cohort-specific
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s10278-022-00734-4 · https://link.springer.com/content/pdf/10.1007/s10278-022-00734-4.pdf · https://api.crossref.org/works/10.1007%2Fs10278-022-00734-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial Intelligence-Based Robust Hybrid Algorithm Design and Implementation for Real-Time Detection of Plant Diseases in Agricultural Environments.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A hybrid learned plant-disease detector was tested on field images and embedded in a UAV workflow across apple, grape and tomato sites, with reported class errors and runtime comparisons. This is a specific field detection result.

Technology: Publisher full text checked: Methods describe image training/evaluation and field implementation; Results report field observations and classification errors.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe image training/evaluation and field implementation; Results report field observations and classification errors.
Limitation / still unknown
Reported field images and UAV trials cover the studied crops and sites; the evidence does not establish performance across other farms or seasons.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2079-7737/11/12/1732

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning in renal pathology.

Source check: Source checked. Not independent scientific confirmation.

Finding: Image-derived biomarkers and pretrained deep-learning models were evaluated for diagnosing three named glomerular pathologies from renal-biopsy slides, with patient-level held-out testing. External replication and prospective utility were not established in the reviewed material.

Technology: Methods apply glomerular biomarker extraction and pretrained CNNs to PAS renal-biopsy images; Results report patient-held-out classification of MCD, MN and TBMN.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods apply glomerular biomarker extraction and pretrained CNNs to PAS renal-biopsy images; Results report patient-held-out classification of MCD, MN and TBMN.
Test and result
Image-derived biomarkers and pretrained deep-learning models were evaluated for diagnosing three named glomerular pathologies from renal-biopsy slides, with patient-level held-out testing.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10479587/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Screening for Biomarkers for Progression from Oral Leukoplakia to Oral Squamous Cell Carcinoma and Evaluation of Diagnostic Efficacy by Multiple Machine Learning Algorithms.

Source check: Source checked. Not independent scientific confirmation.

Finding: Gene-expression analyses identified candidate biomarkers during oral-leukoplakia progression, while machine-learning models evaluated diagnostic efficacy for carcinoma.

Technology: Methods describe progression samples and biomarker/model analysis; Results report candidate marker and diagnostic evaluation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe progression samples and biomarker/model analysis; Results report candidate marker and diagnostic evaluation.
Test and result
Methods describe progression samples and biomarker/model analysis; Results report candidate marker and diagnostic evaluation.
Limitation / still unknown
Methods describe progression samples and biomarker/model analysis; Results report candidate marker and diagnostic evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9738227/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Accurate Diagnosis and Survival Prediction of Bladder Cancer Using Deep Learning on Histological Slides.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Weakly supervised attention-MIL with ResNet50 features and autoencoder reduction distinguishes bladder cancer on slides and generates MIBC survival risk; internal TCGA and independent Renmin cohorts gave diagnostic external AUC .993 and prognostic external C-index .622. Retrospective with modest prognostic discrimination.

Technology: Weakly supervised attention-MIL with ResNet50 features and autoencoder reduction distinguishes bladder cancer on slides and generates MIBC survival risk; internal TCGA and independent Renmin cohorts gave diagnostic external AUC .9

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Weakly supervised attention-MIL with ResNet50 features and autoencoder reduction distinguishes bladder cancer on slides and generates MIBC survival risk; internal TCGA and independent Renmin cohorts gave diagnostic external AUC .993 and prognostic external C-index .622. Retrospective with modest prognostic discrimination.
Limitation / still unknown
Both diagnostic and prognostic evidence is retrospective; external prognostic discrimination was modest (C-index 0.622), and clinical benefit was not evaluated.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2072-6694/14/23/5807

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Research on Classification Method of Medical Ultrasound Image Processing Based on Neural Network.

Source check: Source checked. Not independent scientific confirmation.

Finding: Transfer-learned CNNs classified thyroid-nodule ultrasound images against pathology labels, a disease-specific diagnostic task.

Technology: Cascaded/transfer-learned CNNs process ultrasound images after image-quality and ROI filtering.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Cascaded/transfer-learned CNNs process ultrasound images after image-quality and ROI filtering.
Test and result
The Results section reports classification of thyroid nodular lesions and compares the candidate CNN models on held-out images.
Limitation / still unknown
Retrospective image collection with internal model comparisons; no independent-centre test or clinical benefit evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11390190/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC11390190/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC11390190/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Diagnosis of Depth of Submucosal Invasion in Colorectal Cancer with AI Using Deep Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: CNN trained on 706 endoscopic images/91 patients, evaluated on separate 394-image/49-patient set and prospective 56-patient test against pathology/endoscopists; retrospective sensitivity/specificity are imbalanced (87.2%/35.7%), so the result is a narrow diagnostic proof.

Technology: CNN trained on 706 endoscopic images/91 patients, evaluated on separate 394-image/49-patient set and prospective 56-patient test against pathology/endoscopists; retrospective sensitivity/specificity are imbalanced (87.2%/35.7%), s

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CNN trained on 706 endoscopic images/91 patients, evaluated on separate 394-image/49-patient set and prospective 56-patient test against pathology/endoscopists; retrospective sensitivity/specificity are imbalanced (87.2%/35.7%), so the result is a narrow diagnostic proof.
Limitation / still unknown
Although a prospective 56-patient test was included, sensitivity and specificity were imbalanced (87.2% and 35.7% in the reported retrospective evaluation); this does not establish improved clinical outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2072-6694/14/21/5361

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

NigraNet: An automatic framework to assess nigral neuromelanin content in early Parkinson's disease using convolutional neural network

Source check: Source checked. Not independent scientific confirmation.

Finding: The checked primary abstract reports this specific result: CNN framework (NigraNet) segments substantia nigra on NM-MRI in early PD with manual-reference validation, enabling a distinct neuromelanin quantification result in patients versus healthy volunteers. The abstract identifies the model or learned operation and an observed evaluation; the conclusion is limited to the reported study setting. Original PMC full-text Methods and Results were checked; date 2022-10-31 reconciled to the primary record. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: NigraNet

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher/PMC Methods establish the learned model and its operation in the specific study.
Test and result
The checked primary abstract reports this specific result: CNN framework (NigraNet) segments substantia nigra on NM-MRI in early PD with manual-reference validation, enabling a distinct neuromelanin quantification result in patients versus healthy volunteers. The abstract identifies the model or learned operation and an observed evaluation; the conclusion is limited to the reported study setting. Original PMC full-text Methods and Results were checked; date 2022-10-31 reconciled to the primary record.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9668659/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A deep learning model for CT-based kidney volume determination in dogs and normal reference definition.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: deep-learning CT segmentation/volume calculation on dogs, with groups by sex/neuter status; reference volumes and volume/bodyweight indices reported, no significant group differences for normalized indices. This supplies a veterinary imaging measurement/reference dataset, not evidence of disease causation.

Technology: AI volumetry supports a concrete normal canine kidney-volume reference result; retrospective sample and imaging protocol limit generalization.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
AI volumetry supports a concrete normal canine kidney-volume reference result; retrospective sample and imaging protocol limit generalization.
Limitation / still unknown
The study defines reference volumes in dogs and found no significant sex/neuter differences in normalized indices; it does not establish disease associations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2022.1011804/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Quantifying neuro-motor correlations during awake deep brain stimulation surgery using markerless tracking.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Markerless learned tracking was applied to awake deep-brain-stimulation recordings to measure neuro-motor correlations; the seven-subject study is exploratory.

Technology: The full text Methods were checked for the learned operation and its use in the stated field task.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The full text Methods were checked for the learned operation and its use in the stated field task.
Limitation / still unknown
Exploratory analysis involved seven subjects and does not establish generalizable neuro-motor relationships.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-022-21860-7

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Deep Learning-Aided Automated Method for Calculating Metabolic Tumor Volume in Diffuse Large B-Cell Lymphoma

Source check: Source checked. Not independent scientific confirmation.

Finding: A deep-learning method automated metabolic-tumor-volume measurement in diffuse large B-cell lymphoma PET/CT and was compared with expert reader assessment. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: Deep-learning tumor-volume segmentation

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe semiautomated reader reference and trial cohort; Results assess automated MTV measurement in 100 DLBCL patients.
Test and result
A deep-learning method automated metabolic-tumor-volume measurement in diffuse large B-cell lymphoma PET/CT and was compared with expert reader assessment.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9653575/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Novel Preoperative Prediction Model Based on Deep Learning to Predict Neoplasm T Staging and Grading in Patients with Upper Tract Urothelial Carcinoma.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Deep-learning models predicted pathological T stage and grade from preoperative data in 884 upper-tract urothelial carcinoma patients. A random 80/20 patient split assessed selected CNN-BiGRU/BiGRU models; AUCs ranged 0.760–0.824 across endpoints. This is a retrospective single-center diagnostic-prediction result.

Technology: Five deep-learning architectures compared; XGBoost screening and clinical feature selection preceded model fitting. Random train/test split 80/20; 884 pathologically confirmed patients. Best T-stage CNN-BiGRU AUC 0.760; grade AUC 0.804 (1973 WHO) and 0.824 (2004 WHO).

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Five deep-learning architectures compared; XGBoost screening and clinical feature selection preceded model fitting. Random train/test split 80/20; 884 pathologically confirmed patients. Best T-stage CNN-BiGRU AUC 0.760; grade AUC 0.804 (1973 WHO) and 0.824 (2004 WHO).
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcm/jcm-11-05815/article_deploy/jcm-11-05815.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

High-content phenotyping of Parkinson's disease patient stem cell-derived midbrain dopaminergic neurons using machine learning classification.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary Results and Experimental Procedures show image-derived multivariate phenotypes and classifiers separated LRRK2-mutant from isogenic control dopaminergic neurons and detected compound effects, with a second SNCA triplication model used for validation.

Technology: Results and Experimental Procedures show image-derived multivariate phenotypes and classifiers separated LRRK2-mutant from isogenic control dopaminergic neurons and detected compound effects, with a second SNCA triplication model used for validation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Results and Experimental Procedures show image-derived multivariate phenotypes and classifiers separated LRRK2-mutant from isogenic control dopaminergic neurons and detected compound effects, with a second SNCA triplication model used for validation.
Test and result
LDA, SVM and LightGBM used quantitative image features; the paper applied the trained classifier to independent SNCA-triplication neurons and to treated versus DMSO neurons.
Limitation / still unknown
The isogenic and second-model comparisons support in-vitro disease-model phenotyping; this does not establish patient benefit or a new clinical therapy.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9561636/ · https://europepmc.org/article/MED/36179692 · https://pmc.ncbi.nlm.nih.gov/articles/PMC9561636/#Results:

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of key biomarkers and signaling pathways and analysis of their association with immune cells in immunoglobulin A nephropathy.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Biomarkers and pathways were identified and checked with in-vitro experiments. The publisher Methods and Results were checked; this conclusion is limited to the study cohort, reference standard and validation described there.

Technology: Biomarkers and pathways were identified and checked with in-vitro experiments.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Biomarkers and pathways were identified and checked with in-vitro experiments.
Limitation / still unknown
The reported biomarker/pathway checks are limited to the study cohorts and in-vitro experiments; they do not establish causal disease mechanisms or clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://ceji.termedia.pl/Identification-of-key-biomarkers-and-signaling-pathways-and-analysis-of-their-association · https://ceji.termedia.pl/Identification-of-key-biomarkers-and-signaling-pathways-and-analysis-of-their-association,152711,0,2.html

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting futile recanalization, malignant cerebral edema, and cerebral herniation using intelligible ensemble machine learning following mechanical thrombectomy for acute ischemic stroke.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Ensemble models estimated risks of futile recanalization, malignant cerebral edema and herniation after thrombectomy in an acute-stroke cohort.

Technology: Methods combine clinical and NCCT features with ensemble ML; Results evaluate three post-thrombectomy outcomes against held-out data.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods combine clinical and NCCT features with ensemble ML; Results evaluate three post-thrombectomy outcomes against held-out data.
Limitation / still unknown
Risk estimates were derived from an acute-stroke thrombectomy cohort; no external validation or clinical-impact evaluation is stated in the checked rationale.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9554641/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The Predictive Role of Subcutaneous Adipose Tissue in the Pathogenesis of Obstructive Sleep Apnoea.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Deep-learning and random-forest models predict OSA severity from anthropometric, blood-test and ultrasound SAT measures in 100 patients; a stated 75:25 teaching/testing split and repeated analyses yielded >90% effectiveness after an apparent 97% in-sample classification. Single-center, small cohort; the paper notes SAT/AHI correlations were not significant, so the model is exploratory and does not establish causal pathogenesis or external clinical performance.

Technology: Methods §2.7 describes deep neural-network and random-forest classification using Deepnet/caret and variable-importance methods; Results apply them to airway-obstruction and OSA-category prediction.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Methods §2.7 describes deep neural-network and random-forest classification using Deepnet/caret and variable-importance methods; Results apply them to airway-obstruction and OSA-category prediction.
Limitation / still unknown
The study is a small, single-center cohort of 100 patients; SAT/AHI correlations were not significant, so prediction does not establish causal pathogenesis or external performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9605212/ · https://europepmc.org/article/MED/36294937

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Independent assessment of a deep learning system for lymph node metastasis detection on the Augmented Reality Microscope.

Source check: Source checked. Not independent scientific confirmation.

Finding: Independent evaluation of LYNA on 40 whole-slide images using the augmented-reality microscope reported an AUC near 0.98; this retrospective result is limited to the study setting. External replication and prospective utility were not established in the reviewed material.

Technology: LYNA deep-learning model assessed on whole-slide images with the augmented-reality microscope

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
LYNA deep-learning model assessed on whole-slide images with the augmented-reality microscope
Test and result
Independent evaluation of LYNA on 40 whole-slide images using the augmented-reality microscope reported an AUC near 0.98; this retrospective result is limited to the study setting.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9808066/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Automated Detection Of Enlarged Extraocular Muscle In Graves’ Ophthalmopathy With Computed Tomography And Deep Neural Network

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A specific clinical imaging result is tested against defined extraocular muscle enlargement in prospective cohort; supports narrow diagnostic validation. Do not infer care benefit. Research Square v1 date metadata not independently recovered; DOI history is later journal article 2022-09-22, so hold version date unresolved?

Technology: Journal primary source/paper: prospective observational CT study, 371 participants (199 Graves ophthalmopathy with enlarged extraocular muscles; 172 controls). Deep neural network trained/validated/tested on separated 222/74/75 images; test AUC .946, sensitivity 92.5%, specificity 88.6%. Clinical diagnostic finding is accurate CT identification under study criteria; not deployment/clinical benefit.

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Journal primary source/paper: prospective observational CT study, 371 participants (199 Graves ophthalmopathy with enlarged extraocular muscles; 172 controls). Deep neural network trained/validated/tested on separated 222/74/75 images; test AUC .946, sensitivity 92.5%, specificity 88.6%. Clinical diagnostic finding is accurate CT identification under study criteria; not deployment/clinical benefit.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-022-20279-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 4/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Incidence and Complications of Atrial Fibrillation in a Low Socioeconomic and High Disability United States (US) Population: A Combined Statistical and Machine Learning Approach.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: In US Medicaid data, statistical and machine-learning analyses grouped people by socioeconomic status and disability and compared those groups with atrial-fibrillation incidence and complications. The reported associations are observational and do not establish that socioeconomic or disability status causes AF.

Technology: Statistical clustering and machine-learning analyses characterized joint socioeconomic/disability groups and their AF incidence patterns.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Statistical clustering and machine-learning analyses characterized joint socioeconomic/disability groups and their AF incidence patterns.
Limitation / still unknown
These are observational associations in a socially disadvantaged US Medicaid population; they do not establish causal effects or generalizability to other populations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9448617/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Cooperation between artificial intelligence and endoscopists for diagnosing invasion depth of early gastric cancer

Source check: Source checked. Not independent scientific confirmation.

Finding: Full Springer PDF: AI classifier and three expert endoscopists evaluated invasion depth; a 100-case second-institution test and cooperative decision method were reported. AI+endoscopist F1 was .776, exceeding AI alone .768 and endoscopist majority .662 in reported comparison. This is bounded diagnostic evaluation. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit. This record is limited to the study result and does not establish independent replication or real-world benefit.

Technology: EfficientNetB1 CNN

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full Springer PDF: AI classifier and three expert endoscopists evaluated invasion depth; a 100-case second-institution test and cooperative decision method were reported. AI+endoscopist F1 was .776, exceeding AI alone .768 and endoscopist majority .662 in reported comparison. This is bounded diagnostic evaluation.
Test and result
Full Springer PDF: AI classifier and three expert endoscopists evaluated invasion depth; a 100-case second-institution test and cooperative decision method were reported. AI+endoscopist F1 was .776, exceeding AI alone .768 and endoscopist majority .662 in reported comparison. This is bounded diagnostic evaluation. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit.
Limitation / still unknown
The source-checked evidence does not establish broader external validity or prospective benefit. Affiliation was not available in the retrieved trusted primary metadata. Organization affiliation was not stated in the source material reviewed. Affiliation was not stated in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s10120-022-01330-9 · https://link.springer.com/content/pdf/10.1007/s10120-022-01330-9.pdf · https://api.crossref.org/works/10.1007%2Fs10120-022-01330-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Automated Detection of Myocardial Infarction and Heart Conduction Disorders Based on Feature Selection and a Deep Learning Model

Source check: Source checked. Not independent scientific confirmation.

Finding: On single-label PTB-XL records, the reported overall accuracy was 99.20% using an internal record split. No independent cohort or patient-level validation is reported; this is database-internal classification, not clinical deployment.

Technology: A 14-layer CNN extracts ECG features that feed a multiclass SVM for five diagnostic superclasses.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A 14-layer CNN extracts ECG features that feed a multiclass SVM for five diagnostic superclasses.
Test and result
On single-label PTB-XL records, the reported overall accuracy was 99.20% using an internal record split.
Limitation / still unknown
No independent cohort or patient-level validation is reported; this is database-internal classification, not clinical deployment.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/1424-8220/22/17/6503

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Automatized lung disease quantification in patients with COVID-19 as a predictive tool to assess hospitalization severity.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: retrospective single-center 476 COVID-19 patients; CACOVID-CT quantifies affected lung area and CT severity score, correlated with length of stay, ICU admission, ventilation and death. Internal observational evidence only.

Technology: AI quantification is tied to patient outcomes in a defined clinical study; no causal or clinical-impact claim.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
AI quantification is tied to patient outcomes in a defined clinical study; no causal or clinical-impact claim.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2022.930055/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prehospital diagnostic algorithm for acute coronary syndrome using machine learning: a prospective observational study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A prospectively observed prehospital model used symptoms, vital signs and three-lead ECG to predict acute-coronary-syndrome diagnoses; it supports diagnostic prediction, not improved outcomes.

Technology: The full text Methods were checked for the learned operation and its use in the stated field task.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
The full text Methods were checked for the learned operation and its use in the stated field task.
Limitation / still unknown
This prospective observational evaluation supports diagnostic prediction only; it does not show improved prehospital decisions or patient outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-022-18650-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Designing a new fast solution to control isolation rooms in hospitals depending on artificial intelligence decision.

Source check: Source checked. Not independent scientific confirmation.

Finding: Full Methods and Results describe an MLP trained on 45 hospital cases (31 train, 7 validation, 7 test) using time, cost, facility and space inputs to choose prefabricated isolation units versus retrofit. Reported 43% time reduction/78% lower average cost is tied to described implementation comparison at 45 hospitals, though measurement and external comparator details are limited; preserve that scope. The paper appeared in Biomedical Signal Processing and Control. Organization is unavailable in the deposited bibliographic metadata.

Technology: Multilayer perceptron for isolation-room selection

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Full Methods and Results describe an MLP trained on 45 hospital cases (31 train, 7 validation, 7 test) using time, cost, facility and space inputs to choose prefabricated isolation units versus retrofit. Reported 43% time reduction/78% lower average cost is tied to described implementation comparison at 45 hospitals, though measurement and external comparator details are limited; preserve that scope.
Test and result
Full Methods and Results describe an MLP trained on 45 hospital cases (31 train, 7 validation, 7 test) using time, cost, facility and space inputs to choose prefabricated isolation units versus retrofit. Reported 43% time reduction/78% lower average cost is tied to described implementation comparison at 45 hospitals, though measurement and external comparator details are limited; preserve that scope.
Limitation / still unknown
Reported 43% time reduction/78% lower average cost is tied to described implementation comparison at 45 hospitals, though measurement and external comparator details are limited; preserve that scope.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9412665/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Application of machine learning algorithms in predicting HIV infection among men who have sex with men: Model development and validation

Source check: Source checked. Not independent scientific confirmation.

Finding: The study used 6,346 development records and a prospective 2020 validation cohort to assess screening prediction. The result supports prediction in this surveillance setting, not infection prevention, causal inference or demonstrated clinical benefit; a locked-model evaluation is not established in the reviewed record.

Technology: A random forest predicts HIV infection status from Zhejiang sentinel-surveillance data among men who have sex with men.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A random forest predicts HIV infection status from Zhejiang sentinel-surveillance data among men who have sex with men.
Test and result
The study used 6,346 development records and a prospective 2020 validation cohort to assess screening prediction.
Limitation / still unknown
The result supports prediction in this surveillance setting, not infection prevention, causal inference or demonstrated clinical benefit; a locked-model evaluation is not established in the reviewed record.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/articles/PMC9452878

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Insulator Leakage Current Prediction Using Hybrid of Particle Swarm Optimization and Gene Algorithm-Based Neural Network and Surface Spark Discharge Data

Source check: Source checked. Not independent scientific confirmation.

Finding: A PSO/GA-optimized neural network predicted leakage current from weather and surface-discharge signals on monitored high-voltage towers, a specific grid-monitoring result. Time-separated training/test data from the monitored system and algorithm comparisons; no independent utility or long-term deployment outcome is shown.

Technology: ANN inputs include tower weather measurements and image-derived spark-discharge features; PSO/GA searches optimize the network.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ANN inputs include tower weather measurements and image-derived spark-discharge features; PSO/GA searches optimize the network.
Test and result
The paper compares predicted and measured leakage current on 69-kV/161-kV tower data and reports improved error over baseline networks.
Limitation / still unknown
Time-separated training/test data from the monitored system and algorithm comparisons; no independent utility or long-term deployment outcome is shown. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9436545/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC9436545/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC9436545/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Improves the Prediction Rate of Non-Curative Resection of Endoscopic Submucosal Dissection in Patients with Early Gastric Cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine learning predicted non-curative resection before endoscopic submucosal dissection for early gastric cancer, using internal and external data.

Technology: Methods describe the 4,972-patient outcome and model evaluation; Results report non-curative resection prediction.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe the 4,972-patient outcome and model evaluation; Results report non-curative resection prediction.
Test and result
Methods describe the 4,972-patient outcome and model evaluation; Results report non-curative resection prediction.
Limitation / still unknown
Methods describe the 4,972-patient outcome and model evaluation; Results report non-curative resection prediction.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9367410/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Automatic text classification of actionable radiology reports of tinnitus patients using bidirectional encoder representations from transformer (BERT) and in-domain pre-training (IDPT).

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A BERT classifier with in-domain pretraining sorted tinnitus-related radiology reports by whether they contained actionable findings, supporting follow-up prioritization in the evaluated health-system records. This is a retrospective text-classification result, not evidence of improved patient outcomes.

Technology: BERT with in-domain pretraining classified radiology-report text for actionable findings.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
BERT with in-domain pretraining classified radiology-report text for actionable findings.
Limitation / still unknown
Retrospective classification in the studied health-system reports does not show that automated triage improved follow-up or patient outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9338483/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

ViSTA: A Novel Network Improving Lung Adenocarcinoma Invasiveness Prediction from Follow-Up CT Series.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: ViSTA used serial CT with temporal attention to predict invasiveness of early lung adenocarcinoma nodules, compared with radiologists and baseline methods.

Technology: follow-up CT series and counterpart models; Results compare invasiveness predictions with radiologists.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
follow-up CT series and counterpart models; Results compare invasiveness predictions with radiologists.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9367560/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deployment and validation of an AI system for detecting abnormal chest radiographs in clinical settings.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The deployed VinDr-CXR system was compared with reports at Phu Tho General Hospital; it achieved F1 0.653 (95% CI 0.635–0.671), accuracy 79.6%, sensitivity 68.6% and specificity 83.9% for any abnormality. This establishes real-world performance with a substantial drop from laboratory results.

Technology: CAD combines convolutional image models for abnormality detection/classification within the hospital PACS workflow.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CAD combines convolutional image models for abnormality detection/classification within the hospital PACS workflow.
Limitation / still unknown
Deployment was assessed at one hospital against local reports; performance was lower than laboratory results and does not establish performance at other sites or improved outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2022.890759/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting the oxidation states of Mn ions in the oxygen-evolving complex of photosystem II using supervised and unsupervised machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary source Methods/Results confirm the represented result: Decision tree and k-means models trained on Mn crystallographic compounds assign oxidation states to PSII oxygen-evolving complex structures, a specific structural-biology result with the AI role identified; evaluation compares against XFEL and CryoEM structures.

Technology: Decision tree and k-means models trained on Mn crystallographic compounds assign oxidation states to PSII oxygen-evolving complex structures, a specific structural-biology result with the AI role identified; evaluation compares against XFEL and CryoEM structures.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Decision tree and k-means models trained on Mn crystallographic compounds assign oxidation states to PSII oxygen-evolving complex structures, a specific structural-biology result with the AI role identified; evaluation compares against XFEL and CryoEM structures.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11120-022-00941-8

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Visual Transformer and Deep CNN Prediction of High-risk COVID-19 Infected Patients using Fusion of CT Images and Clinical Data

Source check: Source checked. Not independent scientific confirmation.

Finding: A 3D Swin Transformer fused 64-section CT scans with clinical admission data to classify high-risk COVID-19 outcome among 380 diagnosed inpatients. Five-fold validation compared CT-only, clinical-only and fused models; fusion improved performance, with reported TPR 0.95, FPR 0.40 and AUC 0.77. Retrospective single-hospital data and false positives limit clinical claims.

Technology: A 3D Swin Transformer fused 64-section CT scans with clinical admission data to classify high-risk COVID-19 outcome among 380 diagnosed inpatients.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A 3D Swin Transformer fused 64-section CT scans with clinical admission data to classify high-risk COVID-19 outcome among 380 diagnosed inpatients.
Test and result
A 3D Swin Transformer fused 64-section CT scans with clinical admission data to classify high-risk COVID-19 outcome among 380 diagnosed inpatients.
Limitation / still unknown
Five-fold validation compared CT-only, clinical-only and fused models; fusion improved performance, with reported TPR 0.95, FPR 0.40 and AUC 0.77. Retrospective single-hospital data and false positives limit clinical claims. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.medrxiv.org/content/10.1101/2022.07.26.22278084v1.full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Based on MRI DWI Radiomics Features for Prognostic Prediction in Nasopharyngeal Carcinoma.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Random forest selects DWI radiomic features and fits recurrence/metastasis prognosis models. Results reports validation-set AUCs, highest 0.80 for DWI+ADC; cohort-specific prognostic estimate.

Technology: Random forest selects DWI radiomic features and fits recurrence/metastasis prognosis models.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random forest selects DWI radiomic features and fits recurrence/metastasis prognosis models.
Limitation / still unknown
The prognostic estimates are cohort-specific and retrospective; the best reported validation AUC was 0.80, without prospective impact evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/articles/PMC9264891

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The Application of Artificial Intelligence in Predicting Embryo Transfer Outcome of Recurrent Implantation Failure.

Source check: Source checked. Not independent scientific confirmation.

Finding: A machine-learning system predicts embryo-transfer outcome and transfer-number decisions in recurrent implantation failure using a retrospective dataset; held-out/cross-validation metrics support a narrowly scoped clinical outcome prediction result. External replication and prospective utility were not established in the reviewed material.

Technology: AI models use clinical/embryology variables to predict embryo-transfer outcome.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
AI models use clinical/embryology variables to predict embryo-transfer outcome.
Test and result
A machine-learning system predicts embryo-transfer outcome and transfer-number decisions in recurrent implantation failure using a retrospective dataset; held-out/cross-validation metrics support a narrowly scoped clinical outcome prediction result.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2022.885661/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A novel AI device for real-time optical characterization of colorectal polyps.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary publisher Methods/Results support the AI-enabled result: Standalone CADx device classifies colorectal polyps in real time on prospective colonoscopy videos against blinded histology-based reference, with a clear AI enabling role. Organization is unavailable in the deposited bibliographic metadata.

Technology: Standalone computer-aided polyp characterization device

Publication: Publication status unspecified in annotation · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Methods establish the case-specific AI operation: Standalone CADx device classifies colorectal polyps in real time on prospective colonoscopy videos against blinded histology-based reference, with a clear AI enabling role.
Test and result
The primary publisher Methods/Results support the AI-enabled result: Standalone CADx device classifies colorectal polyps in real time on prospective colonoscopy videos against blinded histology-based reference, with a clear AI enabling role.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41746-022-00633-6

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Strength Predictive Modelling of Soils Treated with Calcium-Based Additives Blended with Eco-Friendly Pozzolans-A Machine Learning Approach

Source check: Source checked. Not independent scientific confirmation.

Finding: On 392 literature-compiled stabilized-soil samples, the study reports classification accuracy of 0.920 and regression R² around 0.900. Evaluation used cross-validation and a held-out split from the compiled data; independent experimental or field validation is not reported.

Technology: Gradient-boosting models map stabilizer/pozzolan composition and curing inputs to unconfined soil-strength values and classes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Gradient-boosting models map stabilizer/pozzolan composition and curing inputs to unconfined soil-strength values and classes.
Test and result
On 392 literature-compiled stabilized-soil samples, the study reports classification accuracy of 0.920 and regression R² around 0.900.
Limitation / still unknown
Evaluation used cross-validation and a held-out split from the compiled data; independent experimental or field validation is not reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/1996-1944/15/13/4575

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Predictor of Immune Checkpoint Blockade Response in Gastric Cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Machine-learning models analyzed gastric-cancer data to predict immune-checkpoint-blockade response and identify candidate response-associated genes.

Technology: responder classification and feature selection; Results report response signature performance and gene associations.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
responder classification and feature selection; Results report response signature performance and gene associations.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9265060/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Pre- and post-dam river water temperature alteration prediction using advanced machine learning models.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The original full text supports the bounded AI-enabled finding: ML models estimate daily water temperature at Yangtze River Cuntan before/after dam construction; the study-specific output addresses measured hydrologic temperature alteration at that site.

Technology: ML models estimate daily water temperature at Yangtze River Cuntan before/after dam construction; the study-specific output addresses measured hydrologic temperature alteration at that site.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ML models estimate daily water temperature at Yangtze River Cuntan before/after dam construction; the study-specific output addresses measured hydrologic temperature alteration at that site.
Limitation / still unknown
The prediction concerns one Yangtze River site and its historical pre/post-dam records; it does not establish causal attribution to the dam or transfer to other rivers.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9244425/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Classification of Toona sinensis Young Leaves Using Machine Learning and UAV-Borne Hyperspectral Imagery

Source check: Source checked. Not independent scientific confirmation.

Finding: UAV-borne hyperspectral imaging and SVM distinguish young from old Toona sinensis leaves; the study reports held-out classification and feature analysis, supporting a specific crop-quality phenotyping result. Held-out validation and wavelength-feature analysis are reported in original Methods/Results.

Technology: SVM classifies leaf age/growth class from UAV hyperspectral signatures.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
SVM classifies leaf age/growth class from UAV hyperspectral signatures.
Test and result
UAV-borne hyperspectral imaging and SVM distinguish young from old Toona sinensis leaves; the study reports held-out classification and feature analysis, supporting a specific crop-quality phenotyping result.
Limitation / still unknown
Held-out validation and wavelength-feature analysis are reported in original Methods/Results. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.940327/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial Intelligence-Based Prognostic Model for Urologic Cancers: A SEER-Based Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: A recurrent neural-network risk model estimated cancer-specific mortality across six named urologic cancer groups using SEER data and a held-out test set. Organization is unavailable in the deposited bibliographic metadata.

Technology: Recurrent neural network for cancer-specific mortality risk

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe SEER cohort, recurrent model and test set; Results report risk stratification across urologic cancers.
Test and result
A recurrent neural-network risk model estimated cancer-specific mortality across six named urologic cancer groups using SEER data and a held-out test set.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9264864/

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Depression in Adolescents Using Mobile and Wearable Sensors: Multimodal Machine Learning-Based Exploratory Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Mobile and wearable sensor models explored prediction of adolescent depression scores and changes; the study is explicitly exploratory and limited to its small cohort.

Technology: sensor modalities, outcomes and cross-validation; Results compare performance across feature/model combinations.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
sensor modalities, outcomes and cross-validation; Results compare performance across feature/model combinations.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9270714/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Recognition Method of Soybean Leaf Diseases Based on an Improved Deep Learning Model.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: An attention-based residual network classifies three soybean leaf diseases on field images; held-out recognition and speed comparisons are reported against ResNet18. Claim is limited to image-based disease recognition, not field-level yield or treatment benefit.

Technology: Residual attention network identifies disease categories from soybean leaf images.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Residual attention network identifies disease categories from soybean leaf images.
Limitation / still unknown
Original paper reports train/test recognition metrics and comparison with ResNet18.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.878834/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Accuracy of Machine Learning Using the Montreal Cognitive Assessment for the Diagnosis of Cognitive Impairment in Parkinson's Disease.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: MoCA domain scores from 397 Parkinson’s patients (2,069 visits) were classified against comprehensive neuropsychological cognitive status; ML is compared with threshold rules. This supports a bounded diagnostic-performance result in an established cohort, not general-population screening.

Technology: Logistic regression, SVM and random forest use MoCA domain scores to classify PD cognitive impairment.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Logistic regression, SVM and random forest use MoCA domain scores to classify PD cognitive impairment.
Limitation / still unknown
The result is limited to 397 Parkinson’s patients (2,069 visits) and does not establish general-population screening performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.e-jmd.org/journal/view.php

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Association of hypoglossal nerve stimulator response with machine learning identified negative effort dependence patterns.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The original full text supports the bounded AI-enabled finding: The model identifies negative-effort-dependence patterns in inspiratory nasal-pressure waveforms and the study relates the patterns to hypoglossal-nerve-stimulator response in 41 patients.

Technology: The model identifies negative-effort-dependence patterns in inspiratory nasal-pressure waveforms and the study relates the patterns to hypoglossal-nerve-stimulator response in 41 patients.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The model identifies negative-effort-dependence patterns in inspiratory nasal-pressure waveforms and the study relates the patterns to hypoglossal-nerve-stimulator response in 41 patients.
Limitation / still unknown
The reported cohort or experimental dataset, comparator/reference and validation are specific to this case; conclusions remain bounded to that setting.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9136201/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting EGFR mutation, ALK rearrangement, and uncommon EGFR mutation in NSCLC patients by driverless artificial intelligence: a cohort study

Source check: Source checked. Not independent scientific confirmation.

Finding: A stacked deep/machine-learning ensemble predicts EGFR and ALK status against molecular reference in a consecutive test cohort; single-center design and reported test accuracy delimit generalization. Methods/Results: 1,089 retrospective training cases and 1,464 consecutive test cases; stacked ensemble of one DL and five ML models; test AUC .883 EGFR and .921 ALK, plus common/uncommon EGFR accuracy .83. Publisher article details and testing results checked.

Technology: 1,089 retrospective training cases and 1,464 consecutive test cases; stacked ensemble of one DL and five ML models; test AUC .883 EGFR and .921 ALK, plus common/uncommon EGFR accuracy .83. Publisher article details and testing results checked.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
1,089 retrospective training cases and 1,464 consecutive test cases; stacked ensemble of one DL and five ML models; test AUC .883 EGFR and .921 ALK, plus common/uncommon EGFR accuracy .83. Publisher article details and testing results checked.
Test and result
1,089 retrospective training cases and 1,464 consecutive test cases; stacked ensemble of one DL and five ML models; test AUC .883 EGFR and .921 ALK, plus common/uncommon EGFR accuracy .83. Publisher article details and testing results checked.
Limitation / still unknown
Methods/Results: 1,089 retrospective training cases and 1,464 consecutive test cases; stacked ensemble of one DL and five ML models; test AUC .883 EGFR and .921 ALK, plus common/uncommon EGFR accuracy .83. Publisher article details and testing results checked. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12931-022-02053-2

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Forecasting Patient Demand at Urgent Care Clinics using Machine Learning

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: statistical/ML ensemble forecasting improved 90-day patient-demand estimates for two Auckland urgent-care clinics on 2011–2021 data; later journal results refine the improvement range, and no deployment or patient-outcome effect is shown.

Technology: Eight statistical/ML algorithms and ensembles evaluated on two clinics’ daily arrivals, with XAI analysis.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Eight statistical/ML algorithms and ensembles evaluated on two clinics’ daily arrivals, with XAI analysis.
Limitation / still unknown
Forecasting used 2011–2021 records from two Auckland urgent-care clinics; no deployment or patient-outcome effect was shown, and later journal results refined the improvement range.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2205.13067 · https://doi.org/10.1049/cit2.12258 · https://mro.massey.ac.nz/server/api/core/bitstreams/5f41c75b-f4ef-4ec8-a242-6f84477addbd/content

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning from multiple experts improves identification of amyloid neuropathologies.

Source check: Source checked. Not independent scientific confirmation.

Finding: CNN ensembles combined five expert annotation schemes to identify amyloid neuropathologies, with held-out evaluation and a prospective blind research filter assessment.

Technology: Methods describe whole-slide image train/holdout split, four-fold models and expert ensembles; Results and prospective phase-two evaluate identification of cored plaques and cerebral amyloid angiopathy.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe whole-slide image train/holdout split, four-fold models and expert ensembles; Results and prospective phase-two evaluate identification of cored plaques and cerebral amyloid angiopathy.
Test and result
Methods describe whole-slide image train/holdout split, four-fold models and expert ensembles; Results and prospective phase-two evaluate identification of cored plaques and cerebral amyloid angiopathy.
Limitation / still unknown
Methods describe whole-slide image train/holdout split, four-fold models and expert ensembles; Results and prospective phase-two evaluate identification of cored plaques and cerebral amyloid angiopathy.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9052651/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Three-Dimensional Reconstruction of a CT Image under Deep Learning Algorithm to Evaluate the Application of Percutaneous Kyphoplasty in Osteoporotic Thoracolumbar Compression Fractures.

Source check: Source checked. Not independent scientific confirmation.

Finding: An improved V-Net reconstructed CT vertebrae for measuring height and kyphosis in a specific PKP-versus-PVP study of osteoporotic compression fractures. External replication and prospective utility were not established in the reviewed material.

Technology: The learned V-Net segments/reconstructs CT vertebrae; Dice and Hausdorff metrics compare the improved network with U-Net, V-Net, and CNN.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The learned V-Net segments/reconstructs CT vertebrae; Dice and Hausdorff metrics compare the improved network with U-Net, V-Net, and CNN.
Test and result
An improved V-Net reconstructed CT vertebrae for measuring height and kyphosis in a specific PKP-versus-PVP study of osteoporotic compression fractures.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9290755/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Severe Disease in Patients With Recent-Onset Psoriatic Arthritis. Prediction Model Based on Machine Learning

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine learning estimated severity in people with recent-onset psoriatic arthritis using data from the two-year REAPSER follow-up.

Technology: Learned prediction addresses a specific prospective clinical cohort outcome; cross-validation only.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned prediction addresses a specific prospective clinical cohort outcome; cross-validation only.
Test and result
The model was assessed by five-fold cross-validation in 158 participants from a prospectively collected cohort.
Limitation / still unknown
The source did not report independent validation or prospective testing of the fitted model.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2022.891863/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning for passive mental health symptom prediction: Generalization across different longitudinal mobile sensing studies.

Source check: Source checked. Not independent scientific confirmation.

Finding: Combined longitudinal sensing data improved held-out sleep/stress prediction in several study-target pairs, with heterogeneous cross-study effects. The paper appeared in PLOS ONE. Organization is unavailable in the deposited bibliographic metadata.

Technology: Machine-learning models for passive mobile-sensing data

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning models predict self-reported sleep/stress from passive mobile sensing features across two longitudinal datasets.
Test and result
Combined longitudinal sensing data improved held-out sleep/stress prediction in several study-target pairs, with heterogeneous cross-study effects.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0266516

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Hybrid Deep Learning Approach for Automatic Detection in Musculoskeletal Radiographs

Source check: Source checked. Not independent scientific confirmation.

Finding: The model is evaluated on a held-out finger-radiograph image set. The claim is limited to the evaluated image set and task, not general musculoskeletal diagnosis or clinical deployment.

Technology: A hybrid deep-learning pipeline classifies musculoskeletal radiographs as normal or abnormal.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A hybrid deep-learning pipeline classifies musculoskeletal radiographs as normal or abnormal.
Test and result
The model is evaluated on a held-out finger-radiograph image set.
Limitation / still unknown
The claim is limited to the evaluated image set and task, not general musculoskeletal diagnosis or clinical deployment.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9138246/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Research on Rice Yield Prediction Model Based on Deep Learning

Source check: Source checked. Not independent scientific confirmation.

Finding: Original source reports a concrete field-yield prediction result from measured rice plots using learned regression over rice-ear image and seed traits. The title overstates deep learning: the decisive model is regression, not a demonstrated deep network. The result is a bounded retrospective plot prediction, not a general yield claim. Methods: 207 paddy plots; 179 effective plot records; ear-area imagery and seed traits used as predictors for measured plot yield. Linear/nonlinear regression evaluates the relationship; OPCA power model R²=0.513 and multi-feature adjusted R²=0.456. Field-specific measured outcome supports eligibility, with limited cohort and data-on-request caveats.

Technology: 207 paddy plots; 179 effective plot records; ear-area imagery and seed traits used as predictors for measured plot yield. Linear/nonlinear regression evaluates the relationship; OPCA power model R²=0.513 and multi-feature adjusted R²=0.456. Field-specific measured outcome supports eligibility, with limited cohort and data-on-request caveats.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
207 paddy plots; 179 effective plot records; ear-area imagery and seed traits used as predictors for measured plot yield. Linear/nonlinear regression evaluates the relationship; OPCA power model R²=0.513 and multi-feature adjusted R²=0.456. Field-specific measured outcome supports eligibility, with limited cohort and data-on-request caveats.
Test and result
207 paddy plots; 179 effective plot records; ear-area imagery and seed traits used as predictors for measured plot yield. Linear/nonlinear regression evaluates the relationship; OPCA power model R²=0.513 and multi-feature adjusted R²=0.456. Field-specific measured outcome supports eligibility, with limited cohort and data-on-request caveats.
Limitation / still unknown
Methods: 207 paddy plots; 179 effective plot records; ear-area imagery and seed traits used as predictors for measured plot yield. Linear/nonlinear regression evaluates the relationship; OPCA power model R²=0.513 and multi-feature adjusted R²=0.456. Field-specific measured outcome supports eligibility, with limited cohort and data-on-request caveats. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://onlinelibrary.wiley.com/doi/10.1155/2022/1922561

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Meibomian Gland Density: An Effective Evaluation Index of Meibomian Gland Dysfunction Based on Deep Learning and Transfer Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Deep-learning segmentation enabled a quantitative meibomian-gland density index, which the authors assessed against clinical meibomian-gland dysfunction status in 85 subjects. The reported AUC was 0.900 for total-eyelid density; the small diagnostic cohort and reuse of earlier training subjects limit independence.

Technology: Mask R-CNN segmented tarsus and ResNet50-U-Net with transfer learning segmented glands; 60 annotated training/validation subjects included 40 training images and 20 upper-lid validation images, while 85 subjects (53 MGD, 32 contro

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Mask R-CNN segmented tarsus and ResNet50-U-Net with transfer learning segmented glands; 60 annotated training/validation subjects included 40 training images and 20 upper-lid validation images, while 85 subjects (53 MGD, 32 controls) were used for clinical efficacy analysis. Density AUC 0.900 for total eyelids. Training overlap with prior report disclosed.
Limitation / still unknown
The diagnostic cohort included 85 subjects and reused earlier training subjects, limiting independence and generalization.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcm/jcm-11-02396/article_deploy/jcm-11-02396.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction of Breast Cancer Histological Outcome by Radiomics and Artificial Intelligence Analysis in Contrast-Enhanced Mammography.

Source check: Source checked. Not independent scientific confirmation.

Finding: Contrast-enhanced mammography radiomics with machine learning predicted pathology-confirmed malignancy and histological features in breast lesions. External replication and prospective utility were not established in the reviewed material.

Technology: The study uses a the machine-learning model evaluated in this study for the stated prediction of breast cancer histological outcome by radiomics and artificial intelligence analysis in contrast-enhanced mammography task.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study uses a the machine-learning model evaluated in this study for the stated prediction of breast cancer histological outcome by radiomics and artificial intelligence analysis in contrast-enhanced mammography task.
Test and result
Contrast-enhanced mammography radiomics with machine learning predicted pathology-confirmed malignancy and histological features in breast lesions.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9102628/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Psychological Stress Identification and Evaluation Method Based on Mobile Human-Computer Interaction Equipment

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary full text reports stress identification from the collected mobile-interaction data and its study evaluation. The result is limited to the measured study setting; real-world longitudinal performance and clinical utility are not established.

Technology: Mobile human-computer interaction signals are used to identify psychological stress in a measured study.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Mobile human-computer interaction signals are used to identify psychological stress in a measured study.
Test and result
The primary full text reports stress identification from the collected mobile-interaction data and its study evaluation.
Limitation / still unknown
The result is limited to the measured study setting; real-world longitudinal performance and clinical utility are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9061069/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-coupled combinatorial mutagenesis enables resource-efficient engineering of CRISPR-Cas9 genome editor activities.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary study couples machine-learning-guided combinatorial mutagenesis with experimental screening and structural/wet-lab validation to engineer Cas9 variants with increased activity. This is a specific genome-editor engineering result, with the assays and cell contexts bounding the claim.

Technology: PMC full Methods/Results: candidate search reduced by about 95%, top-hit enrichment about 7.5×, followed by validation of N888R/A889Q KKH-SaCas9 and derived base editor in human cells. First publication date from PMC 2022-04-25.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
PMC full Methods/Results: candidate search reduced by about 95%, top-hit enrichment about 7.5×, followed by validation of N888R/A889Q KKH-SaCas9 and derived base editor in human cells. First publication date from PMC 2022-04-25.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9039034/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

GENESIS: Gene-Specific Machine Learning Models for Variants of Uncertain Significance Found in Catecholaminergic Polymorphic Ventricular Tachycardia and Long QT Syndrome-Associated Genes

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary Methods show gene-specific LR/MLP models trained with 10-fold CV on curated pathogenic/benign variants for RYR2, KCNQ1, KCNH2 and SCN5A, then applied to ClinVar/exome VUS. Results report model comparisons, blinded KCNQ1 patch-clamp tests (including a discordant V106D prediction), prospective family cosegregation for G245R/V241I, and blinded retrospective clinical correlation of 9/10 cases. Admit narrowly as an AI-enabled channelopathy variant-interpretation finding; it is not a prospective clinical utility trial. Validation is reported in the original source; its scope and limitations are stated in the decision rationale.

Technology: The learned/optimization operation and its relationship to the represented result are stated in the original source.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The learned/optimization operation and its relationship to the represented result are stated in the original source.
Test and result
Original Results establish the stated outcome and scope.
Limitation / still unknown
Validation is reported in the original source; its scope and limitations are stated in the decision rationale. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.ahajournals.org/doi/pdf/10.1161/CIRCEP.121.010326

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A prospective multicenter validation study of a machine learning algorithm classifier on quantitative electroencephalogram for differentiating between dementia with Lewy bodies and Alzheimer's dementia.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A previously developed EEG machine-learning classifier was prospectively evaluated across Japanese centers to distinguish dementia with Lewy bodies from Alzheimer’s disease, providing multicenter external diagnostic validation.

Technology: The source describes the learned operation and its role in producing the specific finding.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
The source describes the learned operation and its role in producing the specific finding.
Limitation / still unknown
This was prospective multicenter external diagnostic validation of an established classifier; it does not show improved care or patient outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1371/journal.pone.0265484 · https://europepmc.org/article/MED/35358240 · https://europepmc.org/articles/PMC8970386 · https://www.ncbi.nlm.nih.gov/research/bionlp/RESTful/pmcoa.cgi/BioC_json/PMC8970386/unicode

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Bridging clinic and wildlife care with AI-powered pan-species computational pathology

Source check: Source checked. Not independent scientific confirmation.

Finding: The first result-bearing preprint reports transfer of a human-lung pathology CNN to veterinary specimens and a canine melanoma prognostic association. The claim is limited to the tested species and small veterinary cohorts; the prostate subgroup trend was nonsignificant and this does not establish broad wildlife or clinical benefit. Full indexed Research Square v1 PDF, Methods/Results and Nature/PMC record checked. Version-of-record statement links the same study; preprint posted 2022-03-29, before journal version 2023-04-26. The study reports 14,570 expert annotations and prognostic analysis in 100 canine cases (melanoma n=88, prostate n=12).

Technology: Full indexed Research Square v1 PDF, Methods/Results and Nature/PMC record checked. Version-of-record statement links the same study; preprint posted 2022-03-29, before journal version 2023-04-26. The study reports 14,570 expert annotations and prognostic analysis in 100 canine cases (melanoma n=88, prostate n=12).

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full indexed Research Square v1 PDF, Methods/Results and Nature/PMC record checked. Version-of-record statement links the same study; preprint posted 2022-03-29, before journal version 2023-04-26. The study reports 14,570 expert annotations and prognostic analysis in 100 canine cases (melanoma n=88, prostate n=12).
Test and result
Full indexed Research Square v1 PDF, Methods/Results and Nature/PMC record checked. Version-of-record statement links the same study; preprint posted 2022-03-29, before journal version 2023-04-26. The study reports 14,570 expert annotations and prognostic analysis in 100 canine cases (melanoma n=88, prostate n=12).
Limitation / still unknown
Full indexed Research Square v1 PDF, Methods/Results and Nature/PMC record checked. Version-of-record statement links the same study; preprint posted 2022-03-29, before journal version 2023-04-26. The study reports 14,570 expert annotations and prognostic analysis in 100 canine cases (melanoma n=88, prostate n=12). The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://assets-eu.researchsquare.com/files/rs-1474064/v1/d6c26e75-387d-46f3-a23d-b1e21b9616bd.pdf?c=1682579356

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Diagnosis of COVID-19 Cases from Chest X-ray Images Using Deep Neural Network and LightGBM

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: DenseNet169 features, ANOVA selection, and LightGBM classification report COVID-19 chest-X-ray discrimination on a small public image dataset; no clinical deployment claim.

Technology: DenseNet169 feature extraction, ANOVA selection, LightGBM classification.

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
DenseNet169 feature extraction, ANOVA selection, LightGBM classification.
Limitation / still unknown
The result comes from a small public image dataset and does not establish clinical deployment or real-world diagnostic utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2203.14275 · https://arxiv.org/pdf/2203.14275

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Structural Health Monitoring of Dams Based on Acoustic Monitoring, Deep Neural Networks, Fuzzy Logic and a CUSUM Control Algorithm

Source check: Source checked. Not independent scientific confirmation.

Finding: The framework identified structural-state changes; signal mobility/complexity tracked model homogeneity and activity correlated with water flow. Evidence comes from a scale-model experiment, not a full-size operating dam; threshold tuning and normal-state data remain limitations described by the authors.

Technology: A deep autoencoder models Hjorth acoustic features, with CUSUM and fuzzy logic detecting experimental changes in a scale-model dam.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A deep autoencoder models Hjorth acoustic features, with CUSUM and fuzzy logic detecting experimental changes in a scale-model dam.
Test and result
The framework identified structural-state changes; signal mobility/complexity tracked model homogeneity and activity correlated with water flow.
Limitation / still unknown
Evidence comes from a scale-model experiment, not a full-size operating dam; threshold tuning and normal-state data remain limitations described by the authors.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9003076/?report=xml

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Rapidly Identifying New Coronavirus Mutations of Potential Concern in the Omicron Variant Using an Unsupervised Learning Strategy

Source check: Source checked. Not independent scientific confirmation.

Finding: Specific emerging SARS-CoV-2 mutation trajectories are a genomic surveillance finding enabled by unsupervised clustering and temporal analysis. Retain no clinical/functional validation caveat; preprint/journal version is one result.

Technology: Primary journal version (same study) reports unsupervised hierarchical clustering/temporal mutation analysis of 4,296 Omicron genomes; identifies core haplotypes and nine emerging mutations with rising trajectories. Mutations are surveillance signals; source states no relevant clinical data for new mutations, and functional effects are speculative.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Primary journal version (same study) reports unsupervised hierarchical clustering/temporal mutation analysis of 4,296 Omicron genomes; identifies core haplotypes and nine emerging mutations with rising trajectories. Mutations are surveillance signals; source states no relevant clinical data for new mutations, and functional effects are speculative. Research Square v1 date confirmed by NIH author CV/PMID record; journal DOI 10.1038/s41598-022-23342-2 (2022-11-09) is same result.
Test and result
Primary journal version (same study) reports unsupervised hierarchical clustering/temporal mutation analysis of 4,296 Omicron genomes; identifies core haplotypes and nine emerging mutations with rising trajectories. Mutations are surveillance signals; source states no relevant clinical data for new mutations, and functional effects are speculative. Research Square v1 date confirmed by NIH author CV/PMID record; journal DOI 10.1038/s41598-022-23342-2 (2022-11-09) is same result.
Limitation / still unknown
Primary journal version (same study) reports unsupervised hierarchical clustering/temporal mutation analysis of 4,296 Omicron genomes; identifies core haplotypes and nine emerging mutations with rising trajectories. Mutations are surveillance signals; source states no relevant clinical data for new mutations, and functional effects are speculative. Research Square v1 date confirmed by NIH author CV/PMID record; journal DOI 10.1038/s41598-022-23342-2 (2022-11-09) is same result.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8887078/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Optimised passive marker device visibility and automatic marker detection for 3-T MRI-guided endovascular interventions: a pulsatile flow phantom study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary article evaluates deep-learning automatic marker detection and MRI visibility across 3-T pulsatile-flow phantom conditions, yielding a specific device-tracking result.

Technology: A learned detector is tested for robustness as MRI parameters and marker characteristics vary.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A learned detector is tested for robustness as MRI parameters and marker characteristics vary.
Limitation / still unknown
Testing was in a 3-T pulsatile-flow phantom, not patients or a clinical intervention; clinical tracking benefit remains unestablished.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s41747-022-00262-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning for Shape Memory Graphene Nanoribbons and Applications in Biomedical Engineering.

Source check: Source checked. Not independent scientific confirmation.

Finding: An active-learning moment-tensor interatomic potential trained against DFT extends shape-memory analysis to graphene-oxide nanoribbons; the source reports a width-dependent suppression of the metastable phase and checks the potential against DFT calculations. External replication and prospective utility were not established in the reviewed material.

Technology: active-learning moment-tensor interatomic potential trained against density-functional-theory calculations

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
active-learning moment-tensor interatomic potential trained against density-functional-theory calculations
Test and result
An active-learning moment-tensor interatomic potential trained against DFT extends shape-memory analysis to graphene-oxide nanoribbons; the source reports a width-dependent suppression of the metastable phase and checks the potential against DFT calculations.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/bioengineering/bioengineering-09-00090/article_deploy/bioengineering-09-00090.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Practical Deep Learning Model in Differentiating Pneumonia-Type Lung Carcinoma from Pneumonia on CT Images: ResNet Added with Attention Mechanism

Source check: Source checked. Not independent scientific confirmation.

Finding: The study reports AUC around 0.82 and diagnostic performance on its independent test set and cross-validation. This is a retrospective, single-center evaluation; prospective and external clinical validation are not established.

Technology: A lesion-segmenting U-Net and attention-augmented ResNet classify pneumonia-type lung carcinoma versus pneumonia on CT.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A lesion-segmenting U-Net and attention-augmented ResNet classify pneumonia-type lung carcinoma versus pneumonia on CT.
Test and result
The study reports AUC around 0.82 and diagnostic performance on its independent test set and cross-validation.
Limitation / still unknown
This is a retrospective, single-center evaluation; prospective and external clinical validation are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8890890/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep neural network prediction of genome-wide transcriptome signatures - beyond the Black-box.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary article uses a deep neural network to predict transcriptome signatures from transcription-factor expression and reports specific regulatory and disease-associated patterns; these are model-derived biological findings, not experimentally confirmed causal mechanisms.

Technology: A deep neural network predicts genome-wide transcriptome signatures from transcription-factor expression.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A deep neural network predicts genome-wide transcriptome signatures from transcription-factor expression.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41540-022-00218-9.pdf · https://pmc.ncbi.nlm.nih.gov/articles/PMC8866467/ · https://www.nature.com/articles/s41540-022-00218-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Combined Free-Energy Calculation and Machine Learning Methods for Understanding Ligand Unbinding Kinetics.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full accepted manuscript links learned transition-state analysis to experimentally compared ligand-unbinding barriers.

Technology: Machine learning interprets transition-state interactions within free-energy calculations of ligand unbinding.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine learning interprets transition-state interactions within free-energy calculations of ligand unbinding.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://discovery.ucl.ac.uk/id/eprint/10146650/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Clinical Events Based on Raw Text: From Bag-of-Words to Attention-Based Transformers.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The study compared bag-of-words, recurrent neural networks and attention/transformer language models on electronic-record narratives to predict readmission and mortality-related events in one health-system cohort. The retrospective evaluation does not establish benefit from using the predictions in care.

Technology: Bag-of-words, recurrent and attention/transformer text models mapped clinical narratives to readmission and mortality-related outcomes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Bag-of-words, recurrent and attention/transformer text models mapped clinical narratives to readmission and mortality-related outcomes.
Limitation / still unknown
Evaluation used retrospective records from one health system; it does not establish prospective prediction or improved clinical outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/fdgth/articles/10.3389/fdgth.2021.810260/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Effect of head motion-induced artefacts on the reliability of deep learning-based whole-brain segmentation

Source check: Source checked. Not independent scientific confirmation.

Finding: Across 110 participants, deep-learning segmentation remained more consistent than FreeSurfer across radiologist-rated image-quality levels. The result concerns segmentation reliability under motion, not diagnostic benefit; conclusions are limited to evaluated sequences and methods.

Technology: FastSurferCNN, Kwyk and ReSeg neural-network segmentation are compared with FreeSurfer on rest and active-motion MRI.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
FastSurferCNN, Kwyk and ReSeg neural-network segmentation are compared with FreeSurfer on rest and active-motion MRI.
Test and result
Across 110 participants, deep-learning segmentation remained more consistent than FreeSurfer across radiologist-rated image-quality levels.
Limitation / still unknown
The result concerns segmentation reliability under motion, not diagnostic benefit; conclusions are limited to evaluated sequences and methods.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41598-022-05583-3.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Analysis of Immune Cells for Diagnosis and Prognosis of Cutaneous Melanoma.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine learning on five melanoma expression cohorts produced diagnostic/prognostic immune-cell signatures and a disease-specific high-risk immune phenotype.

Technology: ssGSEA quantified 24 immune-cell populations; multiple ML feature-selection/classification methods selected diagnostic and prognostic patterns.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
ssGSEA quantified 24 immune-cell populations; multiple ML feature-selection/classification methods selected diagnostic and prognostic patterns.
Test and result
Results include tumor/control immune-cell differences and survival-risk subgroup associations in melanoma datasets.
Limitation / still unknown
Retrospective GEO/TCGA cohorts and cross-dataset analysis; no prospective clinical validation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8813285/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC8813285/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC8813285/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Brain tumor IDH, 1p/19q, and MGMT molecular classification using MRI-based deep learning: an initial study on the effect of motion and motion correction.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: MRI-based deep learning predicts glioma IDH, 1p/19q, and MGMT molecular status and measures the effect of motion and motion correction on those predictions.

Technology: The source compares molecular-classification performance across motion conditions in a brain-tumor imaging cohort.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The source compares molecular-classification performance across motion conditions in a brain-tumor imaging cohort.
Limitation / still unknown
The MRI molecular-status result is limited to the reported cohort and acquisition conditions; no external clinical validation is stated.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/article/MED/35118164 · https://pmc.ncbi.nlm.nih.gov/articles/PMC8794036/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Enriching the Study Population for Ischemic Stroke Therapeutic Trials Using a Machine Learning Algorithm.

Source check: Source checked. Not independent scientific confirmation.

Finding: Gradient boosting identified ischemic-stroke patients likely to have a future event in a trial-relevant time window, evaluating cohort enrichment on linked health records. Organization is unavailable in the deposited bibliographic metadata.

Technology: Gradient-boosting model for stroke cohort enrichment

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods train gradient boosting to predict ischemic stroke within one year; Results assess selection/enrichment performance in a large claims dataset.
Test and result
Gradient boosting identified ischemic-stroke patients likely to have a future event in a trial-relevant time window, evaluating cohort enrichment on linked health records.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8823366/

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Approaches for Predicting Difficult Airway and First-Pass Success in the Emergency Department: Multicenter Prospective Observational Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Seven machine-learning models predicted difficult airway and first-pass intubation success in a prospective multicenter emergency-department cohort and were compared with the modified LEMON reference.

Technology: the JEAN-2 prospective multicenter cohort and outcomes; Results compare model discrimination with the clinical reference.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
the JEAN-2 prospective multicenter cohort and outcomes; Results compare model discrimination with the clinical reference.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8826144/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning based high-throughput phenotyping of chalkiness in rice exposed to high night temperature

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: CNN classification plus Grad-CAM heatmaps quantify rice-grain chalkiness and are applied to high-night-temperature material; the abstract reports image classification and segmentation evaluation.

Technology: CNN image-classification and segmentation models with Grad-CAM

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Quantifies rice-grain chalkiness from images and applies the phenotyping workflow to material exposed to high night temperature.
Limitation / still unknown
The reported result is limited to the study data, samples and conditions; broader application was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13007-022-00839-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Clinical Prognostic Model Based on Machine Learning from the Fondazione Italiana Linfomi (FIL) MCL0208 Phase III Trial.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A machine-learning prognostic score derived from the phase III MCL0208 trial stratified mantle-cell-lymphoma patient outcomes and was checked in external trial cohorts.

Technology: Methods define trial and prognostic model; Results report score stratification and external validation in European MCL cohorts.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Methods define trial and prognostic model; Results report score stratification and external validation in European MCL cohorts.
Limitation / still unknown
The score was derived from the MCL0208 trial and checked in external trial cohorts; this is prognostic stratification evidence, not proof of treatment benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8750124/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial Intelligence to Diagnose Tibial Plateau Fractures: An Intelligent Assistant for Orthopedic Physicians.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary source Methods/Results confirm the represented result: RetinaNet-based AI detection of tibial plateau fractures on X-rays, validated against a reference database for clinical assistance. External replication and prospective utility were not established in the reviewed material.

Technology: RetinaNet-based AI detection of tibial plateau fractures on X-rays, validated against a reference database for clinical assistance.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
RetinaNet-based AI detection of tibial plateau fractures on X-rays, validated against a reference database for clinical assistance.
Test and result
The primary source Methods/Results confirm the represented result: RetinaNet-based AI detection of tibial plateau fractures on X-rays, validated against a reference database for clinical assistance.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11596-021-2501-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Looking for low vision: Predicting visual prognosis by fusing structured and free-text data from electronic health records.

Source check: Source checked. Not independent scientific confirmation.

Finding: Deep-learning models fused structured EHR data and ophthalmology notes to predict whether low vision would improve within a year; on 300 held-out patients, domain-specific text embeddings improved AUROC to about 0.82 versus structured-only and entity-based approaches. This is a bounded retrospective clinical prognosis result. Organization is unavailable in the deposited bibliographic metadata.

Technology: Deep-learning models combining EHR text and structured data

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Author manuscript Methods/Results reviewed: single-center Stanford cohort (5,547), structured/text model comparisons, independent held-out test (300). No intervention or generalization claim; catalog search found no same result.
Test and result
Deep-learning models fused structured EHR data and ophthalmology notes to predict whether low vision would improve within a year; on 300 held-out patients, domain-specific text embeddings improved AUROC to about 0.82 versus structured-only and entity-based approaches. This is a bounded retrospective clinical prognosis result.
Limitation / still unknown
This is a bounded retrospective clinical prognosis result.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://benjamintseng.com/wp-content/uploads/2022/01/Looking-for-low-vision-Predicting-visual-prognosis-by-fusing-structured-and-free-text-data-from-ehrs.pdf

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Breast Cancer Surgery 10-Year Survival Prediction by Machine Learning: A Large Prospective Cohort Study

Source check: Source checked. Not independent scientific confirmation.

Finding: The study compares model performance using separate test and validation sets in a large prospective cohort registry. Despite the prospective cohort source, this is a retrospective model analysis; external prospective benefit is not established.

Technology: Retrospective multi-center registry data are used to compare DNN, KNN, SVM, naive Bayes and Cox models for 10-year breast-cancer survival.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Retrospective multi-center registry data are used to compare DNN, KNN, SVM, naive Bayes and Cox models for 10-year breast-cancer survival.
Test and result
The study compares model performance using separate test and validation sets in a large prospective cohort registry.
Limitation / still unknown
Despite the prospective cohort source, this is a retrospective model analysis; external prospective benefit is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/biology/biology-11-00047/article_deploy/biology-11-00047.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The Impact of Resampling and Denoising Deep Learning Algorithms on Radiomics in Brain Metastases MRI.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Deep-learning resampling/denoising is applied before radiomic feature extraction; Methods specifies MRI workflow. Results measures feature stability changes in brain-metastasis MRI; bounded methodological/clinical imaging result.

Technology: Deep-learning resampling/denoising is applied before radiomic feature extraction; Methods specifies MRI workflow.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep-learning resampling/denoising is applied before radiomic feature extraction; Methods specifies MRI workflow.
Limitation / still unknown
The result concerns radiomic feature stability under the tested resampling/denoising pipelines; it does not establish improved diagnosis or patient outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/articles/PMC8750741

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Spatial Organization and Prognostic Significance of NK and NKT-like Cells via Multimarker Analysis of the Colorectal Cancer Microenvironment.

Source check: Source checked. Not independent scientific confirmation.

Finding: Multiplex imaging plus supervised ML identified NK/NKT-like/T-cell populations and spatial relationships associated with colorectal-cancer survival. External replication and prospective utility were not established in the reviewed material.

Technology: A multiplex immunofluorescence assay and digital image analysis/supervised ML quantified cell markers and distances to tumor cells.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A multiplex immunofluorescence assay and digital image analysis/supervised ML quantified cell markers and distances to tumor cells.
Test and result
Multiplex imaging plus supervised ML identified NK/NKT-like/T-cell populations and spatial relationships associated with colorectal-cancer survival.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://oulurepo.oulu.fi/bitstream/handle/10024/44163/nbnfi-fe202201199473.pdf?sequence=1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Gas-Liquid Two-Phase Flow Pattern Identification of a Centrifugal Pump Based on SMOTE and Artificial Neural Network.

Source check: Source checked. Not independent scientific confirmation.

Finding: SMOTE plus ANN classifies gas-liquid two-phase flow patterns in a centrifugal pump; reported identification rate rises from 89.91% to 94.88% on the study test setup. Organization is unavailable in the deposited bibliographic metadata.

Technology: SMOTE and artificial neural network classifier

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
SMOTE plus ANN classifies gas-liquid two-phase flow patterns in a centrifugal pump; reported identification rate rises from 89.91% to 94.88% on the study test setup.
Test and result
SMOTE plus ANN classifies gas-liquid two-phase flow patterns in a centrifugal pump; reported identification rate rises from 89.91% to 94.88% on the study test setup.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3390/mi13010002

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A novel machine learning derived RNA-binding protein gene-based score system predicts prognosis of hepatocellular carcinoma patients

Source check: Source checked. Not independent scientific confirmation.

Finding: Across nine GEO cohorts and TCGA, the score adds prognostic information alongside TNM staging in validation cohorts. These retrospective cohort associations do not establish treatment response or causal biology.

Technology: A LASSO-Cox RNA-binding-protein score stratifies overall survival in hepatocellular carcinoma and is tested across GEO and TCGA cohorts.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A LASSO-Cox RNA-binding-protein score stratifies overall survival in hepatocellular carcinoma and is tested across GEO and TCGA cohorts.
Test and result
Across nine GEO cohorts and TCGA, the score adds prognostic information alongside TNM staging in validation cohorts.
Limitation / still unknown
These retrospective cohort associations do not establish treatment response or causal biology.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8697767/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Unsupervised Learning for Automated Detection of Coronary Artery Disease Subgroups.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary Methods/Results establish the AI-enabled result within the stated scope and limitations.

Technology: The named learned model operation is stated in Methods and is directly applied to the represented clinical/scientific outcome.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The named learned model operation is stated in Methods and is directly applied to the represented clinical/scientific outcome.
Limitation / still unknown
The identified coronary-disease subgroups are data-derived patterns; they do not establish causal subtypes or prospective clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9075403/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Models That Integrate Tumor Texture and Perfusion Characteristics Using Low-Dose Breast Computed Tomography Are Promising for Predicting Histological Biomarkers and Treatment Failure

Source check: Source checked. Not independent scientific confirmation.

Finding: A prospective study used machine learning on low-dose breast CT texture/perfusion features to predict histological biomarkers and treatment failure. External replication and prospective utility were not established in the reviewed material.

Technology: The study uses a machine-learning model for the stated machine learning models that integrate tumor texture and perfusion characteristics using low-dose breast computed tomography are promising for predicting histological biomarkers and treatment failure task.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study uses a machine-learning model for the stated machine learning models that integrate tumor texture and perfusion characteristics using low-dose breast computed tomography are promising for predicting histological biomarkers and treatment failure task.
Test and result
A prospective study used machine learning on low-dose breast CT texture/perfusion features to predict histological biomarkers and treatment failure.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8656976/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and Validation of a Deep Learning Algorithm to Automatic Detection of Pituitary Microadenoma From MRI.

Source check: Source checked. Not independent scientific confirmation.

Finding: Methods/Results support two-stage deep-learning MRI detection of pituitary microadenomas with reference diagnosis and retrospective temporal/external testing; claim bounded to diagnostic performance, no clinical benefit inferred. The paper appeared in Frontiers in Medicine. Organization is unavailable in the deposited bibliographic metadata.

Technology: Two-stage deep-learning MRI detector

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned imaging detector produces a patient-level diagnostic result; validation remains retrospective.
Test and result
Methods/Results support two-stage deep-learning MRI detection of pituitary microadenomas with reference diagnosis and retrospective temporal/external testing; claim bounded to diagnostic performance, no clinical benefit inferred.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2021.758690/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A machine learning framework for the evaluation of myocardial rotation in patients with noncompaction cardiomyopathy.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Random forest/MLP analysis of 49 rigorously diagnosed noncompaction cardiomyopathy cases links echocardiographic predictors to rigid-body rotation patterns and associated dysfunction.

Technology: prospective echocardiographic data, n=49 after image-quality exclusion; five-fold validation; model identifies echo predictors of RBR. Small single-center exploratory sample; authors say further prospective work is needed.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
prospective echocardiographic data, n=49 after image-quality exclusion; five-fold validation; model identifies echo predictors of RBR. Small single-center exploratory sample; authors say further prospective work is needed.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0260195

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning Enables 60% Accelerated Volumetric Brain MRI While Preserving Quantitative Performance: A Prospective, Multicenter, Multireader Trial.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary prospective multicenter/multireader trial: 40 subjects at five institutions/six scanners; vendor-agnostic CNN enhancement of 60%-faster MRI; quantitative volumetrics and clinical disease classification matched standard of care, with blinded readers rating quality. Small sample, industry-affiliated authors and no downstream workflow outcome.

Technology: Primary prospective multicenter/multireader trial: 40 subjects at five institutions/six scanners; vendor-agnostic CNN enhancement of 60%-faster MRI; quantitative volumetrics and clinical disease classification matched standard of care, with blinded readers rating quality. Small sample, industry-affiliated authors and no downstream workflow outcome.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Primary prospective multicenter/multireader trial: 40 subjects at five institutions/six scanners; vendor-agnostic CNN enhancement of 60%-faster MRI; quantitative volumetrics and clinical disease classification matched standard of care, with blinded readers rating quality. Small sample, industry-affiliated authors and no downstream workflow outcome.
Limitation / still unknown
Primary prospective multicenter/multireader trial: 40 subjects at five institutions/six scanners; vendor-agnostic CNN enhancement of 60%-faster MRI; quantitative volumetrics and clinical disease classification matched standard of care, with blinded readers rating quality. Small sample, industry-affiliated authors and no downstream workflow outcome.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8805755/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial intelligence to diagnosis distal radius fracture using biplane plain X-rays.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The study reports diagnostic accuracy on labeled X-ray patterns, with test performance around 89–94%.

Technology: A CNN diagnoses distal-radius fracture from biplane plain radiographs, a specific fracture-detection task.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A CNN diagnoses distal-radius fracture from biplane plain radiographs, a specific fracture-detection task.
Limitation / still unknown
The diagnostic result is limited to labeled X-ray patterns and the tested dataset; it does not establish prospective clinical performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/article/MED/34823550 · https://pmc.ncbi.nlm.nih.gov/articles/PMC8620959/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A non-invasive, automated diagnosis of Menière's disease using radiomics and machine learning on conventional magnetic resonance imaging: A multicentric, case-controlled feasibility study.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary source Methods/Results confirm the represented result: Multicentre case-control radiomics study on conventional MRI with an MLP classifier diagnosing Meniere's disease (120 patients, 140 controls); a clinically specific retrospective diagnostic assessment with an identified ML role. Organization is unavailable in the deposited bibliographic metadata.

Technology: Radiomics and multilayer perceptron for Meniere disease classification

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Multicentre case-control radiomics study on conventional MRI with an MLP classifier diagnosing Meniere's disease (120 patients, 140 controls); a clinically specific retrospective diagnostic assessment with an identified ML role.
Test and result
The primary source Methods/Results confirm the represented result: Multicentre case-control radiomics study on conventional MRI with an MLP classifier diagnosing Meniere's disease (120 patients, 140 controls); a clinically specific retrospective diagnostic assessment with an identified ML role.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11547-021-01425-w

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Central Serous Chorioretinopathy Recurrence Using Machine Learning

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning models predict recurrence of central serous chorioretinopathy after laser treatment; the paper reports internal and external test-set accuracy and identifies clinical recurrence predictors. This is retrospective recurrence prediction, not evidence that model-guided care reduces recurrence. Internal and external test cohorts are described and Table 2 reports recurrence performance.

Technology: Supervised ML classifiers predict recurrence from clinical/ophthalmic variables.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Supervised ML classifiers predict recurrence from clinical/ophthalmic variables.
Test and result
Machine-learning models predict recurrence of central serous chorioretinopathy after laser treatment; the paper reports internal and external test-set accuracy and identifies clinical recurrence predictors. This is retrospective recurrence prediction, not evidence that model-guided care reduces recurrence.
Limitation / still unknown
Internal and external test cohorts are described and Table 2 reports recurrence performance. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2021.649316/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Post-Therapeutic Visual Acuity and OCT Images in Patients With Central Serous Chorioretinopathy by Artificial Intelligence

Source check: Source checked. Not independent scientific confirmation.

Finding: Learned prognosis models predict post-laser acuity and OCT against observed outcomes, including a separate-center test cohort; this does not establish treatment benefit. Training/validation at ZOC and independent XEC test cohort; comparison with ground-truth and specialist assessments.

Technology: Blended regressors forecast post-laser visual acuity; pix2pixHD generates later OCT from pretreatment scans.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Blended regressors forecast post-laser visual acuity; pix2pixHD generates later OCT from pretreatment scans.
Test and result
Retrospective cohort results compare predicted visual acuity and synthetic OCT with measured outcomes; this is prognostic prediction, not treatment benefit.
Limitation / still unknown
Training/validation at ZOC and independent XEC test cohort; comparison with ground-truth and specialist assessments. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/bioengineering-and-biotechnology/articles/10.3389/fbioe.2021.649221/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Predictive Model Based on a New CI-AKI Definition to Predict Contrast Induced Nephropathy in Patients With Coronary Artery Disease With Relatively Normal Renal Function.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The study reports a specific contrast-induced acute kidney injury prediction result in coronary artery disease patients.

Technology: Machine-learning models use clinical and peri-procedural variables to predict contrast-induced acute kidney injury.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning models use clinical and peri-procedural variables to predict contrast-induced acute kidney injury.
Limitation / still unknown
This is a patient-cohort prediction result; it does not establish causal kidney injury mechanisms or clinical benefit from model use.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/cardiovascular-medicine/articles/10.3389/fcvm.2021.762576/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predictive value of a novel Asian lung cancer screening nomogram based on artificial intelligence and epidemiological characteristics.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary full-text record describes a CNN-derived lung-nodule score combined with epidemiological predictors and tested in an independent multicenter cohort; this is a specific diagnostic-risk result. External replication and prospective utility were not established in the reviewed material.

Technology: convolutional neural network lung-nodule score combined with epidemiological predictors

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
convolutional neural network lung-nodule score combined with epidemiological predictors
Test and result
The primary full-text record describes a CNN-derived lung-nodule score combined with epidemiological predictors and tested in an independent multicenter cohort; this is a specific diagnostic-risk result.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8636223/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Superiority of Multiple-Joint Space Width over Minimum-Joint Space Width Approach in the Machine Learning for Radiographic Severity and Knee Osteoarthritis Progression.

Source check: Source checked. Not independent scientific confirmation.

Finding: CNN-based joint-space-width estimation plus XGBoost predicts knee osteoarthritis progression in OAI, validated against radiologist measurements; a clinically specific retrospective assessment. Organization is unavailable in the deposited bibliographic metadata.

Technology: CNN joint-space-width estimator and XGBoost progression model

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A convolutional neural network estimated joint-space width from radiographs, and XGBoost used those measurements to assess radiographic severity and osteoarthritis progression.
Test and result
CNN-based joint-space-width estimation plus XGBoost predicts knee osteoarthritis progression in OAI, validated against radiologist measurements; a clinically specific retrospective assessment.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/biology/biology-10-01107/article_deploy/biology-10-01107.pdf

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Skin Electrical Resistance as a Diagnostic and Therapeutic Biomarker of Breast Cancer Measuring Lymphatic Regions

Source check: Source checked. Not independent scientific confirmation.

Finding: The study reports AUC 0.68 and includes longitudinal treatment measurements from six patients. Diagnostic performance is modest and the investigational cohort is small; treatment-response evidence from six patients is preliminary.

Technology: A random-forest classifier uses skin electrical-resistance measurements to distinguish malignant from benign breast lesions.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A random-forest classifier uses skin electrical-resistance measurements to distinguish malignant from benign breast lesions.
Test and result
The study reports AUC 0.68 and includes longitudinal treatment measurements from six patients.
Limitation / still unknown
Diagnostic performance is modest and the investigational cohort is small; treatment-response evidence from six patients is preliminary.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8654262/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Hypertrophic Cardiomyopathy Diagnosis Based on Cardiovascular Magnetic Resonance Using Deep Learning Techniques

Source check: Source checked. Not independent scientific confirmation.

Finding: Deep-learning analysis of cardiovascular magnetic-resonance images classified hypertrophic cardiomyopathy and controls; the preprint reports diagnostic performance against its labeled study data.

Technology: Deep-learning cardiovascular-MRI classifier

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Classifies hypertrophic cardiomyopathy and control cases from cardiovascular magnetic-resonance images.
Test and result
The preprint evaluates deep-learning classification of labeled hypertrophic-cardiomyopathy and control CMR images; external clinical validation is not reported in the case checks.
Limitation / still unknown
The reported evaluation does not establish prospective benefit or performance beyond the studied setting.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.21203/rs.3.rs-1005999/v1 · https://europepmc.org/article/PPR/PPR411111 · https://www.researchsquare.com/article/rs-1005999/v1.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Rapid SERS identification of methicillin-susceptible and methicillin-resistant Staphylococcus aureus via aptamer recognition and deep learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A clinically sourced diagnostic proof of concept differentiated susceptibility classes on a small unseen set.

Technology: A CNN classified measured spectra relative to established susceptibility labels.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A CNN classified measured spectra relative to established susceptibility labels.
Limitation / still unknown
The diagnostic proof of concept used a small unseen set; broader clinical validation is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9042729/ · https://doi.org/10.1039/d1ra05778b · https://pubmed.ncbi.nlm.nih.gov/35494737/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Model to Identify Sepsis Patients in the Emergency Department: Algorithm Development and Validation

Source check: Source checked. Not independent scientific confirmation.

Finding: Development included 8,296 patients and a separate external cohort of 1,744, with AUC 0.86 internal and 0.75 external. External validation performance was lower; the authors note heterogeneity and the need for prospective interventional evaluation.

Technology: XGBoost uses emergency-department clinical data to identify sepsis against a Sepsis-3 reference standard.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
XGBoost uses emergency-department clinical data to identify sepsis against a Sepsis-3 reference standard.
Test and result
Development included 8,296 patients and a separate external cohort of 1,744, with AUC 0.86 internal and 0.75 external.
Limitation / still unknown
External validation performance was lower; the authors note heterogeneity and the need for prospective interventional evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2075-4426/11/11/1055

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Evaluating Scalable Supervised Learning for Synthesize-on-Demand Chemical Libraries.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Read ChemRxiv v1 PDF/record and PMC full article. Supervised screening of synthesize-on-demand library; prospective tests found 31/68 PriA-SSB inhibitors, including 1.3 µM hit. Narrowly an in-vitro target-inhibitor discovery; v1 already contains result, predating journal version.

Technology: Read ChemRxiv v1 PDF/record and PMC full article. Supervised screening of synthesize-on-demand library; prospective tests found 31/68 PriA-SSB inhibitors, including 1.3 µM hit. Narrowly an in-vitro target-inhibitor discovery; v1 already contains result, predating journal version.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Read ChemRxiv v1 PDF/record and PMC full article. Supervised screening of synthesize-on-demand library; prospective tests found 31/68 PriA-SSB inhibitors, including 1.3 µM hit. Narrowly an in-vitro target-inhibitor discovery; v1 already contains result, predating journal version.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://chemrxiv.org/engage/chemrxiv/article-details/616dde0df718df8ad1e3422d

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning for Urodynamic Detection of Detrusor Overactivity

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary Methods/Results use pediatric urodynamics files from two machines at one institution (1,154 total, 810 cystometry; file-level train/test split) to detect detrusor overactivity; time-based model AUC 91.9%, frequency specificity 91.7%. Bounded retrospective diagnostic result.

Technology: Primary Methods/Results use pediatric urodynamics files from two machines at one institution (1,154 total, 810 cystometry; file-level train/test split) to detect detrusor overactivity; time-based model AUC 91.9%, frequency specificity 91.7%. Bounded retrospective diagnostic result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Primary Methods/Results use pediatric urodynamics files from two machines at one institution (1,154 total, 810 cystometry; file-level train/test split) to detect detrusor overactivity; time-based model AUC 91.9%, frequency specificity 91.7%. Bounded retrospective diagnostic result.
Limitation / still unknown
Primary Methods/Results use pediatric urodynamics files from two machines at one institution (1,154 total, 810 cystometry; file-level train/test split) to detect detrusor overactivity; time-based model AUC 91.9%, frequency specificity 91.7%. Bounded retrospective diagnostic result. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8865755/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of Apple Leaf Diseases by Improved Deep Convolutional Neural Networks With an Attention Mechanism.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A coordinate-attention EfficientNet model classifies apple leaf diseases on a collected dataset and is compared with conventional CNNs on held-out images; this is an application-specific plant-disease detection result, not a new disease biology finding.

Technology: CA-ENet deep CNN combines coordinate attention with EfficientNet for disease-image classification.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CA-ENet deep CNN combines coordinate attention with EfficientNet for disease-image classification.
Limitation / still unknown
Held-out dataset evaluation, augmentation comparison and baseline CNN metrics are reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2021.723294/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Novel Gene Signatures Predicting Primary Non-response to Infliximab in Ulcerative Colitis: Development and Validation Combining Random Forest With Artificial Neural Network.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: RF/ANN gene signature for primary infliximab nonresponse in UC; GSE16879 and GSE12251 training (n=46 total), independent GSE23597 validation (n=28), response assessed at 4/8 weeks using mucosal healing definition.

Technology: Learned gene signature predicts a specific treatment-response endpoint in independent public data; retrospective and small-cohort.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned gene signature predicts a specific treatment-response endpoint in independent public data; retrospective and small-cohort.
Limitation / still unknown
The gene signature was trained on 46 cases and checked in 28 independent cases; these small transcriptomic cohorts do not establish causal mechanisms or clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2021.678424/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identifying the Location of an Accessory Pathway in Pre-Excitation Syndromes Using an Artificial Intelligence-Based Algorithm

Source check: Source checked. Not independent scientific confirmation.

Finding: A neural network used 12-lead ECG delta-wave polarity to identify accessory-pathway location against successful ablation site in 357 retrospective WPW patients. It was compared with three conventional ECG algorithms and had higher reported accuracy; the single-center retrospective evaluation does not establish prospective benefit.

Technology: Backpropagation neural network trained on 12-lead ECG-derived delta-wave polarity; ablation site was reference. 70/30 patient-level train/internal-validation split; accuracy 85.7% across 14 locations, versus lower accuracy at lower resolution for three conventional algorithms.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Backpropagation neural network trained on 12-lead ECG-derived delta-wave polarity; ablation site was reference. 70/30 patient-level train/internal-validation split; accuracy 85.7% across 14 locations, versus lower accuracy at lower resolution for three conventional algorithms.
Test and result
Backpropagation neural network trained on 12-lead ECG-derived delta-wave polarity; ablation site was reference. 70/30 patient-level train/internal-validation split; accuracy 85.7% across 14 locations, versus lower accuracy at lower resolution for three conventional algorithms.
Limitation / still unknown
A neural network used 12-lead ECG delta-wave polarity to identify accessory-pathway location against successful ablation site in 357 retrospective WPW patients. It was compared with three conventional ECG algorithms and had higher reported accuracy; the single-center retrospective evaluation does not establish prospective benefit. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/jcm/jcm-10-04394/article_deploy/jcm-10-04394.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development, validation, and application of a machine learning model to estimate salt consumption in 54 countries.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary eLife full text/PDF reports model derivation and validation using 21 surveys (49,776 people), then application to 54 countries (166,677 people); version history lists preprint 2021-09-02 and VOR 2022-01-25.

Technology: Supervised regression uses routine survey variables to estimate population mean salt intake; source explicitly limits prediction to population level. Estimates in 54 countries are not direct urine measurements.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Supervised regression uses routine survey variables to estimate population mean salt intake; source explicitly limits prediction to population level. Estimates in 54 countries are not direct urine measurements.
Limitation / still unknown
The country estimates apply a model derived/validated on 21 surveys to 54 countries; these are modeled estimates, not direct measurements in every country.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://elifesciences.org/articles/72930.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting postoperative surgical site infection with administrative data: a random forests algorithm

Source check: Source checked. Not independent scientific confirmation.

Finding: A random-forest model predicted postoperative surgical-site infection from linked administrative records. The combined model had an internally evaluated C-statistic of 0.91.

Technology: Random Forest and high-performance logistic regression select/fit administrative-data SSI predictors.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random Forest and high-performance logistic regression select/fit administrative-data SSI predictors.
Test and result
The Ottawa cohort included 14,351 patients and 795 surgical-site infections. Hospitalization, physician and procedure codes supplied the model inputs.
Limitation / still unknown
The authors identified external validation as a remaining requirement.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12874-021-01369-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Analysis of the Impact of Sustained Load and Temperature on the Performance of the Electromechanical Impedance Technique through Multilevel Machine Learning and FBG Sensors.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Machine-learning models used electromechanical-impedance measurements, with fiber Bragg grating sensors recording load and temperature, to quantify how sustained mechanical load and temperature altered sensor impedance. This was an instrument/materials experiment under the reported conditions.

Technology: Machine-learning regressors analyzed electromechanical-impedance signals alongside fiber Bragg grating load and temperature measurements.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning regressors analyzed electromechanical-impedance signals alongside fiber Bragg grating load and temperature measurements.
Limitation / still unknown
Conclusions are confined to the tested load/temperature conditions and sensor setup; no independent deployment or broader materials validation is reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8434432/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and Validation of a Prediction Model for Elevated Arterial Stiffness in Chinese Patients With Diabetes Using Machine Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A machine-learning model predicted elevated arterial stiffness in Chinese adults with diabetes; an external Japanese validation cohort is reported (AUROC 0.821 per the source intake), supporting cross-cohort prediction but not a causal risk-factor interpretation.

Technology: Gradient-boosted/other ML models estimate elevated arterial stiffness from clinical predictors.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Gradient-boosted/other ML models estimate elevated arterial stiffness from clinical predictors.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2021.714195/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Raman Spectroscopy and Machine Learning for IDH Genotyping of Unprocessed Glioma Biopsies.

Source check: Source checked. Not independent scientific confirmation.

Finding: Raman spectroscopy and machine learning classified IDH mutation status in untreated glioma biopsies for an intraoperative diagnostic task.

Technology: Methods describe 38 untreated tumor specimens and Raman spectra; Results compare XGBoost/SVM classification against IDH status.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe 38 untreated tumor specimens and Raman spectra; Results compare XGBoost/SVM classification against IDH status.
Test and result
Methods describe 38 untreated tumor specimens and Raman spectra; Results compare XGBoost/SVM classification against IDH status.
Limitation / still unknown
Methods describe 38 untreated tumor specimens and Raman spectra; Results compare XGBoost/SVM classification against IDH status.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8392399/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Towards patient-specific prediction of conduction abnormalities induced by transcatheter aortic valve implantation: a combined mechanistic modelling and machine learning approach.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Patient-specific valve simulations supplied contact-pressure features for ML prediction of conduction abnormalities after TAVI, a specific procedure-related clinical risk result.

Technology: patient-specific valve simulation generated contact-pressure features that were combined with ML for post-TAVI conduction abnormality prediction.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
patient-specific valve simulation generated contact-pressure features that were combined with ML for post-TAVI conduction abnormality prediction.
Limitation / still unknown
The model predicts conduction risk from simulated valve-contact features; it does not establish observed patient outcomes or causal procedure effects.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9708019/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC9708019/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC9708019/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Body shape matters: Evidence from machine learning on body shape-income relationship.

Source check: Source checked. Not independent scientific confirmation.

Finding: A graphical autoencoder derives 3D body-shape features that change estimated gender-specific family-income associations versus sparse classical measures; the result is observational and dated. Organization is unavailable in the deposited bibliographic metadata.

Technology: Graphical autoencoder for three-dimensional scans

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
graphical autoencoder applied to 3D scans; deep-learned features used in income regressions and compared with classical anthropometrics, with bootstrap intervals and robustness checks. Data were collected in 2002; authors caution that current generalization is uncertain and associations are not causal.
Test and result
A graphical autoencoder derives 3D body-shape features that change estimated gender-specific family-income associations versus sparse classical measures; the result is observational and dated.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0254785

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Using data-driven algorithms for semi-automated geomorphological mapping.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Springer PDF Methods/Results: Direct Sampling and Random Forest trained using a manually mapped area and evaluated over the adjacent Arolla valley test area against a field-made map of eight classes. Both had similar accuracy/Kappa; DS yielded less spatially noisy classes by accounting for spatial dependence. This is a bounded new geomorphology mapping application. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit.

Technology: Springer PDF Methods/Results: Direct Sampling and Random Forest trained using a manually mapped area and evaluated over the adjacent Arolla valley test area against a field-made map of eight classes. Both had similar accuracy/Kappa; DS yielded less spatially noisy classes by accounting for spatial dependence. This is a bounded new geomorphology mapping application.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Springer PDF Methods/Results: Direct Sampling and Random Forest trained using a manually mapped area and evaluated over the adjacent Arolla valley test area against a field-made map of eight classes. Both had similar accuracy/Kappa; DS yielded less spatially noisy classes by accounting for spatial dependence. This is a bounded new geomorphology mapping application.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s00477-021-02062-5 · https://link.springer.com/content/pdf/10.1007/s00477-021-02062-5.pdf · https://api.crossref.org/works/10.1007%2Fs00477-021-02062-5

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identifying elevated risk for future pain crises in sickle-cell disease using photoplethysmogram patterns measured during sleep: A machine learning approach

Source check: Source checked. Not independent scientific confirmation.

Finding: Sleep heart-rate/PPG models predict future vaso-occlusive pain crises in children/adolescents with sickle-cell disease; a stacked model and cross-validation are evaluated against a random baseline. Scope is exploratory home-monitoring risk prediction, not clinical outcome prevention. Methods describes 252 SCD subjects, feature selection, cross-validation and held-out prediction against baseline; retrospective cohort caveats retained.

Technology: Stacked ML classifiers use overnight heart-rate and photoplethysmogram features to predict subsequent VOC pain crises.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Stacked ML classifiers use overnight heart-rate and photoplethysmogram features to predict subsequent VOC pain crises.
Test and result
Sleep heart-rate/PPG models predict future vaso-occlusive pain crises in children/adolescents with sickle-cell disease; a stacked model and cross-validation are evaluated against a random baseline. Scope is exploratory home-monitoring risk prediction, not clinical outcome prevention.
Limitation / still unknown
Methods describes 252 SCD subjects, feature selection, cross-validation and held-out prediction against baseline; retrospective cohort caveats retained. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2021.714741/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Models to Predict Survival Outcomes According to the Surgical Approach of Primary Radical Hysterectomy in Patients with Early Cervical Cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning models estimated survival outcomes for early cervical-cancer patients according to radical-hysterectomy surgical approach in a multicenter cohort.

Technology: Methods describe predictors and survival model development; Results report validation performance for progression-free and overall survival.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe predictors and survival model development; Results report validation performance for progression-free and overall survival.
Test and result
Methods describe predictors and survival model development; Results report validation performance for progression-free and overall survival.
Limitation / still unknown
Methods describe predictors and survival model development; Results report validation performance for progression-free and overall survival.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8345043/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Modeling Agrobacterium -Mediated Gene Transformation of Tobacco ( Nicotiana tabacum )-A Model Plant for Gene Transformation Studies.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Neural networks model Agrobacterium-mediated tobacco transformation outcomes from experimental factors and identify conditions associated with PCR-verified transgenic plants; compared with stepwise regression on experimental data. This is a bounded experimental-process result, not proof that model-guided optimization improved transformation.

Technology: MLP neural models predict putative/PCR-verified transformation outcomes from experimental conditions.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
MLP neural models predict putative/PCR-verified transformation outcomes from experimental conditions.
Limitation / still unknown
The neural-network analysis modeled experimental outcomes but did not show that model-guided optimization improved transformation success.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2021.695110/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A machine learning approach to brain epigenetic analysis reveals kinases associated with Alzheimer's disease.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary study applies EWASplus machine learning to extrapolate epigenome-wide association signals and reports candidate Alzheimer’s-associated loci/kinases, with targeted bisulfite sequencing for a subset. This supports an association result; it does not establish kinase causality. Organization is unavailable in the deposited bibliographic metadata.

Technology: Machine-learning models for multimodal dementia prediction

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PMC/Nature full Methods/Results: ROS/MAP 717 participants plus three cohorts; targeted sequencing of 559 selected sites from 150 donors, 319 passing QC; 25/38 predicted-positive clusters associated with an AD trait versus 3/10 negative controls; kinase enrichment p=.044.
Test and result
The primary study applies EWASplus machine learning to extrapolate epigenome-wide association signals and reports candidate Alzheimer’s-associated loci/kinases, with targeted bisulfite sequencing for a subset. This supports an association result; it does not establish kinase causality.
Limitation / still unknown
This supports an association result; it does not establish kinase causality.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8298578/

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Combining microfluidics with machine learning algorithms for RBC classification in rare hereditary hemolytic anemia.

Source check: Source checked. Not independent scientific confirmation.

Finding: The pilot analyzed 79 videos from 32 participants and reported leave-one-video-out performance for healthy-versus-disease and subtype classification. Small pilot cohort and video-level evaluation limit patient-level generalization; external clinical validation is not established.

Technology: Pretrained AlexNet transferred image features from red-cell videos; cooperative classifiers combined features from a spleen-mimicking microfluidic assay to classify deformability patterns.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Pretrained AlexNet transferred image features from red-cell videos; cooperative classifiers combined features from a spleen-mimicking microfluidic assay to classify deformability patterns.
Test and result
The pilot analyzed 79 videos from 32 participants and reported leave-one-video-out performance for healthy-versus-disease and subtype classification.
Limitation / still unknown
Small pilot cohort and video-level evaluation limit patient-level generalization; external clinical validation is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://art.torvergata.it/retrieve/8779408d-90dd-4167-9525-e9194cfa64e0/s41598-021-92747-2.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning-Based Stage-Wise Risk Stratification for Early Lung Adenocarcinoma in CT Images: A Multi-Center Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A two-stage 3D deep-learning system classified early lung adenocarcinoma nodules from CT and was checked against surgical pathology, multi-center data and radiologist performance.

Technology: stage-wise CT model and multi-center sets; Results report pathology-grounded model validation and slice-thickness/radiologist comparisons.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
stage-wise CT model and multi-center sets; Results report pathology-grounded model validation and slice-thickness/radiologist comparisons.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8269183/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predictive modelling of hypoxic ischaemic encephalopathy risk following perinatal asphyxia.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: clinical variables from a prospective cohort of infants with perinatal asphyxia were split into training/validation and an independent test set; random forest with 5-fold CV predicted subsequent HIE. This is a bounded clinical risk-prediction study; it does not establish improved cooling outcomes.

Technology: Primary Methods/Results: clinical variables from a prospective cohort of infants with perinatal asphyxia were split into training/validation and an independent test set; random forest with 5-fold CV predicted subsequent HIE. This is a bounded clinical risk-prediction study; it does not establish improved cooling outcomes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
clinical variables from a prospective cohort of infants with perinatal asphyxia were split into training/validation and an independent test set; random forest with 5-fold CV predicted subsequent HIE. This is a bounded clinical risk-prediction study; it does not establish improved cooling outcomes.
Limitation / still unknown
clinical variables from a prospective cohort of infants with perinatal asphyxia were split into training/validation and an independent test set; random forest with 5-fold CV predicted subsequent HIE. This is a bounded clinical risk-prediction study; it does not establish improved cooling outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8261660/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial Intelligence-Assisted Electrocardiography for Early Diagnosis of Thyrotoxic Periodic Paralysis

Source check: Source checked. Not independent scientific confirmation.

Finding: An ECG deep-learning system estimated hypokalemia and supported thyrotoxic periodic paralysis diagnosis, with a separated validation cohort and a prospective clinical evaluation. Methods describe ECG12Net-derived models, development/validation cohorts and prospective workflow; Results report diagnostic performance and the limited prospective cohort.

Technology: Methods describe ECG12Net-derived models, development/validation cohorts and prospective workflow; Results report diagnostic performance and the limited prospective cohort.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe ECG12Net-derived models, development/validation cohorts and prospective workflow; Results report diagnostic performance and the limited prospective cohort.
Test and result
Methods describe ECG12Net-derived models, development/validation cohorts and prospective workflow; Results report diagnostic performance and the limited prospective cohort.
Limitation / still unknown
Methods describe ECG12Net-derived models, development/validation cohorts and prospective workflow; Results report diagnostic performance and the limited prospective cohort. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8294684/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of Clinically Relevant Subgroups of Chronic Lymphocytic Leukemia Through Discovery of Abnormal Molecular Pathways.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Molecular pathway and clustering analyses identified clinically relevant chronic-lymphocytic-leukemia subgroups with distinct pathway patterns, a specific molecular-stratification finding.

Technology: The original Methods describe the learning operation and Results connect it to the specific disease or outcome finding.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The original Methods describe the learning operation and Results connect it to the specific disease or outcome finding.
Limitation / still unknown
The subgroup and pathway result is based on existing transcriptomic cohorts and does not establish causal disease mechanisms or prospective clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fgene.2021.627964 · https://europepmc.org/article/MED/34262590 · https://europepmc.org/articles/PMC8273263 · https://www.ncbi.nlm.nih.gov/research/bionlp/RESTful/pmcoa.cgi/BioC_json/PMC8273263/unicode

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Advanced thyroid carcinomas: neural network analysis of ultrasonographic characteristics.

Source check: Source checked. Not independent scientific confirmation.

Finding: The source-specific clinical question is whether advanced carcinoma has distinctive ultrasound/clinical features; NN evaluates that finding against histopathology. This qualifies narrowly as retrospective diagnostic validation, not deployment or benefit.

Technology: retrospective n=96 histologically confirmed thyroid cancers (32 advanced, 32 non-advanced PTC, 32 non-advanced FTC); nine sonographic/clinical inputs to a neural network proof-of-concept. Results characterize advanced tumors (including size/irregularity patterns) and report classification performance (accuracy 84.4%, CI 75.5–91). S

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
retrospective n=96 histologically confirmed thyroid cancers (32 advanced, 32 non-advanced PTC, 32 non-advanced FTC); nine sonographic/clinical inputs to a neural network proof-of-concept. Results characterize advanced tumors (including size/irregularity patterns) and report classification performance (accuracy 84.4%, CI 75.5–91). Small, single-center retrospective cohort; authors call for prospective multicenter work.
Test and result
retrospective n=96 histologically confirmed thyroid cancers (32 advanced, 32 non-advanced PTC, 32 non-advanced FTC); nine sonographic/clinical inputs to a neural network proof-of-concept. Results characterize advanced tumors (including size/irregularity patterns) and report classification performance (accuracy 84.4%, CI 75.5–91). Small, single-center retrospective cohort; authors call for prospective multicenter work.
Limitation / still unknown
Methods/Results: retrospective n=96 histologically confirmed thyroid cancers (32 advanced, 32 non-advanced PTC, 32 non-advanced FTC); nine sonographic/clinical inputs to a neural network proof-of-concept. Results characterize advanced tumors (including size/irregularity patterns) and report classification performance (accuracy 84.4%, CI 75.5–91). Small, single-center retrospective cohort; authors call for prospective multicenter work.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13044-021-00107-z

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning Approach Reveals the Assembly of Activated Sludge Microbiome with Different Carbon Sources during Microcosm Startup.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: ML classification of activated-sludge microcosm data distinguishes communities grown on acetate, glucose, or starch and identifies taxa associated with these assembly patterns; the finding is specific to the measured 77-day startup experiments.

Technology: Machine-learning analysis applies the stated model to the paper’s biology dataset and evaluates the reported outcome.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Machine-learning analysis applies the stated model to the paper’s biology dataset and evaluates the reported outcome.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/microorganisms/microorganisms-09-01387/article_deploy/microorganisms-09-01387.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and internal validation of a predictive model for COVID-19 mass screening based on symptoms and a simple olfactory test

Source check: Source checked. Not independent scientific confirmation.

Finding: A classification tree combines symptoms and an olfactory test in a prospective 519-person COVID-19 cohort, with PCR reference results; sensitivity is high but specificity is only 0.39, limiting its screening interpretation.

Technology: Classification-tree screening model

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Combines symptoms and an olfactory test in a classification tree to predict COVID-19 status against PCR.
Test and result
In a prospective 519-person cohort, the classification tree combined symptoms and olfactory-test results against PCR; reported specificity was 0.39 despite high sensitivity.
Limitation / still unknown
The reported evaluation does not establish prospective benefit or performance beyond the studied setting.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-622297/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and Evaluation of a Leukemia Diagnosis System Using Deep Learning in Real Clinical Scenarios.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The paper evaluates the system on image data and describes assessment in clinical scenarios.

Technology: A staged deep-learning system detects/classifies leukocytes in bone-marrow images.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A staged deep-learning system detects/classifies leukocytes in bone-marrow images.
Limitation / still unknown
The reported assessment is limited to the study image data and described clinical scenarios; generalizable prospective clinical benefit is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fped.2021.693676

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-based glucose prediction with use of continuous glucose and physical activity monitoring data: The Maastricht Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning on Maastricht CGM data predicts 15- and 60-minute glucose values; external type-1-diabetes transfer is a small proof-of-concept. External replication and prospective utility were not established in the reviewed material.

Technology: 851 participants with CGM; random train/test split, 540 also with accelerometry, and six-person OhioT1DM transfer check. Reported CGM-only error-grid safety >99% at 15 min and >98% at 60 min in main data; type 1 external subset is very small.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
851 participants with CGM; random train/test split, 540 also with accelerometry, and six-person OhioT1DM transfer check. Reported CGM-only error-grid safety >99% at 15 min and >98% at 60 min in main data; type 1 external subset is very small.
Test and result
Machine-learning on Maastricht CGM data predicts 15- and 60-minute glucose values; external type-1-diabetes transfer is a small proof-of-concept.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0253125

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Clinical Evaluation of Automatic Coronary Artery Disease Reporting and Data System (CAD-RADS TM) in Coronary CT Angiography Using Convolutional Neural Networks

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A CNN-based system assigns CAD-RADS categories on coronary CT angiography and is clinically validated against readers and invasive angiography.

Technology: Prospective validation includes 346 patients; reported CAD sensitivity/specificity/accuracy are compared with reader performance.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Prospective validation includes 346 patients; reported CAD sensitivity/specificity/accuracy are compared with reader performance.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-622726/v1.pdf?c=1632309264000

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Handwork vs machine: a comparison of rheumatoid arthritis patient populations as identified from EHR free-text by diagnosis extraction through machine-learning or traditional criteria-based chart revi

Source check: Source checked. Not independent scientific confirmation.

Finding: An SVM extracts rheumatologist-diagnosed RA from EHR free text and the study compares the resulting cohort with criteria-based chart review; this is a specific clinical epidemiology comparison, not only a benchmark.

Technology: Support-vector machine for EHR diagnosis extraction

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Extracts rheumatologist-diagnosed rheumatoid arthritis from EHR free text for comparison with criteria-based chart review.
Test and result
The SVM-extracted EHR cohort was compared with criteria-based chart review for rheumatologist-diagnosed RA.
Limitation / still unknown
The reported evaluation does not establish prospective benefit or performance beyond the studied setting.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13075-021-02553-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

FedDPGAN: Federated Differentially Private Generative Adversarial Networks Framework for the Detection of COVID-19 Pneumonia

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: federated ResNet with DP-GAN augmentation was evaluated for COVID chest-X-ray classification and reports 94.45% IID accuracy; public class-imbalanced image data do not establish external clinical utility or real-world privacy guarantees.

Technology: Federated ResNet plus DP-GAN generation; dataset has 2,000 normal, 1,250 pneumonia, 350 COVID images resized to 28x28.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Federated ResNet plus DP-GAN generation; dataset has 2,000 normal, 1,250 pneumonia, 350 COVID images resized to 28x28.
Limitation / still unknown
Evaluation used public, class-imbalanced image data; IID accuracy does not establish external clinical utility or real-world privacy guarantees.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8204125/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Multimodal Imaging-Based Deep Learning Model for Detecting Treatment-Requiring Retinal Vascular Diseases: Model Development and Validation Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: A multimodal imaging model detected retinal vascular diseases requiring treatment, evaluated on a held-out set across several disease classes. Organization is unavailable in the deposited bibliographic metadata.

Technology: Multimodal retinal image classification model

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods describe multimodal image training/testing; Results report disease-specific AUCs and treatment-requiring disease detection.
Test and result
A multimodal imaging model detected retinal vascular diseases requiring treatment, evaluated on a held-out set across several disease classes.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8204240/

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction of Major Complications and Readmission After Lumbar Spinal Fusion: A Machine Learning-Driven Approach.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: 38,788 California lumbar-fusion patients (2015–2017), 80/20 holdout and cross-validation; XGBoost predicts 30-day major complications/readmission and identifies predictors, with AUROC/calibration comparison. Retrospective registry result, bounded to source population.

Technology: Primary Methods/Results: 38,788 California lumbar-fusion patients (2015–2017), 80/20 holdout and cross-validation; XGBoost predicts 30-day major complications/readmission and identifies predictors, with AUROC/calibration comparison. Retrospective registry result, bounded to source population.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
38,788 California lumbar-fusion patients (2015–2017), 80/20 holdout and cross-validation; XGBoost predicts 30-day major complications/readmission and identifies predictors, with AUROC/calibration comparison. Retrospective registry result, bounded to source population.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1016/j.wneu.2021.05.080 · https://pmc.ncbi.nlm.nih.gov/articles/PMC8338911/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Pulmonary embolism identification in computerized tomography pulmonary angiography scans with deep learning technologies in COVID-19 patients

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary v1 and later arXiv text report a retrospective COVID-19 CTPA PE detection/localization evaluation on 673 scans; validation is limited and does not establish deployment or clinical benefit. Catalog pulmonary-embolism records use different sources/cohorts/results.

Technology: Primary v1 and later arXiv text report a retrospective COVID-19 CTPA PE detection/localization evaluation on 673 scans; validation is limited and does not establish deployment or clinical benefit. Catalog pulmonary-embolism records use different sources/cohorts/results.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Primary v1 and later arXiv text report a retrospective COVID-19 CTPA PE detection/localization evaluation on 673 scans; validation is limited and does not establish deployment or clinical benefit. Catalog pulmonary-embolism records use different sources/cohorts/results.
Limitation / still unknown
Primary v1 and later arXiv text report a retrospective COVID-19 CTPA PE detection/localization evaluation on 673 scans; validation is limited and does not establish deployment or clinical benefit. Catalog pulmonary-embolism records use different sources/cohorts/results.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2105.11187

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Cyber-Physical System for Environmental Monitoring Based on Deep Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A deep-learning classifier assigned environmental-monitoring sensor signals to the study’s defined classes and was evaluated on the reported dataset. The result is limited to that sensing setup and dataset; it does not establish performance in other deployments.

Technology: A deep neural network classified signals from the environmental-monitoring system.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A deep neural network classified signals from the environmental-monitoring system.
Limitation / still unknown
Performance is limited to the reported environmental-monitoring dataset and sensing setup; transfer to other deployment conditions was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8197376/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Sch-net: A Deep Learning Architecture for Automatic Schizophrenia Setection

Source check: Source checked. Not independent scientific confirmation.

Finding: Sch-net uses deep learning on speech to distinguish schizophrenia patients from controls. External replication and prospective utility were not established in the reviewed material.

Technology: Speech dataset contains 28 schizophrenia patients and 28 healthy controls; the source reports ablations and comparisons with feature-engineered/classical models.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Speech dataset contains 28 schizophrenia patients and 28 healthy controls; the source reports ablations and comparisons with feature-engineered/classical models.
Test and result
Sch-net uses deep learning on speech to distinguish schizophrenia patients from controls.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-533815/v1.pdf?c=1631882321000

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning approaches to predict gestational age in normal and complicated pregnancies via urinary metabolomics analysis

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Urinary metabolomics and machine learning estimate gestational age in healthy and complicated pregnancies.

Technology: 302 pregnant participants were recruited; urinary metabolite models are evaluated against gestational age and pregnancy-status reference data.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
302 pregnant participants were recruited; urinary metabolite models are evaluated against gestational age and pregnancy-status reference data.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-541993/v1.pdf?c=1631897497000

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Using Machine Learning to Predict the Diagnosis, Management and Severity of Pediatric Appendicitis.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Single-center cohort analyses report predictor associations and model performance for three clinical outcomes.

Technology: Random-forest models use presentation variables to predict pediatric appendicitis diagnosis, management, and severity.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random-forest models use presentation variables to predict pediatric appendicitis diagnosis, management, and severity.
Limitation / still unknown
Retrospective cohort and clinical diagnosis/management labels support the bounded comparison; external validation is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fped.2021.662183

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Imaging Features by Machine Learning for Quantification of Optic Disc Changes and Impact on Choroidal Thickness in Young Myopic Patients.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: 896 young myopic participants randomly divided 627 training/269 test; LASSO-derived fundus imaging features form optic-disc indicators, evaluated against choroidal thickness and ocular features. Retrospective internal holdout only.

Technology: ML-derived imaging measure is associated with quantitative ophthalmic phenotype; no external/clinical benefit claim.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ML-derived imaging measure is associated with quantitative ophthalmic phenotype; no external/clinical benefit claim.
Limitation / still unknown
This is retrospective internal holdout evidence from 896 participants (627 training, 269 test); it does not establish external or prospective performance.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2021.657566/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A deep convolutional neural network for segmentation of whole-slide pathology images identifies novel tumour cell-perivascular niche interactions that are associated with poor survival in glioblastoma

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary study uses a deep CNN to segment glioblastoma histologic regions on whole-slide images, then integrates regional expression and single-cell evidence to report spatial tumor–vascular/perivascular biology associated with survival. The associations are exploratory and not causal. External replication and prospective utility were not established in the reviewed material.

Technology: Author-repository primary full text, Methods and Results: seven histologic regions from TCGA slides, matched RNA-seq and supporting scRNA-seq from three resections; source reports mutation-stratified survival correlations and cell-type signatures.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Author-repository primary full text, Methods and Results: seven histologic regions from TCGA slides, matched RNA-seq and supporting scRNA-seq from three resections; source reports mutation-stratified survival correlations and cell-type signatures.
Test and result
The primary study uses a deep CNN to segment glioblastoma histologic regions on whole-slide images, then integrates regional expression and single-cell evidence to report spatial tumor–vascular/perivascular biology associated with survival. The associations are exploratory and not causal.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://researchnow-admin.flinders.edu.au/ws/portalfiles/portal/47070570/s41416_021_01394_x.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Reinforcement learning-based decision support system for COVID-19.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full Methods and Results describe a Q-learning reinforcement-learning controller evaluated in COVID-19 epidemic-model simulations/case studies under hospital-saturation and intervention-cost constraints; it reports simulated control-policy outcomes. Treat as a bounded computational epidemiology finding, not observed clinical effectiveness.

Technology: Full Methods and Results describe a Q-learning reinforcement-learning controller evaluated in COVID-19 epidemic-model simulations/case studies under hospital-saturation and intervention-cost constraints; it reports simulated control-policy outcomes. Treat as a bounded computational epidemiology finding, not observed clinical effectiveness.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full Methods and Results describe a Q-learning reinforcement-learning controller evaluated in COVID-19 epidemic-model simulations/case studies under hospital-saturation and intervention-cost constraints; it reports simulated control-policy outcomes. Treat as a bounded computational epidemiology finding, not observed clinical effectiveness.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1016/j.bspc.2021.102676 · https://pmc.ncbi.nlm.nih.gov/articles/PMC8079127/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

The CARESSES Randomised Controlled Trial: Exploring the Health-Related Impact of Culturally Competent Artificial Intelligence Embedded Into Socially Assistive Robots and Tested in Older Adult Care Hom

Source check: Source checked. Not independent scientific confirmation.

Finding: A randomized study evaluated a culturally competent AI socially assistive robot intervention in older-adult care. It reports a narrow emotional-wellbeing outcome and no physical-health effect; small completer groups limit inference.

Technology: Methods and Results describe the England/Japan randomized study, three arms and 33 completers; SF-36 Emotional Wellbeing improved versus care-as-usual, while physical-health effects were not found.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Methods and Results describe the England/Japan randomized study, three arms and 33 completers; SF-36 Emotional Wellbeing improved versus care-as-usual, while physical-health effects were not found.
Test and result
Methods and Results describe the England/Japan randomized study, three arms and 33 completers; SF-36 Emotional Wellbeing improved versus care-as-usual, while physical-health effects were not found.
Limitation / still unknown
Methods and Results describe the England/Japan randomized study, three arms and 33 completers; SF-36 Emotional Wellbeing improved versus care-as-usual, while physical-health effects were not found.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s12369-021-00781-x

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A hybrid data envelopment analysis-artificial neural network prediction model for COVID-19 severity in transplant recipients.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full Springer PDF: retrospective COVID-19 transplant-recipient hospital cohort tested DEA-derived categories with ANN and comparator algorithms; results show improved ANN performance under DEA categorization. Small single-center early-pandemic cohort limits generalization. The admission is restricted to this measured result and study setting; it is not evidence of deployment benefit.

Technology: Full Springer PDF: retrospective COVID-19 transplant-recipient hospital cohort tested DEA-derived categories with ANN and comparator algorithms; results show improved ANN performance under DEA categorization. Small single-center e

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full Springer PDF: retrospective COVID-19 transplant-recipient hospital cohort tested DEA-derived categories with ANN and comparator algorithms; results show improved ANN performance under DEA categorization. Small single-center early-pandemic cohort limits generalization.
Limitation / still unknown
This was a small, single-center early-pandemic transplant cohort; it does not establish deployment benefit or generalization to later populations.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s10462-021-10008-0 · https://link.springer.com/content/pdf/10.1007/s10462-021-10008-0.pdf · https://api.crossref.org/works/10.1007%2Fs10462-021-10008-0

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine Learning with <sup>18</sup>F-Sodium Fluoride PET and Quantitative Plaque Analysis on CT Angiography for the Future Risk of Myocardial Infarction

Source check: Source checked. Not independent scientific confirmation.

Finding: XGBoost combined clinical data, CT plaque measurements and sodium fluoride PET imaging to predict myocardial infarction in a cohort with coronary artery disease.

Technology: Primary Methods/Results: XGBoost combines clinical variables, quantitative CT plaque and 18F-NaF PET in 293 CAD participants; 22 infarctions accrued over median 53 months; repeated held-out 10-fold testing compared models. Small event count limits precision and no external validation is shown.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
XGBoost combines clinical variables, quantitative CT plaque and 18F-NaF PET in 293 CAD participants; 22 infarctions accrued over median 53 months; repeated held-out 10-fold testing compared models. Small event count limits precision and no external validation is shown.
Test and result
In 293 participants, 22 infarctions accrued over a median 53 months. The study compared models by repeated held-out ten-fold testing.
Limitation / still unknown
The small event count limits precision. The source did not report an external cohort.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8717197/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning-assisted high-content analysis of pluripotent stem cell-derived embryos in vitro.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary Results and Experimental Procedures show a supervised high-content image-analysis workflow classified embryo phenotypes and screened 55 factors; BMP4 increased high-grade ITS/ETS morphogenesis and was checked in two additional stem-cell lines. Organization is unavailable in the deposited bibliographic metadata.

Technology: Supervised high-content image-analysis workflow

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Results and Experimental Procedures show a supervised high-content image-analysis workflow classified embryo phenotypes and screened 55 factors; BMP4 increased high-grade ITS/ETS morphogenesis and was checked in two additional stem-cell lines.
Test and result
The primary Results and Experimental Procedures show a supervised high-content image-analysis workflow classified embryo phenotypes and screened 55 factors; BMP4 increased high-grade ITS/ETS morphogenesis and was checked in two additional stem-cell lines.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8185434/ · https://doi.org/10.1016/j.stemcr.2021.03.018

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction and Optimization of Ion Transport Characteristics in Nanoparticle-Based Electrolytes Using Convolutional Neural Networks

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary ChemRxiv v1 full PDF Methods/Results: CNN predicts kMC diffusivity from nanoparticle morphology and topology optimization identifies simulated electrolyte structure. Computational kMC result, not experimentally realized; v1 date precedes journal version.

Technology: Primary ChemRxiv v1 full PDF Methods/Results: CNN predicts kMC diffusivity from nanoparticle morphology and topology optimization identifies simulated electrolyte structure. Computational kMC result, not experimentally realized; v1 date precedes journal version.

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Primary ChemRxiv v1 full PDF Methods/Results: CNN predicts kMC diffusivity from nanoparticle morphology and topology optimization identifies simulated electrolyte structure. Computational kMC result, not experimentally realized; v1 date precedes journal version.
Limitation / still unknown
Primary ChemRxiv v1 full PDF Methods/Results: CNN predicts kMC diffusivity from nanoparticle morphology and topology optimization identifies simulated electrolyte structure. Computational kMC result, not experimentally realized; v1 date precedes journal version. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://chemrxiv.org/engage/chemrxiv/article-details/6086f5a68827d5c8c64a1a8e

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Body fat predicts exercise capacity in persons with Type 2 Diabetes Mellitus: A machine learning approach.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning models on LOOK AHEAD participants estimate treadmill exercise capacity and identify body-fat measures as important predictors in adults with type 2 diabetes.

Technology: 1,348 patients; sex-stratified models and five-fold cross-validation; treadmill capacity is the measured outcome and body-composition features contribute predictive information. Association/prediction only; no causal or intervention claim.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
1,348 patients; sex-stratified models and five-fold cross-validation; treadmill capacity is the measured outcome and body-composition features contribute predictive information. Association/prediction only; no causal or intervention claim.
Test and result
1,348 patients; sex-stratified models and five-fold cross-validation; treadmill capacity is the measured outcome and body-composition features contribute predictive information. Association/prediction only; no causal or intervention claim.
Limitation / still unknown
Methods/Results: 1,348 patients; sex-stratified models and five-fold cross-validation; treadmill capacity is the measured outcome and body-composition features contribute predictive information. Association/prediction only; no causal or intervention claim.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0248039

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A fully automated deep learning-based network for detecting COVID-19 from a new and large lung CT scan dataset.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Full Methods and Results: COVID-CTset contains CT images from 95 COVID-19 patients and 282 normal persons; five-fold testing is described at image and patient-condition stages. A proposed ResNet feature-pyramid CNN and image-selection pipeline classify patient status; authors report correct classification for 234/245 in the patient-condition phase. This is a single-source dataset diagnostic feasibility result, not independent clinical validation.

Technology: Full Methods and Results: COVID-CTset contains CT images from 95 COVID-19 patients and 282 normal persons; five-fold testing is described at image and patient-condition stages. A proposed ResNet feature-pyramid CNN and image-selec

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Full Methods and Results: COVID-CTset contains CT images from 95 COVID-19 patients and 282 normal persons; five-fold testing is described at image and patient-condition stages. A proposed ResNet feature-pyramid CNN and image-selection pipeline classify patient status; authors report correct classification for 234/245 in the patient-condition phase. This is a single-source dataset diagnostic feasibility result, not independent clinical validation.
Limitation / still unknown
Evaluation used a single-source CT dataset (95 COVID-19 patients, 282 controls); no independent clinical validation was reported.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8011666/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

scGNN is a novel graph neural network framework for single-cell RNA-Seq analyses.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The primary study’s scGNN integrates single-cell RNA-seq modalities and reports Alzheimer’s disease-associated oligodendrocyte-subcluster shifts, gene/pathway enrichments and cell-specific regulatory programs. This is an exploratory reanalysis of existing data, not causal or independently experimentally confirmed biology.

Technology: Author/university full primary PDF, Methods/Results and Nature record checked: GSE138852, 13,214 nuclei from six AD and six control brains. Correction dated 2022-05-04 only swaps figure panels and does not alter the result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Author/university full primary PDF, Methods/Results and Nature record checked: GSE138852, 13,214 nuclei from six AD and six control brains. Correction dated 2022-05-04 only swaps figure panels and does not alter the result.
Limitation / still unknown
This is exploratory reanalysis of existing single-cell data; the reported cell programs are not causal or independently experimentally confirmed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://bpb-us-w2.wpmucdn.com/u.osu.edu/dist/0/72768/files/2021/03/Wang_et_al-2021-Nature_Communications-1.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Evaluation of an Automatic Classification Algorithm Using Convolutional Neural Networks in Oncological Positron Emission Tomography.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: CNN prototype PARS applied to PET/CT in research and routine cohorts; research cohort n=119; median Dice between automated suspicious ROI and manual ROI .65, ICC/volumetric comparisons and prognostic analyses reported. This is retrospective clinical imaging validation, not clinical benefit.

Technology: CNN performs a defined oncological PET classification/segmentation task with patient-cohort comparison; modest agreement and retrospective scope.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CNN performs a defined oncological PET classification/segmentation task with patient-cohort comparison; modest agreement and retrospective scope.
Limitation / still unknown
Methods/Results: CNN prototype PARS applied to PET/CT in research and routine cohorts; research cohort n=119; median Dice between automated suspicious ROI and manual ROI .65, ICC/volumetric comparisons and prognostic analyses reported. This is retrospective clinical imaging validation, not clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2021.628179/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Direct Attenuation Correction Using Deep Learning for Cardiac SPECT: A Feasibility Study.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: direct DCNN maps non-corrected SPECT images to CT-attenuation-corrected appearance on 100 retrospective cardiac SPECT datasets. Quantitative agreement/artifact reduction are promising, but no reader study or clinical interpretation was performed.

Technology: Primary Methods/Results: direct DCNN maps non-corrected SPECT images to CT-attenuation-corrected appearance on 100 retrospective cardiac SPECT datasets. Quantitative agreement/artifact reduction are promising, but no reader study

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
direct DCNN maps non-corrected SPECT images to CT-attenuation-corrected appearance on 100 retrospective cardiac SPECT datasets. Quantitative agreement/artifact reduction are promising, but no reader study or clinical interpretation was performed.
Limitation / still unknown
The study used 100 retrospective cardiac-SPECT datasets; no reader study or clinical interpretation was performed.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8612332/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Mining topic and sentiment dynamics in physician rating websites during the early wave of the COVID-19 pandemic: Machine learning approach.

Source check: Source checked. Not independent scientific confirmation.

Finding: TCLDA topic modeling, manual annotation and sentiment analysis of 55,612 reviews from three US physician-rating sites identified topic and emotion trends during March–June 2020. This is a specific empirical public-health/social-science result; review-site users are not representative of all patients and the short window cannot establish pandemic-caused changes. Organization is unavailable in the deposited bibliographic metadata.

Technology: TCLDA topic model and sentiment analysis

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PMC9760788 Methods/Results reviewed: corpus, preprocessing, TCLDA topic coherence plus manual annotation, and topic/emotion trends. The paper itself notes no pre-pandemic comparator and limited period; catalog search found no same result.
Test and result
TCLDA topic modeling, manual annotation and sentiment analysis of 55,612 reviews from three US physician-rating sites identified topic and emotion trends during March–June 2020. This is a specific empirical public-health/social-science result; review-site users are not representative of all patients and the short window cannot establish pandemic-caused changes.
Limitation / still unknown
This is a specific empirical public-health/social-science result; review-site users are not representative of all patients and the short window cannot establish pandemic-caused changes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9760788/

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Feedforward Artificial Neural Network-Based Colorectal Cancer Detection Using Hyperspectral Imaging: A Step towards Automatic Optical Biopsy.

Source check: Source checked. Not independent scientific confirmation.

Finding: The study included 54 consecutive surgical patients; leave-one-patient-out evaluation reported 86% sensitivity and 95% specificity for cancer/adenoma distinction. The cohort is small and the result is a diagnostic feasibility study; prospective external clinical performance is not established.

Technology: A four-layer feedforward neural network classified hyperspectral tissue annotations to distinguish colorectal cancer, adenoma and healthy samples.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A four-layer feedforward neural network classified hyperspectral tissue annotations to distinguish colorectal cancer, adenoma and healthy samples.
Test and result
The study included 54 consecutive surgical patients; leave-one-patient-out evaluation reported 86% sensitivity and 95% specificity for cancer/adenoma distinction.
Limitation / still unknown
The cohort is small and the result is a diagnostic feasibility study; prospective external clinical performance is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/articles/PMC7956537

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A High-Precision Machine Learning Algorithm to Classify Left and Right Outflow Tract Ventricular Tachycardia.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: An XGBoost ECG classifier distinguishes left- versus right-ventricular-outflow-tract ventricular tachycardia in 420 retrospective patients; held-out performance and comparison with localization approaches are reported. This is single-center diagnostic support, not clinical outcome evidence.

Technology: XGBoost classifies ECG patterns for VT origin localization.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
XGBoost classifies ECG patterns for VT origin localization.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2021.641066/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Otitis media detection using tympanic membrane images with a novel multi-class machine learning algorithm.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary study trains/evaluates multiclass ML on 956 otoscope images with otoscopist labels and reports discrimination among normal, acute/chronic otitis media, earwax and other categories. This is a retrospective image-classification result in a small single-site dataset, not prospective diagnostic validation.

Technology: Original Methods examined; learned-model operation and its link to the stated result are summarized in reason.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Original Methods examined; learned-model operation and its link to the stated result are summarized in reason.
Test and result
Original Results/proof passage examined; outcome and scope summarized in reason.
Limitation / still unknown
Original Methods/Results establish the case-specific cohort/experiment, comparator, and validation limitations as summarized in reason.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7959604/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Tool Wear Inspection and Classification by Artificial Neural Network Based on Vision System and Cutting Force Signal.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A feed-forward neural network classified milling tools as normal or worn from cutting-force features; machine-vision measurements supplied tool-wear observations for the reported manufacturing-monitoring experiment. Validation was limited to the tested setup.

Technology: Feed-forward neural network classifier using cutting-force features, with tool-wear observations from machine vision.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Feed-forward neural network classifier using cutting-force features, with tool-wear observations from machine vision.
Limitation / still unknown
Validation is confined to the reported milling setup; broader operational generalization was not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-229749/v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Structural OCT Parameters Associated with Treatment Response and Macular Neovascularization Onset in Central Serous Chorioretinopathy.

Source check: Source checked. Not independent scientific confirmation.

Finding: In a 96-eye, 36-month prospective CSC series, a decision-tree/logistic analysis associated OCT features and age with treatment response and MNV onset. The associations are cohort-specific and do not establish which treatment causes better outcomes.

Technology: a supervised decision-tree approach uses OCT measurements, hyperreflective foci, age and sex to associate imaging features with treatment response; logistic modeling analyzes MNV onset.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
a supervised decision-tree approach uses OCT measurements, hyperreflective foci, age and sex to associate imaging features with treatment response; logistic modeling analyzes MNV onset.
Test and result
among 96 treatment-naive eyes followed for 36 months, OCT hyperreflective foci and age were associated with response to eplerenone/PDT; retinal/RPE/choroidal parameters were associated with MNV risk.
Limitation / still unknown
The result is tied to prospectively observed clinical/OCT/angiography outcomes through 36 months in one case series. No held-out or external model validation is reported; approval is limited to these cohort associations, not treatment-selection performance or causal benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8079563/ · https://pubmed.ncbi.nlm.nih.gov/33606200/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep6mA: A deep learning framework for exploring similar patterns in DNA N6-methyladenine sites across different species.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The full study trains a CNN plus bidirectional LSTM on rice 6mA sites and evaluates transfer to three other plant species; it reports over 90% prediction accuracy and shared sequence-pattern/motif evidence across species. This is an exploratory in-silico biological pattern result using existing datasets, not experimental validation of methylation sites.

Technology: The source directly describes the learned operation and its link to the reported result.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
The source directly describes the learned operation and its link to the reported result.
Limitation / still unknown
The sequence-pattern findings come from existing datasets and computational transfer; methylation sites were not experimentally validated.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7924747/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and validation of a prognostic COVID-19 severity assessment (COSA) score and machine learning models for patient triage at a tertiary hospital

Source check: Source checked. Not independent scientific confirmation.

Finding: The full preprint reports a clinical score and learned models for severe COVID-19 outcomes, with prospective second-wave validation at the same hospital; scope is bounded to that cohort. Organization is unavailable in the deposited bibliographic metadata.

Technology: Support-vector and other machine-learning prognostic models

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
The paper trains support-vector and other ML models from early routine measurements to predict severe hospitalization outcomes.
Test and result
The frozen COSA score and learned models were prospectively evaluated on a second wave of 459 patients without recalibration at the same hospital; model development used the first wave (n=198).
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-165301/latest.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 4/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Early Detection of Plant Viral Disease Using Hyperspectral Imaging and Deep Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: Full paper reports ML-assisted spectral separation of asymptomatic virus-inoculated grapevines, with small-sample and weak image-test limits.

Technology: SVM/RF classifiers use spectral, vegetation-index and pixel features; 2D/3D CNNs extract image features.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
SVM/RF classifiers use spectral, vegetation-index and pixel features; 2D/3D CNNs extract image features.
Test and result
Controlled inoculation study reports asymptomatic grapevine spectral differences and classifier results; image-level held-out accuracy is only 50%.
Limitation / still unknown
The study uses healthy/inoculated vines, cross-validation, and a held-out set from the same experiment; no independent field cohort.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/1424-8220/21/3/742

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Novel Approach for Predicting Atrial Fibrillation Recurrence After Ablation Using Deep Convolutional Neural Networks by Assessing Left Atrial Curved M-Mode Speckle-Tracking Images.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Original publisher full text was reviewed through Methods and Results. The study reports a concrete AI-enabled scientific/clinical finding with identifiable cohort or dataset, comparison and quantitative result; this is eligible as an AI-enabled discovery/diagnostic result, with the stated validation limitations.

Technology: Deep CNN classifies left-atrial curved M-mode speckle-tracking images; Methods specify imaging inputs and model comparison.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Deep CNN classifies left-atrial curved M-mode speckle-tracking images; Methods specify imaging inputs and model comparison.
Limitation / still unknown
The input rationale does not state cohort size or quantitative performance; the record should remain bounded to the reviewed study and not imply clinical benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/cardiovascular-medicine/articles/10.3389/fcvm.2020.605642/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Analysis of heterogeneous genomic samples using image normalization and machine learning.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Image normalization and supervised classification of viral genomic sequence samples distinguished recent from chronic HCV infection, with reported held-out performance; the result is a specific molecular-epidemiology classification.

Technology: <h4>Background</h4>Analysis of heterogeneous populations such as viral quasispecies is one of the most challenging bioinformatics problems. Although machine learning models are becoming to be widely employed for analysis of sequen

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The study normalizes genomic sequence data from heterogeneous viral populations and applies supervised machine learning to distinguish recent from chronic HCV infection, evaluating the classification on held-out samples.
Limitation / still unknown
The classification concerns genomic samples in the reported study; broader epidemiological or clinical utility is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1186/s12864-020-6661-6 · https://europepmc.org/article/MED/33349236 · https://europepmc.org/articles/PMC7751093 · https://www.ncbi.nlm.nih.gov/research/bionlp/RESTful/pmcoa.cgi/BioC_json/PMC7751093/unicode

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of Nitrogen Fixation Genes in Lactococcus Isolated from Maize Using Population Genomics and Machine Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: Population-genomic ML analysis of Lactococcus isolates from Sierra Mixe maize links candidate genetic features to measured nitrogen-fixation phenotypes; the source reports genotype associations, not a causal gene knockout validation. External replication and prospective utility were not established in the reviewed material.

Technology: population-genomic machine-learning classifier linking Lactococcus genotypes to nitrogen-fixation phenotypes

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
population-genomic machine-learning classifier linking Lactococcus genotypes to nitrogen-fixation phenotypes
Test and result
Population-genomic ML analysis of Lactococcus isolates from Sierra Mixe maize links candidate genetic features to measured nitrogen-fixation phenotypes; the source reports genotype associations, not a causal gene knockout validation.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/microorganisms/microorganisms-08-02043/article_deploy/microorganisms-08-02043.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Probing T-cell response by sequence-based probabilistic modeling

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Restricted Boltzmann Machine models of TCR sequences and clone abundance identify antigen-specific motifs and distinguish responsive from generic clones; held-out and leave-one-out validation plus in-vitro antigen-response datasets support a specific T-cell-response modeling result.

Technology: Restricted Boltzmann Machines model T-cell receptor sequences and clone abundance to infer antigen-associated motifs and response likelihood.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Restricted Boltzmann Machines model T-cell receptor sequences and clone abundance to infer antigen-associated motifs and response likelihood.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.biorxiv.org/content/10.1101/2020.12.17.423283v1.full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Development and Prospective Validation of a Deep Learning Algorithm for Predicting Need for Mechanical Ventilation.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A deep-learning model predicted whether patients would require mechanical ventilation using clinical time-series data, with evaluation against ventilation records in an external clinical dataset. The study’s “prospective” wording does not make this a prospective frozen-model workflow evaluation; alarm-scoring choices and absence of care-outcome evidence limit interpretation.

Technology: A deep neural network analyzed clinical time-series data to estimate subsequent need for mechanical ventilation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A deep neural network analyzed clinical time-series data to estimate subsequent need for mechanical ventilation.
Limitation / still unknown
The input review notes external dataset evaluation rather than a prospective frozen-model workflow; alarm-scoring choices and absence of care-outcome evidence limit interpretation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8027289/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning Classification of Canine Behavior Using a Single Collar-Mounted Accelerometer: Real-World Validation.

Source check: Source checked. Not independent scientific confirmation.

Finding: Primary MDPI article plus linked bioRxiv v1 (2020-12-14) confirm deep-learning collar accelerometer behavior classification and real-world owner validation; v1 abstract reports >5,000 videos/2,500 dogs, >11 million device-days, 10,550 dogs/163,110 event responses, eating/drinking sensitivity/specificity and production true-positive rates. Full v1 HTML was accessed; core result-bearing claims agree with journal article. External replication and prospective utility were not established in the reviewed material.

Technology: Learned behavior classifier received real-world user validation at scale; labels are owner reports and validation covers mainly eating/drinking, not every behavior.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned behavior classifier received real-world user validation at scale; labels are owner reports and validation covers mainly eating/drinking, not every behavior.
Test and result
Primary MDPI article plus linked bioRxiv v1 (2020-12-14) confirm deep-learning collar accelerometer behavior classification and real-world owner validation; v1 abstract reports >5,000 videos/2,500 dogs, >11 million device-days, 10,550 dogs/163,110 event responses, eating/drinking sensitivity/specificity and production true-positive rates. Full v1 HTML was accessed; core result-bearing claims agree with journal article.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/2076-2615/11/6/1549

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Ensemble-CVDNet: A Deep Learning based End-to-End Classification Framework for COVID-19 Detection using Ensembles of Networks

Source check: Source checked. Not independent scientific confirmation.

Finding: fine-tuned parallel SqueezeNet/ShuffleNet/EfficientNet-B0 distinguished COVID from other pneumonia/healthy chest X-rays in a public dataset; retrospective image classification only. Organization is unavailable in the deposited bibliographic metadata.

Technology: Ensemble of fine-tuned SqueezeNet, ShuffleNet and EfficientNet-B0 CNNs

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Three fine-tuned pretrained CNNs supply parallel features to an ensemble classifier.
Test and result
fine-tuned parallel SqueezeNet/ShuffleNet/EfficientNet-B0 distinguished COVID from other pneumonia/healthy chest X-rays in a public dataset; retrospective image classification only.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/pdf/2012.09132

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Diagnostic Performance of a New Convolutional Neural Network Algorithm for Detecting Developmental Dysplasia of the Hip on Anteroposterior Radiographs.

Source check: Source checked. Not independent scientific confirmation.

Finding: CNN-based DDH detection was evaluated on held-out radiographs against radiologists; the original study reports no external validation.

Technology: CNN classified cropped hip radiographs for DDH detection.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CNN classified cropped hip radiographs for DDH detection.
Test and result
Original Results report DDH diagnostic performance comparable to experienced pediatric radiology; no external validation is claimed.
Limitation / still unknown
Held-out 10% test set and three-reader comparison; authors note no external validation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://kjronline.org/DOIx.php?id=10.3348/kjr.2020.0051

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of a Transcriptomic Prognostic Signature by Machine Learning Using a Combination of Small Cohorts of Prostate Cancer.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Machine-learning integration of prostate-cancer transcriptomic cohorts produced a prognostic gene signature and evaluated survival stratification in independent datasets; biological mechanisms remain candidates.

Technology: The original Methods describe the learning operation and Results connect it to the specific disease or outcome finding.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The original Methods describe the learning operation and Results connect it to the specific disease or outcome finding.
Limitation / still unknown
The prognostic signature was evaluated in existing transcriptomic cohorts; proposed biological mechanisms remain candidates, not established causal findings.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fgene.2020.550894 · https://europepmc.org/article/MED/33324443 · https://europepmc.org/articles/PMC7723980 · https://www.ncbi.nlm.nih.gov/research/bionlp/RESTful/pmcoa.cgi/BioC_json/PMC7723980/unicode

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prediction of Hypotension Events with Physiologic Vital Sign Signatures in The Intensive Care Unit

Source check: Source checked. Not independent scientific confirmation.

Finding: Specific clinical prediction of observed ICU hypotension, with operational alert burden and temporal validation, qualifies as retrospective clinical prediction validation. Preserve single-center limitation and absence of impact study. External replication and prospective utility were not established in the reviewed material.

Technology: MIMIC-III retrospective ICU data; RF on minute-level vital signs/30-minute overlapping windows predicts initial hypotension (5 of 10 readings SBP≤90 and MAP≤60); temporal split validation. Cohort 1,307 subjects/1,580 stays with events and 1,619/2,279 without; AUROC 0.88 at 60 min, sensitivity 92.4% with 0.79 alerts/subject-hour. Single-center existing-data prediction; not evidence alerts improve outcomes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
MIMIC-III retrospective ICU data; RF on minute-level vital signs/30-minute overlapping windows predicts initial hypotension (5 of 10 readings SBP≤90 and MAP≤60); temporal split validation. Cohort 1,307 subjects/1,580 stays with events and 1,619/2,279 without; AUROC 0.88 at 60 min, sensitivity 92.4% with 0.79 alerts/subject-hour. Single-center existing-data prediction; not evidence alerts improve outcomes.
Test and result
Specific clinical prediction of observed ICU hypotension, with operational alert burden and temporal validation, qualifies as retrospective clinical prediction validation. Preserve single-center limitation and absence of impact study.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13054-020-03379-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning-based classification and segmentation of retinal cavitations on optical coherence tomography images of macular telangiectasia type 2.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Two CNNs classified and segmented MacTel2 retinal cavitations, with outputs checked against manual measurements and longitudinal trial data.

Technology: 99 eyes from 67 trial participants and CNN procedures; Results report manual-volume agreement and time-course comparisons.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
99 eyes from 67 trial participants and CNN procedures; Results report manual-volume agreement and time-course comparisons.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pmc&id=PMC8365558

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Network-based machine learning in colorectal and bladder organoid models predicts anti-cancer drug efficacy in patients.

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary study combines network-based pathway selection and deep neural response models in colorectal and bladder organoids, then reports associations between predicted drug-response biomarkers and patient survival, with negative controls and cell-line validation. This is retrospective translational evidence, not prospective treatment-selection utility.

Technology: PMC/publisher full Methods and Results: 19 colorectal and 9 bladder organoid samples; survival associations in treated cohorts (114 CRC/5-FU; 77 bladder/cisplatin), absent in untreated negative controls, plus isogenic sensitive/resistant cell-line validation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
PMC/publisher full Methods and Results: 19 colorectal and 9 bladder organoid samples; survival associations in treated cohorts (114 CRC/5-FU; 77 bladder/cisplatin), absent in untreated negative controls, plus isogenic sensitive/resistant cell-line validation.
Test and result
PMC/publisher full Methods and Results: 19 colorectal and 9 bladder organoid samples; survival associations in treated cohorts (114 CRC/5-FU; 77 bladder/cisplatin), absent in untreated negative controls, plus isogenic sensitive/resistant cell-line validation.
Limitation / still unknown
PMC/publisher full Methods and Results: 19 colorectal and 9 bladder organoid samples; survival associations in treated cohorts (114 CRC/5-FU; 77 bladder/cisplatin), absent in untreated negative controls, plus isogenic sensitive/resistant cell-line validation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7599252/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Feasibility of the deep learning method for estimating the ventilatory threshold with electrocardiography data.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A deep-learning model estimated ventilatory threshold from single-lead ECG in 260 cardiovascular patients and was compared with cardiopulmonary exercise testing. This is a measurement-feasibility result in the reported cohort, not evidence of interchangeable clinical decisions.

Technology: A deep neural network mapped single-lead ECG signals to ventilatory-threshold estimates.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A deep neural network mapped single-lead ECG signals to ventilatory-threshold estimates.
Limitation / still unknown
The feasibility evaluation involved 260 cardiovascular patients against cardiopulmonary exercise testing; it does not establish interchangeable clinical decisions or patient benefit.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41746-020-00348-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning-based digitization of prostate brachytherapy needles in ultrasound images

Source check: Source checked. Not independent scientific confirmation.

Finding: 3D CNN digitizes prostate brachytherapy needles on ultrasound; evaluation compares with physicist annotations and manual workflow. First publication date 2020-10-27 is corroborated by the primary bibliographic entry. Validation approach and cohort/reference comparison were checked in Methods/Results; conclusion is limited to this study.

Technology: 3D CNN digitizes prostate brachytherapy needles on ultrasound; evaluation compares with physicist annotations and manual workflow.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
3D CNN digitizes prostate brachytherapy needles on ultrasound; evaluation compares with physicist annotations and manual workflow.
Test and result
3D CNN digitizes prostate brachytherapy needles on ultrasound; evaluation compares with physicist annotations and manual workflow.
Limitation / still unknown
Validation approach and cohort/reference comparison were checked in Methods/Results; conclusion is limited to this study. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7821271/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning predicts mortality in septic patients using only routinely available ABG variables: a multi-centre evaluation.

Source check: Source checked. Not independent scientific confirmation.

Finding: An LSTM using arterial blood gas measurements from the first 48 ICU hours predicted 96-hour mortality among septic ICU stays, with AUC 0.88 in multicenter eICU and 0.85 in single-center MIMIC-III, compared with logistic regression and SOFA. This supports a bounded retrospective prognostic result.

Technology: White Rose accepted version Methods/Results read: retrospectively selected eICU (multi-center) and MIMIC-III cohorts, ABG inputs, LSTM and baseline comparisons; no prospective outcome/clinical benefit claim. No same-result catalog match.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
White Rose accepted version Methods/Results read: retrospectively selected eICU (multi-center) and MIMIC-III cohorts, ABG inputs, LSTM and baseline comparisons; no prospective outcome/clinical benefit claim. No same-result catalog match.
Test and result
White Rose accepted version Methods/Results read: retrospectively selected eICU (multi-center) and MIMIC-III cohorts, ABG inputs, LSTM and baseline comparisons; no prospective outcome/clinical benefit claim. No same-result catalog match.
Limitation / still unknown
White Rose accepted version Methods/Results read: retrospectively selected eICU (multi-center) and MIMIC-III cohorts, ABG inputs, LSTM and baseline comparisons; no prospective outcome/clinical benefit claim. No same-result catalog match.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://eprints.whiterose.ac.uk/id/eprint/193813/1/Machine_learning_predicts_mortality_in_septic_patients_using_only_routinely_available_ABG_variables_a_multi-centre_evaluation.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Internet of Things and Machine Learning for Healthy Ageing: Identifying the Early Signs of Dementia.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: In a monitored older-adult cohort, machine-learning analysis of smart-home activity patterns was used to identify dementia onset, with classification accuracy reported by the study. The result is bounded to the monitored sample and does not establish general clinical screening performance.

Technology: Machine-learning classifiers analyzed smart-home activity features to detect dementia onset.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning classifiers analyzed smart-home activity features to detect dementia onset.
Limitation / still unknown
The detection result is limited to monitored older adults and the reported validation; broader screening performance and clinical benefit are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7660294/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning for Distinguishing Normal versus Abnormal Chest Radiographs and Generalization to Unseen Diseases

Source check: Source checked. Not independent scientific confirmation.

Finding: deep-learning CXR classifier was evaluated on large training cohorts and six external datasets; discrimination varied by disease and COVID AUC was limited, while triage time reduction was simulated only. The result-bearing arXiv v1 was posted on 2020-10-22; the later same-result Scientific Reports article is linked in paper_url.

Technology: Deep-learning chest-radiograph classifier

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A deep-learning chest-radiograph classifier was trained on data from five Indian hospital clusters and tested on six external datasets to distinguish normal from abnormal images and assess unseen-disease performance.
Test and result
deep-learning CXR classifier was evaluated on large training cohorts and six external datasets; discrimination varied by disease and COVID AUC was limited, while triage time reduction was simulated only.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2010.11375

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Medical Informatics Platform (MIP): A Pilot Study Across Clinical Italian Cohorts.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A distributed clinical informatics platform was assessed across Italian memory-clinic cohorts for Alzheimer-related classification and physician diagnostic confidence.

Technology: Methods/Results cover distributed clinical cohorts, ML classification from clinical/CSF/MRI data and physician confidence effects.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods/Results cover distributed clinical cohorts, ML classification from clinical/CSF/MRI data and physician confidence effects.
Limitation / still unknown
This was a pilot across the studied Italian memory-clinic cohorts; it does not establish general clinical deployment or improved diagnostic outcomes.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7538836/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Using Machine Learning to Develop a Novel COVID-19 Vulnerability Index (C19VI)

Source check: Source checked. Not independent scientific confirmation.

Finding: A Random Forest COVID-19 Vulnerability Index ranked county-level vulnerability and highlighted socioeconomic disparities; the article uses overdose/case indicators and ecological county data, so the result is a proxy-based index rather than a causal or individual-risk claim. Organization is unavailable in the deposited bibliographic metadata.

Technology: Random-forest COVID-19 Vulnerability Index

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ArXiv v1 abstract and 2021 journal full text checked: RF model built from CDC themes; journal Results report county rankings, feature contributions, and held-out AUC 0.84. Journal and preprint contain the same represented index result.
Test and result
A Random Forest COVID-19 Vulnerability Index ranked county-level vulnerability and highlighted socioeconomic disparities; the article uses overdose/case indicators and ecological county data, so the result is a proxy-based index rather than a causal or individual-risk claim.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2009.10808

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning to Estimate RECIST in Patients with NSCLC Treated with PD-1 Blockade

Source check: Source checked. Not independent scientific confirmation.

Finding: A supervised deep-learning NLP model estimated RECIST response/progression from NSCLC radiology reports, enabling scalable measurement of treatment-specific outcomes. Blinded internal holdout plus independent external institution and expert RECIST ground truth; authors limit use to retrospective research, not real-time care.

Technology: A fully connected network used tokenized longitudinal report text and radiologist RECIST labels to predict response and progression timing.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A fully connected network used tokenized longitudinal report text and radiologist RECIST labels to predict response and progression timing.
Test and result
Internal and external test cohorts showed response classification and progression-date agreement with radiologist RECIST; external test included 97 patients.
Limitation / still unknown
Blinded internal holdout plus independent external institution and expert RECIST ground truth; authors limit use to retrospective research, not real-time care. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7981277/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC7981277/#methods · https://pmc.ncbi.nlm.nih.gov/articles/PMC7981277/#results

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Early Prediction of Cognitive Deficit in Very Preterm Infants Using Brain Structural Connectome With Transfer Learning Enhanced Deep Convolutional Neural Networks.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Transfer-learning-enhanced deep networks using preterm-infant brain structural connectomes predicted cognitive deficits at follow-up, linking neonatal imaging to later developmental outcomes.

Technology: The source documents the learned imaging or signal-analysis operation and its relation to the reported result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The source documents the learned imaging or signal-analysis operation and its relation to the reported result.
Limitation / still unknown
Prediction concerns follow-up cognitive deficits in the studied very-preterm infant cohort; external validation and improved developmental outcomes are not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fnins.2020.00858 · https://europepmc.org/article/MED/33041749 · https://europepmc.org/articles/PMC7530168

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Tumor immune profiles noninvasively estimated by FDG PET with deep learning correlate with immunotherapy response in lung adenocarcinoma.

Source check: Source checked. Not independent scientific confirmation.

Finding: Publisher PDF Methods/Results report model training on FDG-PET to estimate tumor immune CytAct and tests in separate SNUH, TCGA and immunotherapy cohorts; modest retrospective associations bound the claim. Organization is unavailable in the deposited bibliographic metadata.

Technology: Deep-learning model mapping FDG-PET to tumor immune activity

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep model estimates RNA-seq-derived immune activation from PET. Independent-cohort correlations and immunotherapy response associations were observed, but cohorts are small and retrospective.
Test and result
Publisher PDF Methods/Results report model training on FDG-PET to estimate tumor immune CytAct and tests in separate SNUH, TCGA and immunotherapy cohorts; modest retrospective associations bound the claim.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.thno.org/v10p10838.pdf

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning identifies morphological determinants of sex differences in the pre-adolescent brain.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The checked primary abstract reports this specific result: Deep learning on ABCD structural MRI identifies a cerebellar and subcortical sex-difference pattern tied to puberty, a distinct field result with an identified enabling AI role and cohort-scale evaluation. The abstract identifies the model or learned operation and an observed evaluation; the conclusion is limited to the reported study setting. Original PMC full-text Methods and Results were checked; date 2020-08-22 reconciled to the primary record.

Technology: Publisher/PMC Methods establish the learned model and its operation in the specific study.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher/PMC Methods establish the learned model and its operation in the specific study.
Limitation / still unknown
The result is an association in the ABCD structural-MRI cohort; it does not establish causal effects of puberty or generalization beyond the studied sample.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7780846/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Using Machine Learning to Unveil Demographic and Clinical Features of COVID-19 Symptomatic and Asymptomatic Patients

Source check: Source checked. Not independent scientific confirmation.

Finding: A multi-algorithm pipeline identifies demographic and clinical features associated with symptomatic versus asymptomatic COVID-19. Organization is unavailable in the deposited bibliographic metadata.

Technology: Support-vector and other machine-learning classifiers for COVID-19 symptoms

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Clinical data from 3,995 patients in Kuwait are modeled; SVM reports AUC 0.77 and accuracy 70.01%, with feature importance results.
Test and result
A multi-algorithm pipeline identifies demographic and clinical features associated with symptomatic versus asymptomatic COVID-19.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.researchsquare.com/article/rs-52330/v1.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Explainable machine learning for materials discovery: predicting the potentially formable Nd-Fe-B crystal structures and extracting the structure-stability relationship.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: ArXiv v1 directly reports ML stability prediction and structure–stability analysis alongside 20 DFT-screened Nd–Fe–B candidates; candidate formability remains theoretical, not experimental.

Technology: Machine-learning models predict potentially formable Nd-Fe-B compounds and analyze structure-stability relationships alongside DFT screening.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning models predict potentially formable Nd-Fe-B compounds and analyze structure-stability relationships alongside DFT screening.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2008.08793

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Differentiating novel coronavirus pneumonia from general pneumonia based on machine learning

Source check: Source checked. Not independent scientific confirmation.

Finding: The paper reports a specific 100-patient COVID-19 versus general-pneumonia CT classification, with ReliefF-selected image features and 10-fold classifier evaluation; the single-center small cohort limits transportability. Methods/Results: 73 COVID and 27 general pneumonia cases; CT GGO ROI, 34 texture features, ReliefF selection, bagged-tree/SVM/LR/DT/KNN, tenfold CV; best reported ACC 94.16%, sensitivity 88.62%, specificity 100%, AUC .99.

Technology: 73 COVID and 27 general pneumonia cases; CT GGO ROI, 34 texture features, ReliefF selection, bagged-tree/SVM/LR/DT/KNN, tenfold CV; best reported ACC 94.16%, sensitivity 88.62%, specificity 100%, AUC .99.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
73 COVID and 27 general pneumonia cases; CT GGO ROI, 34 texture features, ReliefF selection, bagged-tree/SVM/LR/DT/KNN, tenfold CV; best reported ACC 94.16%, sensitivity 88.62%, specificity 100%, AUC .99.
Test and result
73 COVID and 27 general pneumonia cases; CT GGO ROI, 34 texture features, ReliefF selection, bagged-tree/SVM/LR/DT/KNN, tenfold CV; best reported ACC 94.16%, sensitivity 88.62%, specificity 100%, AUC .99.
Limitation / still unknown
Methods/Results: 73 COVID and 27 general pneumonia cases; CT GGO ROI, 34 texture features, ReliefF selection, bagged-tree/SVM/LR/DT/KNN, tenfold CV; best reported ACC 94.16%, sensitivity 88.62%, specificity 100%, AUC .99. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s12938-020-00809-9

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Stratification of portal vein-invaded hepatocellular carcinoma treated with transarterial chemoembolization monotherapy.

Source check: Source checked. Not independent scientific confirmation.

Finding: The checked primary abstract reports this specific result: Unverified Haiku assessment. Retrospective two-hospital cohort of HCC with portal vein invasion treated with TACE; ANN and prognostic model validated on an independent cohort. The abstract identifies the model or learned operation and an observed evaluation; the conclusion is limited to the reported study setting. Original PMC full-text Methods and Results were checked; date 2020-08-16 reconciled to the primary record.

Technology: Publisher/PMC Methods establish the learned model and its operation in the specific study.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher/PMC Methods establish the learned model and its operation in the specific study.
Test and result
Publisher/PMC Results report the represented result and its cohort-specific scope.
Limitation / still unknown
Validation design and reported performance/comparison were checked in the full text and are stated narrowly.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8562278/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A Novel Hybrid Model Based on a Feedforward Neural Network and One Step Secant Algorithm for Prediction of Load-Bearing Capacity of Rectangular Concrete-Filled Steel Tube Columns.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A hybrid feed-forward neural network and one-step secant optimization model estimated load-bearing capacity of rectangular concrete-filled steel-tube columns from 422 experimental records and was assessed against measured capacities.

Technology: Hybrid feed-forward neural network with one-step secant optimization, trained on experimental column-capacity data.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Hybrid feed-forward neural network with one-step secant optimization, trained on experimental column-capacity data.
Limitation / still unknown
The model was assessed against the 422-instance experimental database; validation outside the represented column designs and test conditions is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/molecules/molecules-25-03486/article_deploy/molecules-25-03486.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Forecasting Corn Yield With Machine Learning Ensembles.

Source check: Source checked. Not independent scientific confirmation.

Finding: Machine-learning ensembles forecast county-level corn yield across three US states and compare held-out time periods against base learners and published yield models; environmental predictors support a bounded regional crop-yield forecasting result. External replication and prospective utility were not established in the reviewed material.

Technology: Weighted ensembles combine ML regressors and environment/management covariates for county-level yield prediction.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Weighted ensembles combine ML regressors and environment/management covariates for county-level yield prediction.
Test and result
Machine-learning ensembles forecast county-level corn yield across three US states and compare held-out time periods against base learners and published yield models; environmental predictors support a bounded regional crop-yield forecasting result.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2020.01120/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identifying prostate cancer and its clinical risk in asymptomatic men using machine learning of high dimensional peripheral blood flow cytometric natural killer cell subset phenotyping data.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Institutional repository copy of the primary eLife manuscript supplies Methods/Results for NK-cell flow-cytometry features and model evaluation in asymptomatic and clinical cohorts; the result is a small-cohort classification proof of concept.

Technology: Machine-learning models classify prostate-cancer status and clinical risk from peripheral natural-killer-cell phenotyping.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning models classify prostate-cancer status and clinical risk from peripheral natural-killer-cell phenotyping.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://irep.ntu.ac.uk/id/eprint/40312/1/1349937_Pockley.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Drawing insights from COVID-19-infected patients using CT scan images and machine learning techniques: a study on 200 patients

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary source Methods/Results confirm the represented result: Retrospective ML (Azure Custom Vision) classification of COVID-19 CT scans in 200 patients with about 91% accuracy; AI is the enabling classifier with a clinically specific retrospective evaluation. The source reports the defined cohort/experiment and its reference or test comparison; interpretation is restricted to that study design.

Technology: Retrospective ML (Azure Custom Vision) classification of COVID-19 CT scans in 200 patients with about 91% accuracy; AI is the enabling classifier with a clinically specific retrospective evaluation.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Retrospective ML (Azure Custom Vision) classification of COVID-19 CT scans in 200 patients with about 91% accuracy; AI is the enabling classifier with a clinically specific retrospective evaluation.
Test and result
Retrospective ML (Azure Custom Vision) classification of COVID-19 CT scans in 200 patients with about 91% accuracy; AI is the enabling classifier with a clinically specific retrospective evaluation.
Limitation / still unknown
The source reports the defined cohort/experiment and its reference or test comparison; interpretation is restricted to that study design. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.1007/s11356-020-10133-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prospective evaluation of an artificial intelligence-enabled algorithm for automated diabetic retinopathy screening of 30 000 patients.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A prospective AI-enabled algorithm triages diabetic-retinopathy screening images in a 30,000-patient programme against human grading.

Technology: The source reports prospective deployment/test in the English screening programme and PPV/NPV against graders; most algorithm-negative cases lacked referable retinopathy.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The source reports prospective deployment/test in the English screening programme and PPV/NPV against graders; most algorithm-negative cases lacked referable retinopathy.
Limitation / still unknown
The prospective program compared AI triage with human grading in 30,000 patients; the source does not establish improved patient outcomes from AI use.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/article/MED/32606081 · https://pmc.ncbi.nlm.nih.gov/articles/PMC8077216/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Concise Polygenic Models for Cancer-Specific Identification of Drug-Sensitive Tumors from Their Multi-Omics Profiles.

Source check: Source checked. Not independent scientific confirmation.

Finding: XGBoost with feature selection yields 118 validated cell-line drug-response models for 62 drugs across five cancer types using four omics profiles; validation is on the source pharmacogenomic datasets, not clinical patient response. External replication and prospective utility were not established in the reviewed material.

Technology: XGBoost polygenic models with feature selection trained on multi-omics pharmacogenomic profiles

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
XGBoost polygenic models with feature selection trained on multi-omics pharmacogenomic profiles
Test and result
XGBoost with feature selection yields 118 validated cell-line drug-response models for 62 drugs across five cancer types using four omics profiles; validation is on the source pharmacogenomic datasets, not clinical patient response.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://mdpi-res.com/d_attachment/biomolecules/biomolecules-10-00963/article_deploy/biomolecules-10-00963.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Identification of Bacterial Blight Resistant Rice Seeds Using Terahertz Imaging and Hyperspectral Imaging Combined With Convolutional Neural Network.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Terahertz and hyperspectral measurements with CNNs classify bacterial-blight-resistant rice seeds against susceptible seeds; the paper reports held-out testing and comparisons with SVM/RF/PLS-DA. This is a specific seed-screening result, with same-pool sampling limiting independent generalization.

Technology: CNNs classify bacterial-blight-resistant and susceptible rice seeds from terahertz and hyperspectral measurements.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CNNs classify bacterial-blight-resistant and susceptible rice seeds from terahertz and hyperspectral measurements.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2020.00821/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

An Unsupervised Machine Learning Approach to Assess the ZIP Code Level Impact of COVID-19 in NYC

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: NYC health-data clustering identifies differentiated ZIP-code case-growth patterns that agree with Health Department statistics; associations are observational, not causal.

Technology: Lasso/RReliefF feature selection and k-means cluster NYC ZIP codes using demographic, mobility and case-growth features.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Lasso/RReliefF feature selection and k-means cluster NYC ZIP codes using demographic, mobility and case-growth features.
Limitation / still unknown
The ZIP-code patterns are observational and do not establish causal effects of COVID-19 or individual-level risk.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2006.08361

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep learning of contagion dynamics on complex networks

Source check: Source checked. Not independent scientific confirmation.

Finding: GNN learned contagion dynamics and was evaluated on synthetic systems and Spain COVID outbreak data, yielding computational dynamics/source-inference results; no intervention or causal public-health claim. External replication and prospective utility were not established in the reviewed material.

Technology: GNN learns local contagion mechanism from network time series.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
GNN learns local contagion mechanism from network time series.
Test and result
GNN learned contagion dynamics and was evaluated on synthetic systems and Spain COVID outbreak data, yielding computational dynamics/source-inference results; no intervention or causal public-health claim.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2006.05410

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Application of explainable ensemble artificial intelligence model to categorization of hemodialysis-patient and treatment using nationwide-real-world data in Japan.

Source check: Source checked. Not independent scientific confirmation.

Finding: K-means plus SVM ensemble on nationwide Japanese dialysis registry data defines patient clusters and predicts one-year mortality, with separate development/validation data. Organization is unavailable in the deposited bibliographic metadata.

Technology: K-means clustering and support-vector-machine ensemble

Publication: Publication status unspecified in annotation · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
nationwide JRDR cohort split into development and validation datasets; K-means patient clusters plus SVM ensemble compared with SVM, deep learning and logistic regression. Reported validation accuracy 0.948; observational prediction, not treatment effect.
Test and result
K-means plus SVM ensemble on nationwide Japanese dialysis registry data defines patient clusters and predicts one-year mortality, with separate development/validation data.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0233491

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Phenotypic Polymyxin Resistance in Klebsiella pneumoniae through Machine Learning Analysis of Genomic Data.

Source check: Source checked. Not independent scientific confirmation.

Finding: The study evaluated genomic prediction on a 619-genome set with phenotype labels and reported predictive performance and genomic determinants. The model is limited to the studied strains and phenotype data; it does not establish clinical utility or resistance mechanisms beyond the tested evidence.

Technology: Machine-learning models used Klebsiella pneumoniae genomic features to predict phenotypic polymyxin resistance and were compared with a rule-based baseline.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning models used Klebsiella pneumoniae genomic features to predict phenotypic polymyxin resistance and were compared with a rule-based baseline.
Test and result
The study evaluated genomic prediction on a 619-genome set with phenotype labels and reported predictive performance and genomic determinants.
Limitation / still unknown
The model is limited to the studied strains and phenotype data; it does not establish clinical utility or resistance mechanisms beyond the tested evidence.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.asm.org/doi/10.1128/mSystems.00656-19

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Pilot study: can machine learning analyses of movement discriminate between leg movements in sleep (LMS) with vs. without cortical arousals?

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: The original full text supports the bounded AI-enabled finding: A supervised movement classifier distinguishes sleep-related leg movements with versus without cortical arousals; the Results report event-detection agreement against scored sleep recordings.

Technology: A supervised movement classifier distinguishes sleep-related leg movements with and without cortical arousals.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A supervised movement classifier distinguishes sleep-related leg movements with and without cortical arousals.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1007/s11325-020-02100-6

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Combining gene expression profiling and machine learning to diagnose B-cell non-Hodgkin lymphoma

Source check: Source checked. Not independent scientific confirmation.

Finding: The primary paper uses a random-forest classifier on targeted gene-expression profiling to classify B-cell non-Hodgkin lymphoma subtypes and reports concordance against expert pathology and assay references. This is a specific hematopathology diagnostic result; the authors frame it as a complement to pathology. Publisher/author-hosted full text, Methods and Results: RT-MLPSeq 137-marker assay, 510 biopsies and seven subtype classes; validation against expert pathology and NanoString/immunostain references.

Technology: Publisher/author-hosted full text, Methods and Results: RT-MLPSeq 137-marker assay, 510 biopsies and seven subtype classes; validation against expert pathology and NanoString/immunostain references.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Publisher/author-hosted full text, Methods and Results: RT-MLPSeq 137-marker assay, 510 biopsies and seven subtype classes; validation against expert pathology and NanoString/immunostain references.
Test and result
Publisher/author-hosted full text, Methods and Results: RT-MLPSeq 137-marker assay, 510 biopsies and seven subtype classes; validation against expert pathology and NanoString/immunostain references.
Limitation / still unknown
Publisher/author-hosted full text, Methods and Results: RT-MLPSeq 137-marker assay, 510 biopsies and seven subtype classes; validation against expert pathology and NanoString/immunostain references. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.nature.com/articles/s41408-020-0322-5.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Machine learning provides evidence that stroke risk is not linear: The non-linear Framingham stroke risk score.

Source check: Source checked. Not independent scientific confirmation.

Finding: An optimal classification tree using Framingham cohorts yields nonlinear stroke-risk profiles and is compared with the revised Framingham score on held-out cohorts.

Technology: tree model trained on 14,195 Framingham offspring examinations; paper evaluates held-out Framingham and external cohorts, with calibration and discrimination comparisons. Retrospective risk-model evidence, not prospective clinical utility.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
tree model trained on 14,195 Framingham offspring examinations; paper evaluates held-out Framingham and external cohorts, with calibration and discrimination comparisons. Retrospective risk-model evidence, not prospective clinical utility.
Test and result
tree model trained on 14,195 Framingham offspring examinations; paper evaluates held-out Framingham and external cohorts, with calibration and discrimination comparisons. Retrospective risk-model evidence, not prospective clinical utility.
Limitation / still unknown
Methods/Results: tree model trained on 14,195 Framingham offspring examinations; paper evaluates held-out Framingham and external cohorts, with calibration and discrimination comparisons. Retrospective risk-model evidence, not prospective clinical utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0232414

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Prognostic Significance of Immune Cell Populations Identified by Machine Learning in Colorectal Cancer Using Routine Hematoxylin and Eosin-Stained Sections.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Pathologist-supervised ML image analysis quantified H&E immune-cell densities/spatial patterns in colorectal cancer and linked stromal lymphocytes/eosinophils to survival.

Technology: Supervised image classifiers identified four immune-cell types and tumor/stroma areas; spatial metrics quantified cell proximity.

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Supervised image classifiers identified four immune-cell types and tumor/stroma areas; spatial metrics quantified cell proximity.
Limitation / still unknown
The survival associations derive from pathologist-supervised image analysis in the studied colorectal-cancer cohort; they do not establish causal effects or prospective utility.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://oulurepo.oulu.fi/bitstream/handle/10024/30228/nbnfi-fe2021042611913.pdf

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Artificial Neural Network and Cox Regression Models for Predicting Mortality after Hip Fracture Surgery: A Population-Based Comparison.

Source check: Source checked. Not independent scientific confirmation.

Finding: Retrospective population cohort of 10,534 hip-fracture procedures compares ANN with Cox regression; holdout test AUROC .93 vs .88 and sensitivity .96 vs .92 for mortality. External replication and prospective utility were not established in the reviewed material.

Technology: Retrospective population cohort of 10,534 hip-fracture procedures compares ANN with Cox regression; holdout test AUROC .93 vs .88 and sensitivity .96 vs .92 for mortality.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Retrospective population cohort of 10,534 hip-fracture procedures compares ANN with Cox regression; holdout test AUROC .93 vs .88 and sensitivity .96 vs .92 for mortality.
Test and result
Retrospective population cohort of 10,534 hip-fracture procedures compares ANN with Cox regression; holdout test AUROC .93 vs .88 and sensitivity .96 vs .92 for mortality.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3390/medicina56050243

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Virtual mouse brain histology from multi-contrast MRI via deep learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: Held-out mouse samples and co-registered histology comparisons showed predicted maps tracked tissue patterns and distinguished shiverer from control tissue. This is an ex vivo mouse proof of concept limited by tissue, sample and MRI-resolution conditions; it does not establish human in-vivo histology. The recorded date is the first located result-bearing version (2020-05-03); the eLife article is the later journal version.

Technology: A convolutional neural network mapped multi-contrast mouse-brain MRI signals to predicted histology-related contrasts.

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A convolutional neural network mapped multi-contrast mouse-brain MRI signals to predicted histology-related contrasts.
Test and result
Held-out mouse samples and co-registered histology comparisons showed predicted maps tracked tissue patterns and distinguished shiverer from control tissue.
Limitation / still unknown
This is an ex vivo mouse proof of concept limited by tissue, sample and MRI-resolution conditions; it does not establish human in-vivo histology.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://elifesciences.org/articles/72331

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Towards an Effective and Efficient Deep Learning Model for COVID-19 Patterns Detection in X-ray Images

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Primary journal article and arXiv v1 report deep-learning COVID-19 X-ray classification on a small, imbalanced dataset with a five-patient held-out test set; authors caution that broader heterogeneous validation is needed before clinical use.

Technology: Primary journal article and arXiv v1 report deep-learning COVID-19 X-ray classification on a small, imbalanced dataset with a five-patient held-out test set; authors caution that broader heterogeneous validation is needed before clinical use.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Primary journal article and arXiv v1 report deep-learning COVID-19 X-ray classification on a small, imbalanced dataset with a five-patient held-out test set; authors caution that broader heterogeneous validation is needed before clinical use.
Limitation / still unknown
Primary journal article and arXiv v1 report deep-learning COVID-19 X-ray classification on a small, imbalanced dataset with a five-patient held-out test set; authors caution that broader heterogeneous validation is needed before clinical use.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC8055440/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Diagnosing COVID-19 Pneumonia from X-Ray and CT Images using Deep Learning and Transfer Learning Algorithms

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: CNN and modified AlexNet were tested for COVID classification on mixed-source X-ray/CT images; test sensitivities are promising but based on only 50 X-rays and 17 CTs, with no clinical utility established.

Technology: CNN and modified AlexNet classify mixed-source COVID/normal images.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CNN and modified AlexNet classify mixed-source COVID/normal images.
Limitation / still unknown
The test sets were small (50 X-rays and 17 CTs) and mixed-source; no clinical utility or deployment was established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2004.00038 · https://doi.org/10.1117/12.2588672 · https://arxiv.org/pdf/2004.00038

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Deep Learning Radiomics Based on Contrast-Enhanced Ultrasound Might Optimize Curative Treatments for Very-Early or Early-Stage Hepatocellular Carcinoma Patients.

Source check: Source checked. Not independent scientific confirmation.

Finding: Cox-CNN radiomics extracted CEUS features to estimate treatment-specific progression-free survival in early HCC and identify candidate treatment-selection differences. Organization is unavailable in the deposited bibliographic metadata.

Technology: CNN-based Cox model for CEUS radiomics and treatment outcomes

Publication: Publication status unspecified in annotation · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
A CNN-based Cox model learned spatiotemporal CEUS features separately for RFA and surgical resection, with radiomics signatures added to nomograms.
Test and result
Cox-CNN radiomics extracted CEUS features to estimate treatment-specific progression-free survival in early HCC and identify candidate treatment-selection differences.
Limitation / still unknown
The reviewed study does not establish validation or benefit beyond its reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://karger.com/lic/article/9/4/397/202376/Deep-Learning-Radiomics-Based-on-Contrast-Enhanced

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Peak Outpatient and Emergency Department Visit Forecasting for Patients With Chronic Respiratory Diseases Using Machine Learning Methods: Retrospective Cohort Study.

Source check: Source checked. Not independent scientific confirmation.

Finding: A retrospective cohort included 13,218 visits; random forests achieved the best reported AUC (0.809) in the combined dataset. The retrospective forecast does not establish improved service delivery or transport to other health systems.

Technology: Machine-learning models forecasted peak outpatient and emergency department demand for patients with chronic respiratory diseases using visit, weather and pollution features.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Machine-learning models forecasted peak outpatient and emergency department demand for patients with chronic respiratory diseases using visit, weather and pollution features.
Test and result
A retrospective cohort included 13,218 visits; random forests achieved the best reported AUC (0.809) in the combined dataset.
Limitation / still unknown
The retrospective forecast does not establish improved service delivery or transport to other health systems.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.jmir.org/2020/3/e13075/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

CBCT-based synthetic CT generation using deep-attention cycleGAN for pancreatic adaptive radiotherapy.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: Deep-attention CycleGAN generates synthetic CT from CBCT for pancreatic adaptive radiotherapy; the paper evaluates image and dosimetric agreement in the reported patients. First publication date 2020-03-28 is corroborated by the primary bibliographic entry.

Technology: Deep-attention CycleGAN generates synthetic CT from cone-beam CT for pancreatic adaptive radiotherapy.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Deep-attention CycleGAN generates synthetic CT from cone-beam CT for pancreatic adaptive radiotherapy.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7762616/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

DeepMILO: a deep learning approach to predict the impact of non-coding sequence variants on 3D chromatin structure

Source check: Source checked. Not independent scientific confirmation.

Finding: This method paper reports a specific cancer-genomics result (recurrent predicted loop disruption and association with BCL2/MYC expression), supported by learned sequence modeling and genomic expression analysis. Keep associations and predictions provisional. Methods/Results: CNN/RNN learn DNA sequence features for CTCF anchor/orientation and loop predictions. Test AUPRC varies by negative set (0.684 hardest); application to WGS of 1,834 patients across 12 cancer types predicts 672 insulator loops disrupted in ≥10% of patients; variants at anchors associated with BCL2/MYC upregulation in malignant lymphoma. Association/predictions computational, with validation on published loop datasets/deletion examples; does not prove causal mechanism.

Technology: CNN/RNN learn DNA sequence features for CTCF anchor/orientation and loop predictions. Test AUPRC varies by negative set (0.684 hardest); application to WGS of 1,834 patients across 12 cancer types predicts 672 insulator loops disrupted in ≥10% of patients; variants at anchors associated with BCL2/MYC upregulation in malignant lymphoma. Association/predictions computational, with validation on published loop datasets/deletion examples; does not prove causal mechanism.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
CNN/RNN learn DNA sequence features for CTCF anchor/orientation and loop predictions. Test AUPRC varies by negative set (0.684 hardest); application to WGS of 1,834 patients across 12 cancer types predicts 672 insulator loops disrupted in ≥10% of patients; variants at anchors associated with BCL2/MYC upregulation in malignant lymphoma. Association/predictions computational, with validation on published loop datasets/deletion examples; does not prove causal mechanism.
Test and result
CNN/RNN learn DNA sequence features for CTCF anchor/orientation and loop predictions. Test AUPRC varies by negative set (0.684 hardest); application to WGS of 1,834 patients across 12 cancer types predicts 672 insulator loops disrupted in ≥10% of patients; variants at anchors associated with BCL2/MYC upregulation in malignant lymphoma. Association/predictions computational, with validation on published loop datasets/deletion examples; does not prove causal mechanism.
Limitation / still unknown
Methods/Results: CNN/RNN learn DNA sequence features for CTCF anchor/orientation and loop predictions. Test AUPRC varies by negative set (0.684 hardest); application to WGS of 1,834 patients across 12 cancer types predicts 672 insulator loops disrupted in ≥10% of patients; variants at anchors associated with BCL2/MYC upregulation in malignant lymphoma. Association/predictions computational, with validation on published loop datasets/deletion examples; does not prove causal mechanism. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13059-020-01987-4

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting Body Mass Index From Structural MRI Brain Images Using a Deep Convolutional Neural Network.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A deep convolutional network predicted body-mass index from structural brain MRI, reporting held-out prediction and image-feature associations; the result is a specific brain-imaging/phenotype association.

Technology: The source documents the learned imaging or signal-analysis operation and its relation to the reported result.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The source documents the learned imaging or signal-analysis operation and its relation to the reported result.
Limitation / still unknown
Prediction and image-feature associations are limited to the reported MRI dataset; causal interpretation is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://doi.org/10.3389/fninf.2020.00010 · https://europepmc.org/article/MED/32265681 · https://europepmc.org/articles/PMC7104804

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Temporal Pattern Detection to Predict Adverse Events in Critical Care: Case Study With Acute Kidney Injury.

Source check: Source checked. Not independent scientific confirmation.

Finding: On MIMIC data from 22,542 patients, random forest with structural trend features achieved 81.3% accuracy, above two specified temporal-feature baselines under 20-fold cross-validation. This is retrospective benchmark performance, not prospective evidence that alerts prevent acute kidney injury.

Technology: Structural temporal-pattern features representing local and global trends were combined with machine-learning classifiers to predict acute kidney injury in ICU data.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Structural temporal-pattern features representing local and global trends were combined with machine-learning classifiers to predict acute kidney injury in ICU data.
Test and result
On MIMIC data from 22,542 patients, random forest with structural trend features achieved 81.3% accuracy, above two specified temporal-feature baselines under 20-fold cross-validation.
Limitation / still unknown
This is retrospective benchmark performance, not prospective evidence that alerts prevent acute kidney injury.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.jmir.org/2020/3/e14272/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Evaluation via Supervised Machine Learning of the Broiler Pectoralis Major and Liver Transcriptome in Association With the Muscle Myopathy Wooden Breast

Source check: Source checked. Not independent scientific confirmation.

Finding: Supervised transcriptome SVM/MLP models distinguish broilers with wooden-breast myopathy and identify differential-expression regulators; cross-validation and held-out splits are reported. The result is a bounded disease-associated transcriptomic signature, not causal validation of regulators. Five-fold CV and separate held-out split reported; cohort and expression comparisons are in original Methods/Results.

Technology: SVM and multilayer perceptron classify breast/liver transcriptomes and rank gene regulators.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
SVM and multilayer perceptron classify breast/liver transcriptomes and rank gene regulators.
Test and result
Supervised transcriptome SVM/MLP models distinguish broilers with wooden-breast myopathy and identify differential-expression regulators; cross-validation and held-out splits are reported. The result is a bounded disease-associated transcriptomic signature, not causal validation of regulators.
Limitation / still unknown
Five-fold CV and separate held-out split reported; cohort and expression comparisons are in original Methods/Results. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2020.00101/full

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Acoustic Indoor Localization Augmentation by Self-Calibration and Machine Learning.

Source check: Source checked. Not independent scientific confirmation.

Finding: Full paper reports experimental ML-based localization in a real cafeteria; evidence is limited to its marked zones and one setup.

Technology: Random-forest and neural classifiers learn TDoA-signature zones; physical experiments use marked table zones and labeled timestamps.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random-forest and neural classifiers learn TDoA-signature zones; physical experiments use marked table zones and labeled timestamps.
Test and result
Real cafeteria experiments show learned zone localization; results section reports 91% and 88% association, while conclusion gives a differing 94% summary.
Limitation / still unknown
Test placements near trained points provide an in-situ evaluation; transfer to other venues is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.mdpi.com/1424-8220/20/4/1177

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Predicting antimicrobial resistance in Pseudomonas aeruginosa with machine learning-enabled molecular diagnostics.

Source check: Source checked. Not independent scientific confirmation.

Finding: Across 414 clinical Pseudomonas aeruginosa isolates, genomic/transcriptomic/phenotypic ML identified marker combinations predicting resistance to four antibiotic classes; transcriptomics improved predictions over genome alone and included previously unassociated markers. External replication and prospective utility were not established in the reviewed material.

Technology: ML integrates gene presence, sequence variation and expression profiles to predict antimicrobial resistance.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
ML integrates gene presence, sequence variation and expression profiles to predict antimicrobial resistance.
Test and result
Across 414 clinical Pseudomonas aeruginosa isolates, genomic/transcriptomic/phenotypic ML identified marker combinations predicting resistance to four antibiotic classes; transcriptomics improved predictions over genome alone and included previously unassociated markers.
Limitation / still unknown
External replication, causal effects and prospective utility were not established in the reviewed material.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.15252/emmm.201910264

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

A deep learning image-based intrinsic molecular subtype classifier of breast tumors reveals tumor heterogeneity that may affect survival.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: A deep image classifier predicts PAM50 breast-tumor subtypes from H&E slides in held-out tumors and examines heterogeneity related to survival; this is a specific retrospective pathology result.

Technology: A deep image classifier predicts PAM50 breast-tumor subtypes from H&E slides and examines subtype heterogeneity.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A deep image classifier predicts PAM50 breast-tumor subtypes from H&E slides and examines subtype heterogeneity.
Limitation / still unknown
Results are limited to the reported models, cohorts, datasets or experiments; broader generalization and clinical or operational benefit were not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://link.springer.com/article/10.1186/s13058-020-1248-3

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Quantifying the Impact of Chronic Ischemic Injury on Clinical Outcomes in Acute Stroke With Machine Learning

Source check: Source checked. Not independent scientific confirmation.

Finding: Automated lesion analysis and machine learning quantified how chronic ischemic injury relates to outcomes after acute stroke in a 1,704-scan clinical cohort.

Technology: Methods apply lesion/tissue segmentation and high-dimensional ML to admission CTs; Results relate chronic injury patterns to clinical outcomes.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Methods apply lesion/tissue segmentation and high-dimensional ML to admission CTs; Results relate chronic injury patterns to clinical outcomes.
Test and result
Methods apply lesion/tissue segmentation and high-dimensional ML to admission CTs; Results relate chronic injury patterns to clinical outcomes.
Limitation / still unknown
Methods apply lesion/tissue segmentation and high-dimensional ML to admission CTs; Results relate chronic injury patterns to clinical outcomes. The source does not establish benefits beyond the reported evaluation.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC6992664/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: other. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Monitoring canid scent marking in space and time using a biologging and machine learning approach.

Source check: Source checked. Not independent scientific confirmation.

Finding / tested result: KNN biologging detected scent-mark postures in video-validated dogs and enabled a bounded field observation in three guardian dogs; substantial attrition and small deployment limit generalization.

Technology: KNN biologging detected scent-mark postures in video-validated dogs and enabled a bounded field observation in three guardian dogs; substantial attrition and small deployment limit generalization.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
KNN biologging detected scent-mark postures in video-validated dogs and enabled a bounded field observation in three guardian dogs; substantial attrition and small deployment limit generalization.
Limitation / still unknown
The field observation involved only three guardian dogs with substantial attrition; generalization to other populations or settings is limited.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC6969016/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open the interactive profile for the timeline, or browse all organizations and model families.