Source check: Source checked. Not independent scientific confirmation.
Finding: A deep-learning model combined electronic portal imaging device data with daily machine-state features to predict gamma pass rates and assess detection under simulated machine and multileaf-collimator deviations. This is an engineering quality-assurance result, not evidence of improved patient outcomes.
Technology: Deep-learning prediction using EPID images and daily machine-state features
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Deep-learning prediction using EPID images and daily machine-state features
- Test and result
- Plan-paired tests and simulated deviation scenarios evaluated gamma-pass-rate prediction and deviation detection.
- Limitation / still unknown
- The tests do not establish patient-outcome benefit or performance in prospective clinical deployment.
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.2026.1911466
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.
Source check: Source checked. Not independent scientific confirmation.
Finding: An ensemble-based angiography system detected 1,705 of 1,828 manually defined coronary lesions in 420 patients, compared with 69% detection by visual estimation. Models trained on separate 2016 data were tested retrospectively on selected 2022 images from three hospitals. The result establishes reference-based lesion detection, not improved treatment outcomes.
Technology: MPXA-2000; DeepLabV3+, U-Net++ and U²-Net segmentation ensemble
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Learned vessel segmenters generated lumen boundaries; an ensemble and downstream lesion analysis located and quantified lesions for comparison with manual QCA and cardiologist visual estimates.
- Test and result
- Retrospective three-center evaluation of 1,076 selected angiography images from 420 patients. Nine view-specific segmentation models combined DeepLabV3+/EfficientNet-B4, U-Net++/EfficientNet-B4 and U²-Net; training images from 2016 were separate from 2022 evaluation data. Against blinded manual QCA, the system detected 1,705/1,828 lesions (93%) versus 69% for visual estimation; on multiple-lesion images it identified all reference lesions in 86% versus 33%.
- Limitation / still unknown
- Manual QCA is the anatomical reference, not physiological significance or clinical outcomes. Selected frames excluded occlusions and poor-quality images; only reference lesions meeting size/stenosis criteria were counted. Lesion observations repeat within patients, and reported significance does not establish cluster-adjusted uncertainty. No prospective procedural or patient benefit was tested; training details in the supplementary material were not independently inspected.
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/PMC11880186/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 21 Nov 2026. Review scope: Read actual primary cohort, separate model-training data, named segmentation ensemble, blinded manual/AI/visual-reference Methods, detection definition, statistical Methods, Results and limitations. Supplementary training tables were not independently inspected. Exact primary identity/date and fresh primary PubMed correction/retraction relationships checked..
No follow-up recorded; this is not a claim that no follow-up 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.