AI Science Tracker
EN

Serial CT and clinical features support learned early classification of malignant stroke edema

Washington University in St. Louis; Jagiellonian University; Helsinki University Hospital · 20 Aug 2021 · biology

Discovery

Serial CT and clinical features support learned early classification of malignant stroke edema

Source check: Source checked. Not independent scientific confirmation.

Finding: An LSTM combining serial clinical assessments and automated CT measurements identified malignant edema cases in an internal analysis of 598 stroke patients. Twenty patients required decompressive surgery or died with substantial midline shift. The source reports stronger internal precision than regression and a fully connected network, but rare events, imaging-based selection and absent external validation limit clinical interpretation.

Technology: Long short-term memory recurrent neural network combining serial clinical and CT-derived measurements

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

AI contribution
The LSTM learned nonlinear dependencies across baseline and follow-up clinical/imaging measurements, enabling the represented patient-endpoint discrimination beyond the separately assessed regression and feed-forward models.
Test and result
A retrospective imaging substudy of a prospectively enrolled three-center stroke cohort retained 598 patients with baseline and follow-up CT, including twenty malignant-edema endpoints: twelve decompressive hemicraniectomies and eight deaths with midline shift of at least five millimeters. The main Methods describe nested stratified tenfold cross-validation with patients held out from training/tuning. The source reports LSTM recall 100%, precision 87% and AUPRC 0.97, compared with regression precision 32% and AUPRC 0.74 using the automated feature set.
Limitation / still unknown
Only twenty positive events and selection for available repeat imaging leave substantial uncertainty about transportability and overfitting. External cohorts, prospective locked-model evaluation and clinical decision benefit were not tested. The outcome combines a clinical intervention decision with edema-related death; CT and clinical assessment timing varied. Supplemental implementation/imputation details were not independently read, so fold-specific preprocessing isolation is not authenticated here. Model feature attribution is predictive association, not causal edema 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/PMC8858326/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation 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 clinical eligibility/outcome/timing, automated feature and LSTM/nested patient-cross-validation Methods, endpoint/model Results and limitations. Supplemental implementation/imputation Methods were not independently read; no claim that every preprocessing operation was fold-isolated. Exact original article front DOI/title and fresh PubMed primary relationship metadata 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.

Open in the full catalog for filters, charts and related discoveries. Browse biology discoveries.

Related catalog entries

Links use documented relationships where available, plus shared catalog context. A shared topic or model family does not establish a scientific dependency.