An image model labels hybrid nerve-tumor slides in a small held-out set
Source check: Source checked. Not independent scientific confirmation.
Finding: A retrospective study of 115 nerve-sheath tumors compared image classifiers for hybrid neurofibroma/schwannoma and related diagnoses. In 17 held-out hybrid-only slides from another center/scanner, the best reported model reached mean accuracy 0.96. This selected positive-class test does not establish specificity against other tumors or clinical utility.
Technology: ResNet50 features and clustering-constrained attention multiple instance learning
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Learned image features and multiple-instance classifiers predicted diagnostic slide labels; expert histology supplied reference diagnoses.
- Test and result
- The 115 tumors supplied 375 histological sections across three German centers. Authors report patient-separated cross-validation; a 17-slide hybrid-only holdout assessed scanner/center transfer. The reported best mCLAM/ResNet50 configuration with stain normalization achieved accuracy 0.96 ± 0.05 and F1 0.98 ± 0.03.
- Limitation / still unknown
- The holdout contains only hybrid-tumor slides, so it does not measure external three-class specificity. Its distinct patient count is not clear from the main text. Different configurations were compared on the test set, and best test performance is not a locked-model estimate. Expert-curated referral cases, artifacts and center/scanner confounding limit generalization; prospective clinical 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://pmc.ncbi.nlm.nih.gov/articles/PMC13524541/
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: 21 Nov 2026. Review scope: Primary full-text bibliographic dates, computational classification Methods, held-out Results and explicit cohort/reference limitations..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 1/5 · 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.