Development of a multi-task deep learning system for classification of nine common knee abnormalities on MRI: a large-scale, multicentre, stepwise validation study
Source check: Source checked. Not independent scientific confirmation.
Finding: Results report AUCs on internal and external tests, per-abnormality accuracy, and reader performance/assistance outcomes.
Technology: A meniscus segmentation U-Net and nine binary classifiers with attention-guided localization process three MRI planes and classify each named abnormality.
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- A meniscus segmentation U-Net and nine binary classifiers with attention-guided localization process three MRI planes and classify each named abnormality.
- Test and result
- Results report AUCs on internal and external tests, per-abnormality accuracy, and reader performance/assistance outcomes. Reference labels were independently made by musculoskeletal radiologists with consensus; patient-level data splits and two geographically separate external cohorts; stepwise reader validation. Performance varies by finding/age, and retrospective diagnostic performance does not prove patient benefit.
- Limitation / still unknown
- normalities across 13,419 patients/14,962 examinations.
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/PMC12508577/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..
No follow-up recorded; this is not a claim that no follow-up 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.