AI Science Tracker
EN

Ascites-image models stratify recorded platinum resistance in ovarian cancer

Fudan University Shanghai Cancer Center; Northern Jiangsu People’s Hospital · 28 Jun 2026 · medicine

Discovery

Ascites-image models stratify recorded platinum resistance in ovarian cancer

Source check: Source checked. Not independent scientific confirmation.

Finding: OVCAP analyzed pretreatment ascites cytology from 438 women with advanced epithelial ovarian cancer. Internal and two additional-site validation cohorts yielded resistance-classification AUCs of 0.894, 0.863 and 0.828. This retrospective treatment-response study does not establish that model-guided treatment improves outcomes.

Technology: Prov-GigaPath features, TransMIL and multi-scale late fusion

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

AI contribution
Pathology foundation-model embeddings and attention-based multiple-instance learning estimated response risk from image patches; trained multi-scale fusion combined the prediction streams.
Test and result
The 2019–2023 cohort comprised 367 Xuhui-campus cases split into 254 training and 113 internal-validation patients, plus 51 Pudong-campus and 20 Northern-Jiangsu validation patients. Patient-level three-fold training cross-validation selected among 15 encoders and hyperparameters. The retained multi-scale predictor was applied unchanged to the held-out cohorts. Resistance meant progression during therapy or recurrence/progression within six months after completing platinum treatment; the reported internal/external AUCs were .894/.863/.828.
Limitation / still unknown
The three sites represent two institutions; one additional-site cohort shares the same cancer center as development. The smallest cohort contains only 20 people. Retrospective selection of patients with evaluable outcomes, histologic concentration in high-grade serous cancer and preparation/scanner differences constrain transfer. AUCs are not treatment-benefit estimates. Attention-selected morphology and integrated single-cell associations are hypothesis-generating, not established mechanisms of resistance. Clinical vignettes and comparison with later-available kinetic biomarkers do not demonstrate 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/PMC13330694/

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: Original model/clinical cohort, outcome definition, patient-level training selection and held-out Results; exploratory biological interpretation and clinical benefit kept bounded..

No follow-up recorded; this is not 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.

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