AI-assisted histomorphological stratification of endometrial cancer: real-world validation of foundation models for molecular subtyping
AI-assisted analysis of cancer tissue slides shows promise for classifying endometrial cancers accurately, though real-world performance varies by laboratory conditions.
This real-world validation study benchmarks foundation model AI (UNI and CTransPath) for H&E-based molecular subtyping of endometrial cancer across heterogeneous diagnostic conditions, demonstrating robust performance for p53abn classification (AUROC 0.844) and quantifying scanner hardware and stain normalization effects. While subtype-specific performance varies, the study establishes important real-world baseline performance for AI pathology workflows in gynecological oncology.
What the study was
- Study design
- Real-world retrospective validation; heterogeneous image quality routine diagnostic cases; foundation model benchmarking (CTransPath vs UNI); stain normalization impact quantified
- Population
- Endometrial cancer patients with confirmed molecular subtype classification (POLEmut n=16, MMRd_MSI n=79, NSMP n=176, p53abn n=18)
- Sample size
- 289
- Category
- Diagnostics
- Maturity
- Exploratory
- Journal
- npj Precision Oncology
Why it surfaced
Real-world validation of AI foundation models for EC molecular subtyping addresses the gap between controlled validation and clinical deployment. Small subtype sample sizes (POLEmut n=16, p53abn n=18) limit precise performance estimates. Score 7 reflects importance of real-world validation framing and high-quality journal (npj Precision Oncology).
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.