Patch-to-slide fusion deep learning model for histological diagnosis of early pregnancy loss including hydatidiform mole
An AI tool can help pathologists more accurately diagnose early pregnancy loss types, potentially reducing the need for costly molecular testing.
A multicenter deep learning model (n=1287) achieved AUROC 0.930 on independent testing for histological classification of early pregnancy loss subtypes including hydatidiform mole, and significantly improved pathologist diagnostic accuracy when used as an assist tool. This addresses a real clinical need where current gold standards combining pathology and molecular profiling are too costly for routine implementation.
What the study was
- Study design
- Multicenter development and independent validation cohort (n=1287 patients, 1380 H&E slides + 1057 p57 + 646 Ki-67 whole-slide images)
- Population
- Patients with early pregnancy loss and hydatidiform mole at multiple hospital centers in China
- Sample size
- 1287
- Category
- Diagnostics
- Maturity
- Validated
- Journal
- NPJ Digital Medicine
Why it surfaced
NPJ Digital Medicine publication, large multicenter cohort, independent validation with strong AUROC, AI-assist demonstrated to improve pathologist performance in a resource-limited setting; represents a near-term implementable diagnostic tool.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.