Development and validation of artificial intelligence-based model for bladder cancer immunophenotyping using whole slide images
AI analyzing routine pathology slides can identify bladder cancer types most likely to respond to immunotherapy, faster and more consistently than pathologists alone.
An AI system using spatial cell-graph networks on routine H&E slides classified muscle-invasive bladder cancer immunophenotypes with AUC 0.922-0.956, outperforming both junior and senior pathologists while cutting review time. Predicted Inflamed tumors showed enriched CD8+ infiltration and stronger correlation with ICI response, enabling reproducible, scalable immunotherapy selection from standard pathology slides.
What the study was
- Study design
- Multicenter retrospective diagnostic study with immunotherapy efficacy validation cohort
- Population
- Muscle-invasive bladder cancer (MIBC) patients undergoing cystectomy (2014-2024) at 2 Chinese hospitals + TCGA cohort; ICI-treated cohort for efficacy validation
- Category
- Diagnostics
- Maturity
- Validated
- Journal
- NPJ Precis Oncol
Why it surfaced
Multicenter-validated AI pathology system enabling routine-slide immunophenotyping of MIBC for immunotherapy selection. Human-AI comparison shows clear outperformance of pathologists with reduced review time — strong translational pathway.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.