AI Framework Predicts Urothelial Carcinoma Outcomes via H&E Images

anantm · x · 2026-08-23

Emory University team developed a novel AI framework using digitized H&E images for urothelial carcinoma analysis. The automated TLS-TSR pipeline was validated across 884 patients from 5 independent cohorts, closely matching expert pathologist assessments. The combined TLS-TSR score independently predicted progression-free survival (PFS) in every cohort after adjusting for stage and treatment. In the mUC cohort, TLS density alone tracked with immune checkpoint inhibitor (ICI) response (AUC 0.745). This is the first unified AI framework to quantify stromal architecture and immune organization together from H&E, providing a fully automated, reproducible readout for pathology labs.

Original post →

More from Research

Research channel →