A stromal-immune computational pathology signature for prognosis and immune checkpoint inhibitor response in localized, locally advanced and metastatic urothelial carcinoma: A multicenter retrospective study.
A deep-learning biomarker combining immune and stromal features better predicts immunotherapy response in bladder cancer than existing tests, potentially guiding treatment selection.
This multicenter retrospective study developed and externally validated a combined TLS-TSR risk score in 884 UC patients across five cohorts (FAHZU, Emory, TCGA, QDPH, TRRC) using deep learning-based nuclei classification and automated stromal segmentation. The integrated score outperformed individual biomarkers (C-index 0.65–0.69), remained an independent PFS predictor in multivariable analysis, and was strongly associated with ICI response—establishing it as a clinically actionable computational biomarker for UC management.
What the study was
- Study design
- retrospective_cohort_multicenter
- Population
- Urothelial carcinoma patients across five independent cohorts
- Sample size
- 884
- Category
- Diagnostics
- Maturity
- Validated
- Journal
- Eur J Cancer
Why it surfaced
Multicenter external validation across 5 cohorts in 884 patients is exceptional for computational pathology; ICI response prediction AUC 0.745 is clinically meaningful; directly bridges AI diagnostics and precision immunotherapy selection.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.