Logo image
Back
IP20-12 ASSOCIATION OF A COMPUTATIONAL HISTOLOGY AI (CHAI) PROGNOSTIC BIOMARKER WITH HIGH GRADE RECURRENCE INDEPENDENT OF TRADITIONAL RISK FACTORS IN LOW GRADE NON-MUSCLE INVASIVE BLADDER CANCER
Abstract   Peer reviewed

IP20-12 ASSOCIATION OF A COMPUTATIONAL HISTOLOGY AI (CHAI) PROGNOSTIC BIOMARKER WITH HIGH GRADE RECURRENCE INDEPENDENT OF TRADITIONAL RISK FACTORS IN LOW GRADE NON-MUSCLE INVASIVE BLADDER CANCER

Roger Li, Yair Lotan, Vikram M. Narayan, Vrishab Krishna, Jay D. Raman, Michael A. O'Donnell, Patrick J. Hensley, John A. Taylor, Viswesh Krishna, Asit Tarsode, …
The Journal of urology, Vol.215(5S), p.e423
05/2026
DOI: 10.1097/01.JU.0001191388.74345.c9.12

View Online

Abstract

INTRODUCTION AND OBJECTIVES: Accurate risk classification is essential for managing non–muscle-invasive bladder cancer (NMIBC), yet existing clinicopathologic models show limited prognostic accuracy and occasional discordance. Previous biomarker models using H&E pathology analyzed with the Computational Histology Artificial Intelligence (CHAI) platform risk stratify high grade (HG) NMIBC more accurately than traditional approaches. Given low grade (LG) disease treatment and surveillance options vary in intensity within and across risk groups, we sought to determine if a new CHAI biomarker could also stratify the risk of HG recurrence in patients with LG NMIBC. METHODS: For 783 patients with available pathologist-selected H&E slide(s) representative of NMIBC diagnosis, whole slide images were analyzed with the CHAI platform using deep learning to extract quantitative histologic features for algorithmic assessment. A novel biomarker signature using only histologic feature data was created in the development cohort (n=275) to prognosticate HG RFS in NMIBC. CHAI biomarker stratification of HG RFS was analyzed in 508 LG NMIBC validation cases with Kaplan Meier analysis and log rank test. RESULTS: Of 508 LG NMIBC validation cases, AUA risk distribution included 107 (21%) low risk, 377 (74%) intermediate risk (IR) and 24 (5%) unknown. Median follow up was 40.3 months. Cases identified as CHAI high risk had worse HG RFS vs the CHAI low risk group (HR 3.06, p<0.05). Recurrence rates at 12, 24 and 36 months were 20%, 24% and 34% in the CHAI high risk group and 5.8%, 9.4%, 14% in the CHAI low risk group. In an IBCG-defined IR NMIBC subset (LG tumors with ≥1 IBCG risk factor, n=425) the CHAI biomarker risk groups still stratified HG RFS (HR 2.65, p<0.05) and were independent of prognostic IBCG risk factor groups (HR 2.86, p<0.05). In this subset, recurrence rates at 12, 24 and 36 months were 21%, 24% and 34% in the CHAI high risk group and 6.9%, 11% and 16% in the CHAI low risk group. CONCLUSIONS: This novel CHAI biomarker identifies patients with LG NMIBC who have significantly increased risk of HG recurrence utilizing only routine H&E stained tumor specimens, providing independent prognostic value beyond guideline risk factors. Such patients may be candidates for escalated treatment, including intravesical therapy and/or surveillance intensification.

Details

Metrics

1 Record Views
Logo image