Related Experiment Video
Updated: Mar 19, 2026
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
Prospective evaluation of artificial intelligence-assisted monitoring of the effective withdrawal time on adenoma
Thomas Ka Luen Lui1, Carla Pui-Mei Lam1, Vivien Wai-Man Tsui1
1Department of Medicine, Queen Mary Hospital, University of Hong Kong, Hong Kong, China.
Background/Aims:
Standard withdrawal time (SWT) is a recognized colonoscopy quality metric but may not capture mucosal examination quality. We developed an artificial intelligence (AI)-derived metric, effective withdrawal time (EWT), to quantify clear mucosal visualization. This study prospectively evaluated EWT versus SWT on adenoma detection rate (ADR) and other metrics.
Methods:
In this prospective single-center study, patients undergoing colonoscopy were enrolled. EWT was calculated in real-time using an AI system with endoscopists blinded to results. Multivariable analyses assessed the association of EWT and SWT with binary (e.g., ADR) and count outcomes (e.g., adenoma or polyp per colonoscopy [APC or PPC]), adjusting for patient and procedural characteristics.
Results:
We analyzed 193 colonoscopies by 16 endoscopists. After adjustment, each 1-minute EWT increase was associated with higher odds of adenoma detection (ADR: odds ratio [OR], 1.38; 95% confidence interval [CI], 1.15-1.67) and a > 2-fold increase in adenomas detected per procedure (APC: incidence rate ratio [IRR], 2.14; 95% CI, 1.97-2.32). Both parameters were significantly better with EWT than SWT (OR, 1.38 vs. 1.06; P< 0.01 for ADR; IRR, 2.14 vs. 1.15; P= 0.02 for APC). EWT demonstrated stronger correlation with APC (R2 = 0.97) and PPC (R2 = 0.96) than SWT (R2 = 0.67 and 0.68, respectively; P< 0.01).
Conclusions:
AI-derived EWT is a novel quality indicator reflecting mucosal examination thoroughness, serving as a stronger, more independent predictor of ADR than SWT. (Clinical Trials.gov identifier: NCT06063720).
