Bowel preparation and adenoma detection in AI-assisted colonoscopy.

BACKGROUND AND AIM: Artificial intelligence (AI)-assisted colonoscopy improves adenoma detection rate (ADR), but it is currently unknown whether it performs better with excellent vs good bowel cleansing. This study aims to evaluate the association between bowel cleansing with ADR and adenomas per colonoscopy (APC) in average-risk patients undergoing AI-assisted colonoscopy. METHODS: Prospective single-centre study. Consecutive average-risk patients undergoing AI-assisted colonoscopy were enrolled from December 2025 to March 2026. Demographic and endoscopic variables were collected. Bowel preparation quality was assessed using the Boston Bowel Preparation Scale (BBPS); adequate cleansing was defined as BBPS ≥ 6 and classified as good (6-7) or excellent (≥ 8). Univariable and multivariable regression analyses identified variables associated with ADR and APC. RESULTS: A total of 425 patients were included. The overall ADR and APC were 0.35 and 0.6, respectively. Bowel preparation quality was not associated with ADR or APC. On multivariable analysis, ADR was associated with: BMI (OR 1.05, p = 0.041), female sex (OR 0.39, p < 0.001) and afternoon examinations (OR 0.39, p = 0.013). Female sex (IRR 0.53, p < 0.001) and higher BMI (IRR 1.05, p = 0.031) were independently associated with APC. CONCLUSIONS: Excellent bowel preparation quality was not associated with higher ADR or APC compared with good bowel preparation in this cohort. Female sex, BMI, and timing of colonoscopy were independently associated with ADR, despite the use of computer-aided detection systems.