Artificial intelligence–assisted colonoscopy may mitigate the negative impact of inadequate bowel preparation on adenoma detection, according to a single-center retrospective study evaluating real-world performance of computer-aided detection (CADe).

Investigators analyzed 3163 CADe-assisted colonoscopies and compared outcomes with 2516 non-AI procedures performed prior to implementation. Bowel preparation quality was assessed using the Boston Bowel Preparation Scale (BBPS), with scores of 6 or less categorized as inadequate.

Among AI-assisted procedures, polyp detection rate (PDR) and adenoma detection rate (ADR) were similar regardless of bowel preparation quality. PDR was 71.6% in inadequately prepared cases versus 74.3% in adequately prepared cases, while ADR was 56.9% versus 54.2%, respectively. In contrast, non-AI colonoscopies showed significantly lower detection rates with inadequate preparation, particularly in the proximal colon.

The authors reported that “PDR and ADR were statistically similar regardless of bowel preparation quality” with CADe, suggesting that “AI can compensate for inadequate bowel preparation.” They proposed that continuous AI scanning may identify lesions that could be missed when endoscopists are focused on clearing debris.

However, not all outcomes were unaffected. Sessile serrated lesion detection remained dependent on preparation quality, with higher detection rates observed in adequately prepared colonoscopies even with AI support. This finding indicates that optimal bowel preparation remains important for detecting subtle lesions.

These findings support the integration of AI tools into endoscopic practice while reinforcing that bowel preparation quality remains an important factor in overall colonoscopy performance.

strong="">
Li, TY, Mansour NM. Artificial intelligence assisted colonoscopy can compensate for inadequate bowel preparation. Presented at: Digestive Disease Week; May 2–5, 2026; Chicago, Illinois.