Accurately measuring the actual observation time (AOT), excluding biopsy or polypectomy time, remains a challenge during colonoscopy. Bowel preparation is another critical factor affecting the quality of colonoscopy. We aimed to evaluate the impact of comprehensive artificial intelligence (AI)-based monitoring of AOT and bowel preparation on adenoma detection rate (ADR). 

We developed an AI model that comprehensively assesses both withdrawal time and bowel preparation quality simultaneously. We validated the performance of the AI model by analyzing a total of 1,489 consecutive colonoscopy videos. The AI model assessed both crude withdrawal time (CWT), defined as the time from cecal intubation to anal withdrawal, and AOT. Each was classified into four groups: < 6 min, 6–9 min, 9–12 min, and ≥ 12 min. Bowel preparation was classified as adequate or inadequate. Colonoscopy quality indicators such as ADR and adenomas per colonoscopy (APC) were compared across combined categories of AOT and preparation status. 

The mean age of the 1,489 patients was 60.9 ± 12.9 years, and 71.1% were male. As both CWT and AOT increased, the ADR increased significantly. ADR increased from 6.6% in patients with CWT < 6 min to 71.2% in those with CWT ≥ 12 min (p < 0.001), and from 11.3% in patients with AOT < 6 min to 70.2% in those with AOT ≥ 12 min (p < 0.001). ADR was significantly higher in the 6–9 min AOT group than in the 6–9 min CWT group (23.5% vs. 16.5%, p = 0.005). Advanced ADR was also significantly higher in the 6–9 min AOT group than in the 6–9 min CWT group. In addition, APC was significantly greater in the 6–9 min AOT group than in the 6–9 min CWT group (0.33 [95% confidence interval (CI) 0.27 – 0.39] vs. 0.19 [95% CI 0.15 – 0.22], p = 0.002). The ADR was significantly higher in the AI-assessed adequate preparation group than in the inadequate preparation group (41.1% vs. 31.0%, p = 0.001). APC was also significantly higher in the AI-assessed adequate preparation group. In the integrated analysis, the odds ratio (95% CI) for ADR were 2.76 (1.67–4.79) with adequate preparation and AOT < 9 min, 8.53 (4.91–15.46) with inadequate preparation and AOT ≥ 9 min, and 12.07 (7.4–20.76) with both adequate preparation and AOT ≥ 9 min, compared with inadequate preparation and AOT < 9 min. Similar trends were observed for other outcomes, including multiple advanced ADR and APC.

An integrated metric combining AI-assessed AOT and bowel preparation status offers a more detailed assessment of adenoma detection outcomes, including ADR and APC. Comprehensive AI-based monitoring of quality parameters may enhance colonoscopy performance without increasing the burden on endoscopists.