Related Experiment Video
Updated: Oct 25, 2026
Structured Approach to Colonoscopy Technique Optimization: A Single-Center Experience with Novice Endoscopists
Published on: July 11, 2025
Effectiveness of artificial intelligence-assisted colonoscopy for colorectal lesion detection: a systematic review
Xia Liu1, Jian Qin1, Yuxia Zhao1
1Department of Basic Medicine, Sichuan Vocational College of Health and Rehabilitation, Zigong, Sichuan, 643000, China.
Background:
Colonoscopy is the gold standard for colorectal cancer detection and prevention, but its performance depends on operator skill and lesion features. Artificial intelligence (AI)-assisted colonoscopy has been introduced to improve adenoma detection.
Methods:
A comprehensive search was conducted in PubMed, Embase, the Cochrane Library, Web of Science, ClinicalTrials.gov, and major Chinese databases to identify randomized controlled trials (RCTs) comparing AI-assisted and conventional colonoscopy. Meta-analyses were performed using Review Manager 5.4 and Stata 14.0. The quality of evidence was graded using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system, while trial sequential analysis (TSA) was applied to assess the robustness and information sufficiency of the results.
Results:
A total of 46 RCTs encompassing 37,012 participants were included. Compared with conventional colonoscopy, AI-assisted colonoscopy was associated with a higher ADR (RR = 1.22, 95% CI = 1.17-1.27, P < 0.00001), polyp detection rate (PDR: RR = 1.22, 95% CI = 1.16-1.29, P < 0.00001), polyp detection per colonoscopy (PPC: MD = 0.28, 95% CI = 0.21-0.36), and adenoma detection per colonoscopy (APC: MD = 0.22, 95% CI = 0.15-0.29). The detection rate of sessile serrated lesions (SSL) was also higher in the AI-assisted group (RR = 1.20, 95% CI: 1.05-1.36, P = 0.007), whereas no significant difference was observed in advanced adenoma detection rate (AADR).
Conclusions:
AI-assisted colonoscopy was associated with improvements in several detection-related quality indicators, including ADR, PDR, PPC, APC, and SSL detection rate. However, given the very low GRADE certainty for ADR and PDR and the lack of long-term outcome data, these findings should be interpreted cautiously.
Clinical Trial Number:
Not applicable.
