Artificial intelligence for polyp detection during colonoscopy: a systematic review and meta-analysis.

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Artificial intelligence (AI) systems are increasingly utilized in colonoscopy to enhance polyp detection and overall procedure quality.
• AI aims to improve the accuracy and efficiency of identifying gastrointestinal lesions during endoscopic examinations.

Purpose of the Study:

• To evaluate the efficacy of AI-based polyp detection systems in improving the detection rates of polyps and colorectal cancer.
• To conduct a systematic review and meta-analysis of prospective trials comparing colonoscopy with and without AI assistance.

Main Methods:

• Systematic searches were conducted in major medical databases (MEDLINE, EMBASE, Cochrane CENTRAL).
• Five randomized trials were included, with independent reviewers assessing study eligibility, risk of bias, and certainty of evidence.
• Colonoscopy outcomes with and without AI were compared using relative risks and mean differences for polyp and adenoma detection.

Main Results:

• AI-assisted colonoscopy demonstrated significantly higher adenoma detection rates (ADRs) and polyp detection rates (PDRs) compared to standard colonoscopy.
• ADRs increased from 19.3% without AI to 29.6% with AI (RR 1.52), and PDRs increased from 30.6% to 45.4% (RR 1.48), both with high certainty.
• AI improved the detection of small adenomas (≤5mm) but showed no significant difference in detecting advanced or larger adenomas; data on cancer detection were unavailable.

Conclusions:

• AI-based polyp detection systems effectively increase the detection of small, non-advanced adenomas and polyps during colonoscopy.
• While AI enhances the detection of numerous polyps, its current impact on identifying advanced adenomas requires further investigation.