Real-Time Artificial Intelligence Versus Standard Colonoscopy in the Early Detection of Colorectal Cancer: A Systematic Review and Meta-Analysis.

Abstract

<b>Background:</b> Colonoscopy remains the gold standard for colorectal cancer screening. Deep learning systems with real-time computer-aided polyp detection (CADe) demonstrate high accuracy in controlled research settings and preliminary randomized controlled trials (RCTs) report favorable outcomes in clinical settings. This study aims to evaluate the efficacy of AI-assisted colonoscopy compared to standard colonoscopy focusing on Polyp Detection Rate (PDR) and Adenoma Detection Rate (ADR), and to explore their implications for clinical practice. <b>Methods:</b> A systematic search was conducted using multiple indexing databases for RCTs comparing AI-assisted to standard colonoscopy. Random-effect models were utilized to calculate pooled odds ratios (ORs) with 95% confidence intervals. The risk of bias was assessed using the Cochrane Risk of Bias Tool, and heterogeneity was quantified using I statistics. <b>Results:</b> From 22,762 studies, 12 RCTs (n = 11,267) met the inclusion criteria. AI-assisted colonoscopy significantly improved PDR (OR 1.31, 95% CI 1.08-1.59, <i>p</i> = 0.005), despite heterogeneity among studies (I² = 79%). While ADR showed improvement with AI-assisted colonoscopy (OR 1.24, 95% CI, 0.98-1.58, <i>p</i> = 0.08), the result was not statistically significant and had high heterogeneity (I² = 81%). <b>Conclusions:</b> AI-assisted colonoscopy significantly enhances PDR, highlighting its potential role in colorectal cancer screening programs. However, while an improvement in the ADR was observed, the results were not statistically significant and showed considerable variability. These findings highlight the promise of AI in improving diagnostic accuracy but also point to the need for further research to better understand its impact on meaningful clinical outcomes.