Effectiveness of CADe in Colonoscopy for Improved Adenoma and Polyp Detection Under Diverse Clinical Conditions

Colonoscopy is the gold standard for colorectal cancer prevention, but its effectiveness is limited by missed lesions and operator-dependent variability. Computer-aided detection (CADe) has been introduced to mitigate these limitations, yet its value for endoscopists with different levels of expertise remains unclear. This study evaluated the real-world effectiveness of CADe, focusing on its differential impact on experts and trainees under diverse clinical conditions.

We conducted a retrospective, comparative study of consecutive colonoscopies performed at a tertiary center. To reduce potential selection bias, we applied 1:1 propensity score matching (PSM), yielding 724 matched patients (362 CADe vs. 362 non-CADe). Primary outcomes were adenoma detection rate (ADR) and polyp detection rate (PDR), stratified by operator expertise, bowel preparation quality, and lesion characteristics.

The CADe group demonstrated significantly higher overall ADR (40.6% vs. 30.4%, p=0.004) and PDR (60.2% vs. 50.6%, p=0.011). In subgroup analyses, experts achieved a significant increase in ADR (44.7% vs. 34.0%, p=0.022) even within a lower-risk patient cohort, supporting a refinement effect. Trainees showed a numerical increase in ADR (32.2% vs. 25.9%, p=0.251) in a higher-risk cohort, with performance approaching the baseline detection rates of experts. CADe was particularly effective for detecting small polyps (≤5 mm; p=0.002) and in patients with optimal bowel preparation (BBPS ≥7; p=0.008). Significant improvements were observed in the transverse colon and descending colon.

CADe significantly enhances adenoma and polyp detection in real-world practice. Its utility varies by operator experience, serving as a refinement tool for experts to detect subtle lesions and a safety net for trainees to bridge the experience gap. Adequate bowel preparation is essential for maximizing these benefits.