Withdrawal time (WT) is a well-established quality indicator in colonoscopy, directly associated with adenoma detection rate (ADR). Prolonged WT may enhance ADR but can increase operator fatigue and procedural inefficiency. Computer-aided detection (CADe) has been developed to assist endoscopists in real-time polyp recognition. This study aimed to evaluate whether CADe allows effective shortening of WT while maintaining ADR in colonoscopy.

This was a prospective observational study. Each native 6-minute withdrawal was proportionally rescaled to a 5-minute reference while preserving frame cadence and CADe alert dynamics. The primary outcome was the paired difference in polyp detection rate (PDR) between the original inspection and the 5-minute reference. Secondary outcomes were optical adenoma detection rate (oADR) and polyps per colonoscopy (PPC), with paired comparisons via McNemar’s test and Newcombe confidence intervals. Subgroup analyses were conducted by age and sex.

Among 269 colonoscopies, oADR was preserved between conditions (34.9% vs 35.3%; Δ = 0.4%; 95% CI −2.1% to 3.7%; p = 0.73), whereas PDR increased (42.0% vs 47.6%; Δ = 5.6%; 95% CI 1.1%–9.9%; p = 0.02). Signals appeared stronger in patients <60 years and in males (exploratory). No carcinoma-suspected events were observed. Given CADe’s propensity to increase detection of diminutive/nonadvanced lesions with minimal time increases, findings should be viewed as feasibility signals.

Time-rescaled modeling suggests that CADe can preserve adenoma-suspected detection while improving overall polyp recognition under a simulated 5-minute withdrawal, supporting the biological plausibility of efficiency-preserving workflows. Real-time, histology-verified, multicenter trials should confirm these findings using AI-specific protocol and trial reporting standards and diagnostic reporting principles.