Impact of computer-aided polyp detection in screening colonoscopy: a multi-center randomized study

Aims This multicenter clinical trial explored the role of a novel computer-aided detection (CADe) system in screening colonoscopies performed at tertiary referral centers. The primary objective was to compare the adenoma detection rate (ADR) achieved with CADe to that of standard colonoscopy without CADe, both using ultra-HD monitors. Furthermore, we investigated whether the effects of CADe differed between expert and non-expert endoscopists, and according to the operator's fatigue variables.

Methods Between August 2023 and September 2024, consecutive patients were randomized in a 1:1 fashion to receive screening colonoscopy (i.e., first colonoscopy during the last 5 years in over 40-year-old patients) with either ultra-HD-Iscan-CADe+or ultra-HD-Iscan-CADe- across five Italian centers. For CADe+procedures we adopted the Pentax Medical DISCOVERY, a novel artificial intelligence device providing real-time visual alerts to highlight colorectal lesions suspicious for polyps. Participating physicians were divided into low-volume and high-volume categories (i.e.,<o≥200 colonoscopies / year) to determine whether CADe could help close gaps in detection performance. Parameters measuring operators’ fatigue were collected (i.e., working hours, number of exams performed).

Results We enrolled a total of 421 patients, with 409 patients included in the final data analysis set. The patients’ demographics, bowel cleansing, and withdrawal time were similar in all the centers and between the two treatment arms. The main indications were familial history of colorectal cancer, abdominal pain, and positive fecal occult blood test. Overall, we did not observe clinically relevant changes (p>0.05) in terms of ADR (CADe+N=61, 29.9%, vs CADe- N=69, 33.7%), advanced ADR (CADe+N=12, 5.9%, vs CADe- N=11, 5.4%), polyp detection rate (CADe+N=84, 41.2%, vs CADe- N=93, 45.4%), serrated polyp detection rate (CADe+N=12, 5.9% vs CADe- N=16, 7.8%). The hyperplastic polyp detection rate (CADe+N=22, 26.8%, vs CADe- N=23, 25.6%), based on a per-lesion analysis, was also comparable between the two treatment arms. This figure was also confirmed when stratifying according to either the operator’s expertise or the daily working-related fatigue. However, a trend towards a leveling effect was observed between high- and low-volume operators for ADR, with low-volume operators showing a slight increase when supported by CADe (ADR+3.5%).

Conclusions In the everyday clinical setting of tertiary referral centers, we did not identify a clinically meaningful advantage in using CADe for colorectal polyp detection compared to standard colonoscopies with ultra-HD imaging. Low-volume operators might benefit more from CADe support. These findings should be confirmed in large clinical trials performed in primary and secondary care settings.

Publication History

Article published online:
27 March 2025