Abstract

Background: Advent of Artificial Intelligence (AI) with the development of computer-aided polyp detection (CADe) could be an important support against colorectal cancer (CRC). Our primary aim was to estimate polyp detection rate (PDR) and adenoma detection rate (ADR) improvement in CADe colonoscopy (CAD-EYE, Fujifilm, Tokyo, Japan) when compared to traditional colonoscopy, and to assess how these results could impact on our clinical activity.

Materials and Methods: We retrospectively collected data from all consecutive 40-to80-years old subjects undergoing colonoscopy for primary CRC screening, post-polypectomy surveillance or the presence of any gastrointestinal symptoms from November 2019 to February 2020 (WL-HD colonoscopy, control arm, 450 patients) and from October 2020 to January 2021 (CAD-EYE colonoscopy, study arm, 250 patients).

Results: Higher PDR (159/250 [63.60%] vs 163/450 [36.22%]; p < 0,001) and ADR (115/250 [46%]

vs 138/450 [30,67%]; p < 0,001) was found in the CADe group in comparison to the control group. CADe colonoscopy detected more lesions, especially diminutive lesions (RRR 5.07; 95% CI 3.44 - 7.46; p < < 0.001001) and 6 - 9 mm lesions (RRR 2,75; 95% CI 1.64 - 4.60; p < 0.001).

Regarding lesions’ histology, CADe is associated to a higher detection of non-advanced adenomas (RRR 2.97; 95% CI 2.02 - 4.38; p < 0.001) and serrated lesions (RRR 14.02; 95% CI 5.37 - 36.62; p < 0.001), while no significant improvement was found for advanced adenomas (RRR 1.35; 0.34 - 5.34; p = 0.667) and adenocarcinomas (RRR 1.3; 95% CI 0.52-3.23; 0.574).

Conclusion: Our study showed a 75,6% and 49,98% relative increase in PDR and ADR in the CADe group (absolute increase of 27,38% and 15,33% respectively). Real-time AI-aided colonoscopy applied in the routine endoscopic activity could significantly improve diagnostic ability especially for diminutive lesions detection. Further studies are needed to evaluate a possible colon cancer incidence reduction following AI-colonoscopy increased ADR.

Keywords: Artificial Intelligence (AI); Computer-Aided Polyp Detection (CADe); Colorectal Cancer (CRC); Polyp Detection Rate (PDR); Adenoma Detection Rate (ADR)