eP160 - Computer-aided Detection of Polyps with Artificial Intelligence During Colonoscopy in the Hands of an Experienced Colorectal Surgeon

Purpose/Background: Computer-aided detection (CADe) using artificial intelligence-based tools to assist in polyp detection during colonoscopy is available. Initial studies of CADe demonstrated encouraging outcomes with higher Polyp Detection Rates (PDR) and Adenoma Detection Rates (ADR). However, more recent research has produced mixed findings, especially in the hands of an experienced endoscopist. In this study, we aim to investigate the hypothesis that CADe, when used by an experienced colorectal surgeon, will not significantly improve either polyp or adenoma detection rates during screening or surveillance colonoscopy.Methods/Interventions: This case-control study retrospectively reviewed data collected in a prospective database. A single colorectal surgeon with a nine-year history of performing over 400 colonoscopies annually participated in the study. Beginning on July 1, 2022, CADe (GI Genius Intelligent endoscopy module by Medtronics Inc.) was incorporated into the practice of all colonoscopies. The study only included colonoscopies for colon and rectal cancer screening and colon polyp surveillance. We compared the number and characteristics of polyps during two periods: July 1, 2022, to June 30, 2023 (intervention group with CADe) and July 2019 to June 2022 (control group without CADe).Results/Outcomes: A total of 821 patients (383 pts, 47% in the interventional group and 438 pts, 53% in the control group) are included in the study. Patient characteristics including age, gender, ASA grade, indications for colonoscopy, and quality metrics of colonoscopy are comparable between control and intervention groups as shown in Table 1. The mean withdrawal time in the interventional group is shortened by 1.1 minutes and is statistically significant (9.34 vs 10.44; p < 0.0001).Characteristics of the polyps including number, size, location, and pathology noted in the control and interventional groups are shown in Table 2. The PDR in the intervention group (38%) was not statistically significant(p=0.26) when compared to the control group (42%). Similarly, the ADR is not statistically significant (p=0.20) between the interventional (30%) and the control groups (34%).Conclusion/

Discussion: Artificial intelligence-assisted colonoscopy, when performed by an experienced colorectal surgeon, does not appear to significantly improve polyp or adenoma detection rates. The primary advantage observed in the study is a reduction in withdrawal time. To validate these findings, it is advisable to conduct a multi-institutional prospective study involving a diverse cohort of skilled surgeons."Characteristics of patients

Characteristics of polyps
"