Abstract
INTRODUCTION: Artificial intelligence (AI) tools for polyp detection during colonoscopy have shown promise in improving polyp detection rate (PDR) and adenoma detection rate (ADR). However, in real-world settings, the results are inconsistent, particularly among experienced endoscopists. This study evaluates the impact of AI on ADR and PDR when used by a single high-volume, experienced endoscopist. METHODS: A retrospective analysis was conducted from July 2019 to June 2023, involving colonoscopies performed by a single high-volume endoscopist. Patients undergoing screening or surveillance colonoscopy were included, while those with diagnostic indications were excluded. Outcomes were compared between two groups: pre-AI-assisted colonoscopy (pre-AIAC) (July 2019-June 2022) and AI-assisted colonoscopy (AIAC) (July 2022-June 2023). RESULTS: Of 1,273 colonoscopies, 814 met the inclusion criteria. Quality metrics, including the Boston Bowel Prep Score and withdrawal time greater than six minutes, were comparable between the groups. ADR was 151 (35%) in the pre-AIAC group and 116 (31%) in AIAC group, with no statistical significance (p = 0.2). Similarly, PDR, 184 (42%) vs. 145 (38%); p = 0.22, and adenomas per colonoscopy, 0.62 (1.23) vs. 0.51 (0.97); p = 0.22, showed no statistically significant differences. However, the mean withdrawal time was significantly shorter in the AIAC group (9.34 vs. 10.44 minutes; p < 0.0001). CONCLUSION: AIAC may not significantly improve adenoma or PDRs when performed by an experienced endoscopist. However, it was associated with a significant reduction in mean withdrawal time. Further research is needed to explore these findings in broader clinical settings.