322 Artificial Intelligence-Aided Colonoscopy: A Retrospective Analysis of Effect on Procedure Time

Abstract

INTRODUCTION: Artificial intelligence (AI) for colonoscopy promises to assist polyp discovery, in situ diagnosis and automated procedure documentation. Widespread adoption will require ease of use and value, defined as: improved quality at reduced cost. The aim of this study is to determine if use of AI lengthens procedure time. METHODS: Data was collected prospectively at a single academic medical center at point-of-care using UCICQD, inclusive of all GIQuIC data and locations, sizes and pathology of all polyps stripped of all PHI and IRB-exempt. AI-assisted colonoscopies (AC) and conventional colonoscopy (CC) were compared for differences in means of total procedure time (TT). RESULTS: A total of 8988 CCs and 166 ACs were analyzed without exclusions. Insertion time to cecum averaged 13.0 minutes in CC (95% CI 12.6-13.3 minutes) vs 11.4 minutes (95% CI 10.3-12.4 minutes) in AC, P = 0.007. Withdrawal time averaged 18.2 minutes in CC (95% CI 17.9-18.4 minutes) vs 18.1 minutes (95% CI 16.0-20.1 minutes) in AC, P = 0.03. Total procedure time averaged 31.1 minutes in CC (95% CI 30.7-31.6 minutes) vs 29.4 minutes (95% CI 27.2-31.7 minutes) in AC, P = 0.79. CONCLUSION: While excitement is high for AI assistance during colonoscopy, concerns about potential negative effects on efficiency in high volume ASC's remain. This retrospective comparison of AC and CC demonstrates that the use of AI is not associated with increased procedure time.