Comparative Analysis of Lesion Detection Rates in Colonoscopy for Gastroenterology Residents Assisted by Artificial Intelligence.

Abstract

Introduction: colorectal cancer (CRC) is a highly prevalent neoplasm worldwide. Colonoscopy is the gold standard for its screening, with the adenoma detection rate (ADR) and the serrated polyp detection rate (SDR) being the most important quality indicators. Artificial intelligence (AI) has emerged as a tool to enhance these rates, particularly for endoscopists in training. Materials and methods: an observational, comparative, and ambispective study was conducted between March 2024 and October 2025. A total of 179 screening colonoscopies performed by a single gastroenterology resident were included. Polyp detection rates (PDR), adenoma detection rates (ADR), and serrated lesion detection rates (SDR) were calculated monthly, comparing procedures performed with and without AI assistance during two specific training periods. Results: overall detection rates were PDR 50.84%, ADR 29.05%, and SDR 5.59%. Following the AI training periods, an increasing trend in the resident's detection rates was observed, culminating in October 2025 with a PDR of 69.23%, ADR of 38.46%, and SDR of 15.38%. Comparative analysis during the training period showed improved performance, although with monthly variability. Conclusions: artificial intelligence assistance during resident training proved beneficial, with an observed trend of improvement in colorectal lesion detection rates. This suggests that AI is a valuable tool in endoscopic training for developing cognitive skills and increasing colonoscopy performance.