Orzeszko Z, Gach T, Necka S, Ochwat K, Major P, Szura M. The implementation of computer-aided detection in an initial endoscopy training improves the quality measures of trainees' future colonoscopies: a retrospective cohort study. Surg Endosc. 2025 Aug;39(8):5276-5286. doi: 10.1007/s00464-025-11890-3

PMID: 40588603

Abstract

Introduction: The implementation of computer-aided detection (CADe) systems has resulted in a growing number of young endoscopists being trained using AI-enhanced devices. The potential impact of AI-enhanced training on the trainees' future performance is undefined. This study aimed to evaluate the quality indicators of endoscopists trained in an AI environment compared to those trained conventionally.

Methods and procedure: In this retrospectively study, the independent performance of six endoscopists was evaluated after they had undergone initial training using either CADe (group A) or conventional endoscopy (group B: without CADe). Quality indicators and detection rates of laterally spreading tumors (LSTs) were compared between the two groups.

Results: A total of 6000 patients were included in the analysis. Groups were equal demographically and had similar cecal intubation rate. Withdrawal time (WT) was longer in the AI-trained group (mean difference 0.8 min; 95% confidence interval [CI]: 0.6-1.0). AI-trained group had also a significantly improved adenoma detection rate (ADR) by 5.3% (95% CI: 2.9-7.6%) and sessile lesion detection rate (SDR) by 5.4% (95% CI: 3.8-7.0%). AI-assisted training enhanced the detection of non-granular LSTs smaller than 20 mm by 0.2% (95% CI 0.1% to 0.4%) and was identified as a factor of high-quality performance in terms of ADR and SDR (OR 1.27, 95% CI: 1.14-1.42; OR 1.93, 95% CI: 1.10 to 3.37, respectively).

Conclusions: Endoscopists trained in colonoscopy using AI exceeded the aspirational targets of the quality guidelines when those trained conventionally achieved minimum quality measures.