The impact of artificial intelligence on the adenoma detection rate : Comparison between experienced, intermediate and trainee endoscopists' adenoma detection rate.

Abstract

Background: Artificial intelligence (AI) is a promising tool to achieve a high adenoma detection rate (ADR). The aim of this study is to evaluate the impact of a computer-aided detection (CADe) device on the ADRs of endoscopists with different levels of expertise.

Methods: Data were collected from patients who underwent colonoscopy with CADe within a 12-month period. Endoscopists were divided into three groups, a trainee group (< 500 colonoscopies), an intermediate group (500-1000 colonoscopies) and an expert group (> 2000 colonoscopies). Endoscopists with the same definition of experience without CADe support served as the control cohort. For the differences in ADR between the groups a 2-sided 95% confidence interval (CI) and odds ratios (OR) were calculated.

Results: In this study 335 patients (155 females, 177 males) with a mean age 62.1 years (SD ± 16.2 years) were included in the CADe cohort. In this cohort 508 polyps were resected. The ADRs for the groups and control groups (without CADe) were as follows: 42.9% (95% CI: 28.5-57.2%) and 21.5% (95% CI: 11.3-31.8%) in the trainee group, 41.3% (95% CI: 33.5-49.0%) and 36.8% (95% CI: 27.9-45.6%) in the intermediate group and 39.8% (95% CI: 30.9-48.8%) and 33.3% (95% CI: 26.3-40.4%) in the expert group. There were no significant differences among the CADe groups when trainees were compared to experts (p = 0.72, OR 1.13, 95% CI: 0.58-2.16) or when intermediate endoscopists were compared to experts (p = 0.81, OR 1.06, 95% CI: 0.65-1.74).

Conclusion: The use of AI appears to provide an opportunity to match the ADR-based quality of colonoscopy at an early stage of endoscopy training with experts.