Artificial intelligence assisted colorectal lesion detection in private practices a randomized controlled study

Abstract Computer-aided colonoscopy (CAC) may improve polyp detection and characterization compared to traditional colonoscopy (TC). However, recent studies also reported no relevant effect on adenoma detection rate (ADR). This study evaluates the real-time polyp detection system EndoMind during screening and surveillance colonoscopy in a multicenter randomized controlled trial. From November 2021 to November 2022, 933 individuals undergoing colorectal cancer screening or post-polypectomy surveillance were recruited and randomized in five outpatient treating centers (10 examiners; >10 years of experience). 914 Patients were included in the intention to treat analysis (CAC:452, TC:462) and detected lesions were framed on the primary monitor in the CAC group. More than 94% of the examinations were screening or surveillance colonoscopies with overall similar patient characteristics. The ADR (CAC:34.5% vs TC:32.9%; p = 0.656) was not significantly different between the groups. The effect of CAC on ADR remains a controversial discussion. Diverging study setups and patient collectives complicate consistent comparisons. Future studies should focus on large-scale real-world populations. ClinicalTrials.gov:NCT05006092 (registered: 2021-08-06).