A RECENT study examining AI-assisted colonoscopy (AIAC) with computer-aided detection (CADe) technology found that, while AIAC improves adenoma detection rates (ADR), it also increases the detection of non-neoplastic lesions. This leads to more resection of benign lesions during colonoscopies, raising concerns about unnecessary interventions.
The study, conducted at a single centre, compared the outcomes of AIAC and non-AIAC colonoscopies performed for screening, surveillance, and faecal immunochemical test-positive patients. Data from 441 non-AIAC colonoscopies (November 2022–April 2023) were compared with 599 AIAC colonoscopies (May–October 2023). The cohorts were balanced for patient demographics, endoscopists, AI technology used, procedure time, and the number of polyps detected.
Results showed that 37 cases (8.4%) in the non-AIAC cohort involved polypectomies revealing only benign lesions, while 74 cases (12.4%) in the AIAC cohort involved the resection of benign lesions alone. This difference was statistically significant (P=0.04). The most common benign lesions resected included benign colonic mucosa, lymphoid aggregates, and hyperplastic polyps.
These findings suggest that AIAC significantly increases the likelihood of a colonoscopy leading to a therapeutic procedure for removing non-adenomatous lesions. This raises concerns about the potential for unnecessary interventions and additional healthcare costs without further benefit in colorectal cancer prevention.
The authors call for the development of hybrid CADe and computer-aided diagnosis (CADx) systems to address these concerns, helping to mitigate excessive benign lesion resections while maintaining the benefits of AI in cancer prevention.
Ada Enesco, EMJ
Reference
Herman et al. Artificial intelligence-assisted colonoscopy is associated with higher conversions from screening to therapeutic exams. S836. Am J Gastroenterol. 2024;119(10S):S575-S576.
