The following is a summary of “Evaluation of Computer-Aided Detection During Colonoscopy in the Community (AI-SEE): A Multicenter Randomized Clinical Trial,” published in the October 2023 issue of Gastroenterology by Wei, et al.
Artificial intelligence (AI) has garnered growing attention in gastroenterology, particularly on reducing the miss rates during colonoscopy. Computer-aided detection (CADe) systems have been explored extensively to assist in polyp detection. For a study, researchers sought to evaluate the use of CADe during colonoscopy in community-based, non-academic gastroenterology practices.
A randomized controlled trial, AI-SEE, was conducted from September 28, 2020, to September 24, 2021. The trial assessed the impact of CADe on polyp detection in four community-based endoscopy centers in the United States. Patients were randomized to undergo colonoscopy with or without CADe (specifically, EndoVigilant). The primary outcome measures included adenomas per colonoscopy and adenomas per extraction, which quantified the percentage of polyps removed that were adenomas. Secondary endpoints encompass serrated polyps per colonoscopy, non-adenomatous and non-serrated polyps per colonoscopy, adenoma and serrated polyp detection rates, and procedural time.
A total of 769 patients participated in the study, with 387 in the CADe group and similar demographic characteristics in both groups. The use of CADe did not significantly affect the number of adenomas detected per colonoscopy compared to the non-CADe group (0.73 vs. 0.67, P = 0.496). However, CADe did increase the identification of non-adenomatous, non-serrated polyps per colonoscopy (0.90 vs. 0.51, P < 0.0001), subsequently detecting fewer adenomas per extraction in the CADe group. Adenoma detection rates (35.9% vs. 37.2%, P = 0.774) and serrated polyp detection rates (6.5% vs. 6.3%, P = 1.000) were similar between the CADe and non-CADe groups. The CADe group exhibited a longer mean withdrawal time than the non-CADe group (11.7 vs. 10.7 minutes, P = 0.003). However, in cases where no polyps were identified, both groups had similar mean withdrawal times (9.1 vs. 8.8 minutes, P = 0.288). No adverse events were reported.
The study did not find a statistically significant difference in adenoma detection using CADe. Further research was required to understand better the factors contributing to significant benefits from CADe in some endoscopists but not others.