Novel Artificial intelligence systems in detecting adenomas in colonoscopy: a systemic review and network meta-analysis.

Abstract Background: The use of artificial intelligence (AI) in detecting colorectal neoplasia during colonoscopy holds the potential to enhance adenoma detection rates (ADRs). This meta-analysis compares the efficacy of multiple AI-assisted systems developed to detect adenomas during colonoscopy. Methods: An extensive literature search of the Embase, Medline, and Cochrane Library databases was conducted, employing both Medical Subject Headings (MeSH) and pertinent keywords from the inception of these databases until March 4, 2024. This search aimed to identify randomized controlled trials (RCTs) comparing AI-assisted and conventional colonoscopy methods in the detection of colorectal neoplasia. Notable AI-assisted systems such as GI-Genius (Medtronic), CAD-EYE (Fujifilm), Endoangel, Endoscreener, and EndoAID were scrutinized. The primary measure of interest was the ADR, for which data was aggregated using a random effects model. This model facilitated the calculation of pooled odds ratios (OR) for ADR with a 95% confidence interval (CI), thereby summarizing the efficacy of each system under review. Additionally, the Surface Under the Cumulative Ranking Curve (SUCRA) probabilities were used to assess relative efficacy rankings. Subgroup analyses were conducted, and heterogeneity within each group was meticulously evaluated. Results: Our search criteria yielded 17 eligible RCTs encompassing 10,547 participants. EndoAngel ranked first for efficacy with an OR for ADR 1.84 ( SUCRA, 0.9), closely followed by EndoAID with an OR of 1.64 (95% CI, 1.20 to 2.26; SUCRA, 0.7). CAD-EYE and GI-Genius were jointly ranked third, each demonstrating a similar efficacy (OR, 1.46; CI, 1.21 to 1.75; SUCRA, 0.5) and (OR, 1.45; CI, 1.22 to 1.73; SUCRA, 0.5) respectively. Endoscreener was placed just above the control group, with an OR 1.37 (95% CI, 1.20 to 1.56; SUCRA, 0.4). Notably, the GI-Genius group exhibited a higher degree of study heterogeneity (I^2 = 67%). Conclusion: Our findings advocate for EndoAngel's superior efficacy in adenoma detection relative to traditional colonoscopy techniques. These results highlight the need for further research to refine and enhance the capabilities of AI-assisted adenoma detection systems, many of which are still in developmental stages. Keywords: Artificial intelligence; Adenoma detection rate; colorectal neoplasia; colonoscopy;