Efficacy of artificial intelligence in reducing miss rates of GI adenomas, polyps, and sessile serrated lesions: a meta-analysis of randomized controlled trials

Xi-Feng Jin; Hong-Yan Ma; Jun-Wen Shi; Jian-Ting Cai

doi:10.1016/j.gie.2024.01.004

Efficacy of artificial intelligence in reducing miss rates of GI adenomas, polyps, and sessile serrated lesions: a meta-analysis of randomized controlled trials

Gastrointest Endosc. 2024 May;99(5):667-675.e1. doi: 10.1016/j.gie.2024.01.004. Epub 2024 Jan 4.

Authors

Xi-Feng Jin¹, Hong-Yan Ma², Jun-Wen Shi², Jian-Ting Cai³

Affiliations

¹ Department of Gastroenterology, The Second Affiliated Hospital Zhejiang University School of Medicine, Hangzhou, China. Electronic address: xfj2022@zju.edu.cn.
² Tengzhou Central People's Hospital, Shandong Province, Zaozhuang, China.
³ Department of Gastroenterology, The Second Affiliated Hospital Zhejiang University School of Medicine, Hangzhou, China.

PMID: 38184117
DOI: 10.1016/j.gie.2024.01.004

Abstract

Background and aims: The aim of this study was to determine if utilization of artificial intelligence (AI) in the course of endoscopic procedures can significantly diminish both the adenoma miss rate (AMR) and the polyp miss rate (PMR) compared with standard endoscopy.

Methods: We performed an extensive search of various databases, encompassing PubMed, Embase, Cochrane Library, Web of Science, and Scopus, until June 2023. The search terms used were artificial intelligence, machine learning, deep learning, transfer machine learning, computer-assisted diagnosis, convolutional neural networks, gastrointestinal (GI) endoscopy, endoscopic image analysis, polyp, adenoma, and neoplasms. The main study aim was to explore the impact of AI on the AMR, PMR, and sessile serrated lesion miss rate.

Results: A total of 7 randomized controlled trials were included in this meta-analysis. Pooled AMR was markedly lower in the AI group versus the non-AI group (pooled relative risk [RR], .46; 95% confidence interval [CI], .36-.59; P < .001). PMR was also reduced in the AI group in contrast with the non-AI control (pooled RR, .43; 95% CI, .27-.69; P < .001). The results showed that AI decreased the miss rate of sessile serrated lesions (pooled RR, .43; 95% CI, .20 to .92; P < .05) and diminutive adenomas (pooled RR, .49; 95% CI, .26-.93) during endoscopy, but no significant effect was observed for advanced adenomas (pooled RR, .48; 95% CI, .17-1.37; P = .17). The average number of polyps (Hedges' g = -.486; 95% CI, -.697 to -.274; P = .000) and adenomas (Hedges' g = -.312; 95% CI, -.551 to -.074; P = .01) detected during the second procedure also favored AI. However, AI implementation did not lead to a prolonged withdrawal time (P > .05).

Conclusions: This meta-analysis suggests that AI technology leads to significant reduction of miss rates for GI adenomas, polyps, and sessile serrated lesions during endoscopic surveillance. These results underscore the potential of AI to improve the accuracy and efficiency of GI endoscopic procedures.