AI bests high-definition white light colonoscopy in adenoma detection
Artificial intelligence-based computer-aided polyp detection decreased adenoma miss rate and increased first-pass adenoma per colonoscopy detection, according to research published in Clinical Gastroenterology and Hepatology.
“When we initiated this study, there was single-center data published in China that suggested that use of a deep-learning based computer aided detection system (CADe) might increase adenoma detection rate and therefore might improve colonoscopy quality, but there was no prospective data examining how a CADe system might perform in a diverse U.S. patient population presenting for screening or surveillance colonoscopy,” Jeremy R. Glissen Brown, MD, fellow in gastroenterology and hepatology at Beth Israel Deaconess Medical Center and Harvard Medical Center, told Healio Gastroenterology. “In addition, there were no studies that looked on a granular level at the miss rate of a CADe system combined with a human operator compared with a human operator alone.”
To assess the comparative adenoma miss rate for CADe-assisted colonoscopy compared with routine high-definition white light (HDWL) colonoscopy, researchers enrolled 232 patients presenting for colorectal cancer screening or surveillance in the CADe tandem colonoscopy study. Patients underwent CADe colonoscopy first, followed immediately by HDWL colonoscopy (n = 116) or vice versa. Secondary endpoints included sessile serrate lesion miss rate and adenoma per colonoscopy; the final study cohort consisted of 223 patients.
After colonoscopy, researchers observed a decreased adenoma miss rate among patients who underwent CADe first compared with patients who underwent HDWL first (20.12% vs. 31.25%; OR = 1.8; 95% CI, 1.08-3.02). Further, the CADe-first group demonstrated a lower sessile serrate lesion miss rate (7.14% vs. 42.11%) and polyp miss rate (20.7% vs. 33.71%) as well as a higher first-pass ADR rate (50.44% vs. 43.64%) and adenoma per colonoscopy rate (1.19 vs. 0.9). Multivariate logistic regression analysis revealed randomization to HDWL first vs. CADe first (OR = 1.88), older age ( 65 years; OR = 1.74) and location (right colon vs. other locations; OR = 1.79) correlated with missed adenomas. Overall, researchers noted a relative reduction of 35.61% in adenoma miss rate among patients who underwent CADe first, with an absolute difference of 11.09%.
“Our study is the first to show the potential benefit of CADe for use during colonoscopy in a diverse patient population presenting for screening or surveillance colonoscopy in the U.S.,” Glissen Brown said. “This study suggests that standard use of CADe during screening and surveillance colonoscopy might improve colonoscopy quality. We will likely see the approval of multiple CADe systems; future studies should continue to look at CADe among a variety of patient populations and should look at the role of the technology in different care settings.”
World’s first external independent randomized controlled trial of artificial intelligence demonstrates more accurate colonoscopy screening and surveillance utilizing Wision AI’s technology compared to the standard of care. https://www.globenewswire.com/news-release/2021/09/29/2305339/0/en/World-s-First-External-Independent-Randomized-Controlled-Trial-of-Artificial-Intelligence-Demonstrates-More-Accurate-Colonoscopy-Screening-and-Surveillance-Utilizing-Wision-AI-s-Te.html. Published Sept. 29, 2021. Accessed Sept. 29, 2021.