AI improves adenoma detection during colonoscopy

Compared with other endoscopic interventions such as high-definition (HD) colonoscopy, artificial intelligence (AI) has exhibited markedly better adenoma detection rate (ADR), as shown by the results of a systematic review and meta-analysis.

“AI improved ADR and polyp detection rate (PDR) compared with majority of endoscopic interventions and should be encouraged for use in clinic setting,” the researchers said.

To assess the impact of AI versus other endoscopic interventions aimed at increasing ADR such as distal attachment devices, dye-based/virtual chromoendoscopy, water-based techniques, and balloon-assisted devices, the researchers carried out a broad search of PubMed/Medline, Embase, and Cochrane through 6 May 2022 for randomized controlled trials (RCTs) comparing ADR for any of these interventions.

A frequentist approach and random effects model were then used to perform a network meta-analysis. The research team calculated the relative risk (RR) and 95 percent confidence intervals (CIs) for proportional outcome.

Ninety-four RCTs, including 61,172 patients (mean age 58.1 years, 45.8 percent female), and 20 discrete study interventions met the eligibility criteria. [J Clin Gastroenterol 2024;58:143-155]

Results of the meta-analysis revealed that AI significantly improved ADR relative to autofluorescence imaging (RR, 1.33, 95 percent CI, 1.06‒1.66), dye-based chromoendoscopy (RR, 1.22, 95 percent CI, 1.06‒1.40), endocap (RR, 1.32, 95 percent CI, 1.17‒1.50), endocuff (RR, 1.19, 95 percent CI, 1.04‒1.35), endocuff vision (RR, 1.26, 95 percent CI, 1.13‒1.41), and endoring (RR, 1.30, 95 percent CI, 1.10‒1.52).

AI also demonstrated better ADR than other interventions such as linked colour imaging (RR, 1.21, 95 percent CI, 1.08‒1.36), full-spectrum endoscopy (RR, 1.40, 95 percent CI, 1.19‒1.65), flexible spectral imaging colour enhancement (RR, 1.26, 95 percent CI, 1.09‒1.46), HD colonoscopy (RR, 1.41, 95 percent CI, 1.28‒1.54), narrow band imaging (RR, 1.33, 95 percent CI, 1.18‒1.48), water exchange (RR, 1.22, 95 percent CI, 1.06‒1.42), and water immersion (RR, 1.47, 95 percent CI, 1.19‒1.82).

Breakthrough

Recent studies have demonstrated the beneficial effect of using AI to improve ADR when compared with HD colonoscopy, but none of these studies compared AI to other endoscopic interventions. [Gastroenterology 2020;159:512-520; Endoscopy 2021;53:277-284]

“Our study is the first one to provide an indirect comparison of AI to other endoscopic intervention using the network meta-analysis technique,” the researchers said. “The results of our study demonstrated superiority of AI compared with majority of endoscopic interventions in improving both ADR and PDR.”

Additionally, serrated poly lesion detection rate (SPDR) is another essential metric being considered during colonoscopy. Sessile serrated adenoma or polyps and traditional serrated adenomas can become cancerous and must be removed to prevent colorectal cancer (CRC). [Gastrointest Endosc 2019;90:721-731; Eur J Gastroenterol Hepatol 2020;32:1284-1292; Gastroenterology 2010;138:2088-2100]

However, the current study did not display improved SPDR with AI relative to other interventions. One possible explanation was the lack of screening or surveillance population in AI trials. Only two studies evaluated SPDR using AI.

“Colonoscopy is an effective screening modality that decreases the incidence and mortality associated with CRC by detecting premalignant polyps,” the researchers said. “Current efforts are directed to increase the ADR which in turn reduces interval CRC.” [Am J Gastroenterol 2021;116:458-479]