PHILADELPHIA—The use of artificial intelligence–assisted colonoscopy can increase adenoma detection rates in the right colon among early trainees, according to research presented at ACG 2024.

Image
AI-assisted colonoscopy may help fellows increase their adenoma detection rates in the right colon, according to a study presented at ACG 2024.
The investigators, led by Denis Nguyen, MD, a third-year internal medicine resident at the Keck School of Medicine at the University of Southern California, in Los Angeles, conducted a randomized controlled trial (ClinicalTrial.gov Identifier: NCT05423964) to compare the ADRs in AI-assisted versus conventional colonoscopies (abstract P1931). They assessed overall ADR, as well in the right colon (RADR) and left colon (LADR), with stratification by year of fellowship.

 

Sixteen fellows were randomized to AI-assisted (via GI Genius, Medtronic) versus conventional colonoscopy, with those in the AI-assisted group performing a total of 401 colonoscopies and those in the conventional group performing 644 colonoscopies during the study (March to October 2023).

The mean ADR and LADR were each similar between the arms, but the mean RADR was 7.6 percentage points higher (95% CI, 1.7%-13.5%) in the AI-assisted group (24.1%) than the conventional group (16.5%). When stratifying by year of fellowship (1-3), there were no significant differences between the AI-assisted and conventional arms with respect to overall ADR, RADR or LADR.

When considering screening colonoscopies only (AI-assisted, 57; conventional, 73), RADR was higher for the AI-assisted than conventional colonoscopies for first-year fellows (33.3% vs. 5.5%; difference, 27.8 percentage points; 95% CI, 7.2%-35.2%) and second-year fellows (33.3% vs. 10.7%; difference, 22.6 percentage points; 95% CI, 5.0%-40.2%). However, the statistical significance was not as high for third-year fellows (40.0% vs. 22.2%; difference, 17.8 percentage points; 95% CI, –9.6% to 45.2%). There were no significant differences in LADR between the AI-assisted and conventional colonoscopies for any year of fellowship.

To the investigators, the findings in the right colon were of particular interest, given the American College of Gastroenterology’s emphasis on the need for training to detect right colon lesions. Dr. Nguyen told Gastroenterology & Endoscopy News that considering that the RADR was higher in first- and second-year fellows but not third-year fellows who used AI-assisted versus conventional colonoscopy, “we can deduce that the fellows themselves improve” over time, “AI might not be as helpful in those providers who already have high ADRs.”

To Dr. Nguyen, the primary “benefit of AI-assisted colonoscopy is in education, especially to help early learners.”

—Natasha Albaneze, MPH

Dr. Nguyen reported no relevant financial disclosures.