AI Found More Adenomas During Surveillance Colonoscopy: Digestive and Liver Disease | October 2026

What Did the Study Find?

GENIAL-CO randomised adults aged 50–74 undergoing post-polypectomy surveillance within an organised Italian colorectal-cancer screening programme to computer-aided detection or standard colonoscopy. Of 973 enrolled patients, 914 with adequate bowel preparation were analysed—455 with CADe and 459 controls.

CADe did not significantly improve the primary endpoint: adenoma detection was 66.6% versus 61.0% (P=0.160). Advanced-adenoma detection was similarly unchanged at 7.3% versus 6.3%, while sessile-serrated-lesion detection was 17.1% versus 14.2%. CADe did increase adenomas per colonoscopy from 1.32 to 1.80, but required longer withdrawal—16.35 versus 14.52 minutes, an additional 1.83 minutes per procedure. AI Medical Compendium

Why Does It Matter?

Finding additional small adenomas does not necessarily translate into more patients with clinically consequential neoplasia. In a high-performing surveillance programme with an already high baseline ADR, routine CADe added procedure time without significantly increasing advanced-lesion detection. Results may differ in lower-ADR settings, but they argue against assuming that AI automatically improves every colonoscopy population.

GastroAGI Takeaway

Judge colonoscopy AI by advanced-neoplasia and patient-level detection—not simply by how many additional diminutive adenomas the software highlights.