AI-Assisted Optical Diagnosis (CADx) for Diminutive Colorectal Polyps

This study evaluates whether an artificial intelligence system (GI Genius, Medtronic), already approved by Health Canada, can help doctors accurately identify, in real time during colonoscopy, which small colorectal polyps (5 mm or less) need to be monitored (adenomas) versus those that do not (for example, hyperplastic polyps). For each small polyp found, the endoscopist will first record a diagnosis without the help of the artificial intelligence system, then activate the system and record a second diagnosis after seeing its assessment. Both diagnoses will be compared to the final result from standard pathology testing, which remains the reference standard. This is an observational diagnostic accuracy study: it does not change any clinical care. All polyps continue to be removed and sent for pathology analysis as usual, whether or not the artificial intelligence system agrees with the doctor. The study will take place during colonoscopies already scheduled for standard clinical reasons (screening, surveillance, or diagnostic work-up), with no additional visits, blood draws, imaging, or sedation. Approximately 840 participants will be enrolled across three Canadian centres (Santé Québec - CHUM, McGill University Health Centre, and St. Paul's Hospital, Vancouver). The goal is to determine whether this AI-assisted approach helps doctors reach the internationally recognized performance thresholds (at least 80% sensitivity and 80% specificity) needed to support clinical adoption of real-time optical diagnosis, which could eventually reduce unnecessary pathology testing.