Artificial Intelligence-assisted Colonoscopy on Detection of Missed Proximal Lesions

Centers 1. Queen Mary Hospital, Hong Kong, China (Co-ordinating Center) 2. Tan Tock Seng Hospital, Singapore, Singapore 3. Institute of Gastroenterology and Hepatology, Vietnam Union of Science and Technology Association, Hanoi, Vietnam Study population Inclusion: All adult patients, aged 40 or above, undergoing outpatient colonoscopy in the participating centers will be recruited. Exclusion: * history of inflammatory bowel disease * history of colorectal cancer * previous bowel resection (apart from appendectomy) * Peutz-Jeghers syndrome, familial adenomatous polyposis or other polyposis syndromes * bleeding tendency or severe comorbid illnesses for which polypectomy is considered unsafe. Post-randomization exclusion: * Cecum could not be intubated for various reasons * Boston Bowel Preparation Scale (BBPS) score of the proximal colon is \<2 Study design This is a prospective randomized trial comparing the miss rates of proximal colonic lesions by AI assisted colonoscopy or conventional colonoscopy (Fig. 1). The study will be conducted in the Endoscopy Centre of the participating hospitals. Randomization Eligible patients in each center will be randomly allocated in a 1:1 ratio to undergo tandem colonoscopy of the proximal colon first with AI-assistance and follow by conventional white light colonoscopy (Group 1) or conventional white light colonoscopy without AI assistance follow by conventional colonoscopy (Group 2). Proximal colon refers to colonic segment proximal to the splenic flexure. Randomization will be conducted in blocks of 4 by computer generated random sequences and stratified according to indications of colonoscopy (symptomatic vs screening/surveillance). Patients will be blinded to the group assignment.