Colonoscopy is clinically used as the gold standard for detection of colon cancer (CRC) and removal of adenomatous polyps. Despite the success of colonoscopy in reducing cancer-related deaths, there exists a disappointing level of adenomas missed at colonoscopy. "Back-to-back" colonoscopies have indicated significant miss rates of 27% for small adenomas (< 5 mm) and 6% for adenomas of more than 10 mm in diameter. Studies performing both CT colonography and colonoscopy estimate that the colonoscopy miss rate for polyps over 10 mm in size may be as high as 12%. The clinical importance of missed lesions should be emphasized because these lesions may ultimately progress to CRC.

Limitations in human visual perception and other human biases such as fatigue, distraction, level of alertness during examination increases such recognition errors and way of mitigating them may be the key to improve polyp detection and further reduction in mortality from CRC. In the past years, a number of CAD systems for detection of polyps from endoscopy images have been described. However, the benefits of traditional CAD technologies in colonoscopy appear to be contradictory, therefore they should be improved to be ultimately considered useful. Recent advances in artificial intelligence (AI), deep learning (DL), and computer vision have shown potential to assist polyp detection during colonoscopy.

Average experienced endoscopists (each having performed <2000 screening colonoscopies) will perform the endoscopic procedure.

Inclusion Criteria:

• All 40-80 years-old subjects undergoing a colonoscopy

Exclusion Criteria:

• subjects with personal history of CRC, or IBD.
• patients with inadequate bowel preparation (defined as Boston Bowel Preparation Scale > 2 in any colonic segment).
• patients with previous colonic resection.
• patients on antithrombotic therapy, precluding polyp resection.
• patients who were not able or refused to give informed written consent.

Based on the observed prevalence of adenomas (35%) among patients undergoing colonoscopies at our center within the last 12 months, a sample size of 322 subjects per arm could allow for a 90% power to show the non-inferiority (primary end-point) of the AI-aided arm by excluding that the one-side 95% CI will exclude a difference of 10% in favour of the standard group. Such sample size will also have a 80% power to detect as statistical significant (α=0.05; two-sided test) a 10% absolute increase in the detection rate of adenomas in the AI-aided arm (secondary end-point).

Performing a sub-stratification according to operator experience, we will planning to enroll the 322 subjects per arm performed by experts and 322 subjects per arm performed by average operator.