Machine Learning for Automated, Video-Based Measurement of Colonoscopy Skill to Reduce Colorectal Cancer (CRC) Risk

We propose to reduce CRC risk by developing an automatically abstracted measure of colonoscopy technical skill, generated by machine learning analysis of video-recorded colonoscopies, that correlates with the ability of a colonoscopist to prevent CRC by identifying pre-malignant polyps. This automated algorithm will generate a novel quality measure that will be validated in colonoscopists of varying quality using existing measures and applicable to all individual clinicians performing colonoscopy, regardless of procedure volume. This novel quality measure will provide a rapid, scalable method of colonoscopy quality assessment. Ensuring high quality colonoscopy on a large scale will reduce the incidence of CRC.