Colonoscopy is one of the most effective ways to prevent colorectal cancer. It allows doctors to examine the colonic wall for polyp(s) and remove those that are adenomatous. Nevertheless, two major difficulties are encountered by endoscopists during colonoscopy: namely missed polyps and misclassified polyps. To solve the issues, we develop a real-time computer-aided diagnosis system for polyp detection and classification during colonoscopy. The system, named AIdoscopist, displays the diagnosed result of each polyp to doctors in real time and assists doctors in making instant decision of whether a polyp should be resected or not. The ultimate goal is to make colonoscopy a safer and more cost-effective procedure. Currently, the system has been:

• Built with advanced deep learning method and incorporated both spatial and temporal information in the prediction model;
• Tested with full colonoscopy videos collected from 300 patients and achieved comparable results with a group of endoscopists; and
• Evaluated on two patients for real-time processing and displaying the results