A Fast, Automatic System for Detecting Flat Polyps in Colonoscopy Video

Abstract Colorectal cancer is the second most common cause of cancer death in the United States, with an estimated 140,000 new cases leading to 50,000 deaths this year. The best treatment is to detect and treat the cancer before it becomes invasive and spreads. The most common form of detection is the use of optical colonoscopy in which the clinician visually inspects the surface of the colon through an endoscope to detect the presence of polyps. Studies have shown that even the best clinicians will sometimes miss polyps, especially the more subtle flat polyps, and that many cancers that develop in the years immediately following a colonoscopy likely originate from missed polyps. Many approaches have been used to attempt to improve polyp miss rates but have yielded little to no benefit. We propose to develop an approach for automatically detecting the presence of flat polyps during a colonoscopy procedure to reduce the rate of missed polyps in colonoscopy. This project is a close collaboration between Kitware, Inc. and the University of North Carolina at Chapel Hill (UNC). Kitware is scientific computing company known for creating high-quality open-source software and our co-investigators from UNC bring a long history of medical image analysis and clinical experience. The proposal team has done preliminary work on developing features for detecting polyps and has a strong history of collaboration. In this project, we propose to combine the UNC team's expertise with Kitware's algorithm implementation and software development experience to create the foundation for a system that can be used to automatically detect the presence of flat polyps in colonoscopy video to allow the clinician to return to the site of the polyp and remove it if it was originally missed. The specific aims of the proposed project are to (a) develop and evaluate a set of features that can be used to detect the presence of flat polyps during colonoscopy and (b) accelerate the computation of these features to be useful in a clinical scenario. The successful completion of this work will yield a proven set of features and an implementation that will form the basis of an automatic polyp detection system.