Objective Method for Colon Cleansing Evaluation Using Color CIELAB Features

摘要

An objective method for identification of certain types of regions in a colonoscopic frame image, such as intestinal mucosa, intestinal residues or artifacts was developed. The novelty of the method consists in the prior selection of several rectangular image fragments representing these particular regions, their automatic processing in order to obtain very precise color sets (colorprints) for the intestinal mucosa, residues and artifacts and the possibility of improving these colorprints in the CIELAB color space using an expansion technique. The resulting colorprints can be applied directly to colonoscopy video sequences that can be processed in several hours. Based on the number of pixels contained by each type of region, it can be automatically determined whether or not a frame contains information relevant to the diagnosis and the exactly percentage of residues within each frame. This percentage is further used in computing an equivalent Boston Bowel Preparation Score, more accurate, more objective than the human evaluation actually practiced in the majority of gastroenterology clinics.