Electronic colon cleansing by colonic material tagging and image segmentation for polyp detection: Detection model and method evaluation

Virtual colonoscopy provides a safe, non-invasive approach to screening colonic polyps using medical imaging and computer graphics technologies, where computer aided or automated detection plays an important role. For model-based detection using geometrical features, the entire colonic lumen must be delineated first. A critical step to obtain a cleansed colonic lumen relies on bowel preparation. We have been developing an electronic colon-cleansing technology which minimizes the difficulties associated with the bowel preparation using low-residue diet and contrast solutions, and removes the tagged colonic materials electronically from the abdominal image by image-segmentation methods. In this work, we focused on the model description and method evaluation using volunteer studies. Both subjective and objective criteria were employed in the evaluation. The objective criteria measure whether the electronic cleansing results could meet the requirements of the detection model of geometrical features, while the subjective criteria evaluate whether the entire colonic lumen could be obtained. The results showed that the electronic cleansing technology could extract a clean colonic lumen if the volunteer follows the bowel preparation instructions. The objective validation demonstrated that the error due to partial volume effect can be minimized to a low level and the segmentation method showed a high reproducibility and repeatability. This work provides the evidence that the routine physical colon-washing procedure could be eliminated for automated polyp detection by virtual colonoscopy.