Virtual colonoscopy with computer-assisted polyp detection

We have developed a computer-assisted polyp detection (CAPD) algorithm that identifies potential lesions during virtual colonoscopy (VC) examinations. The system works by generating a surface-rendered three-dimensional (3D) model of the colon using helical computed tomography (CT) data, calculating colon wall thickness at regularly spaced intervals along the entire 3D surface, grouping areas of abnormal wall thickness into potential lesions, refining the list of lesions using shape descriptors, and presenting a final electronic list of potential lesions to a radiologist. After a radiologist selects a lesion from the list, the computer automatically positions his or her viewpoint towards the lesion for final interpretation. Although VC is an exciting new technique because of its ability to visualize colon polyps and masses, the analysis of VC images is often tedious, time-consuming, and prone to error of omission. Computer-assisted detection promises to overcome these barriers by providing an efficient and accurate means for analyzing VC examinations.