An improved method of colon segmentation in computed tomography Colonography images using domain knowledge

Computed Tomography Colonography is a medical imaging technique used for finding the polyps of different shapes and sizes in colon. First step in this diagnosis procedure is the effective segmentation of colon in presence of contrast used for tagging fecal and endoluminal fluids. In this paper, an improved colon segmentation method is proposed with minimal user intervention. The method is hybrid approach, which is composed of three parts: adaptive smoothing for de-noising the colon lumen by preserving the edges followed by edge based colon boundary recognition, connected component labeling for colon segments delineation and finally colon assessment on axial slices to aid radiologists in decision making. The main finding of the study was, colon was properly segmented at mucous membrane including the base of all the soft tissue structures which is required for the polyp height measurement and the finer details of the colonic content was absolutely preserved. The method was evaluated using retrospectively collected CT scans of 40 patients. The results were qualitatively and quantitatively validated by radiologist through 2D MPR and 3D visualizations. The method was implemented with multithreading for parallel processing in a high performance computer. And it took approximately 2 minutes for processing and segmenting 500 CTC images. The accuracy of result achieved through overlap computation for subset of images was 94.614± 7754%.