A knowledge based approach for colon segmentation in CT colonography images

Computed Tomography Colonography (CTC) is a medical imaging and diagnosis procedure for finding the polyps of different shapes and sizes in large intestine using computer based software. Segmenting the colon exactly at the colon wall in presence of oral contrast used for fecal tagging and completely cleansed colon is an important prerequisite, based on which the measurement of polyp relies on. The objective of the study was to segment the colon for polyp analysis. This paper proposes an expert system with boundary based semi-automatic segmentation method which uses a) adaptive smoothing for de-noising the colon lumen by preserving the edges, b) the canny operator for colon boundary recognition, c) connected component labelling for colon segments delineation and prominently d) the translation of Radiologist's perspective of colon assessment on axial slices in to the decision making system. This was a retrospective study and the method was applied on 40 patient's dataset. The main finding of the study was, the proposed approach accurately identified the colon wall. This avoids the misclassification of polyps. The novelty of this approach is discussed. Multithreading concept in a high performance computer was implemented with parallel processing. It takes ~2 minutes for segmenting 500 CTC images. The results were validated by Radiologist through 2D and 3D visualization.