Automatic Segmentation of Colon Using Multilevel Morphology and Thesholding

The segmentation of colon may faces problems such as presence of diaphragm, air packets and other extra colonic components such as bones, lungs, stomach and small bowels. The proposed method uses morphological based Otsu thresholding method for the accurate segmentation of colon. Before starting the segmentation process, diaphragm present at the bottom of the abdomen is removed by using erosion followed by dilation. In segmentation process, the air packets and some non-colon parts are completely removed by implementing the Otsu thresholding. Then the other non-colon components such as small bowels are removed by morphological operations. The effectiveness of this segmentation scheme is compared with the existing methods and is analyzed. This method could obtain an average of accuracy of 99.05%, specificity of 99.70% and sensitivity of 87% in comparison with other existing methods.