Local region based active contours for colon tissue segmentation

In this study, a new active contour model based on local region information is proposed to segment colon tissue in CT image. Due to the large amount data of three-dimensional abdominal CT medical image and heterogeneous intensity regions, the traditional active contour models cannot get the ideal result. They usually have the sensitive limitations on the image noise and initialization. Therefore, the local statistical information of object and background has been used in this paper. After redefining the global energy function, a better segmentation result of the heterogeneous intensity images can be realized. The key point is that the initialization contour of colon can be automatically obtained and a signed distance function constraint item has been added into energy function. This model can extract the colon tissue accurately and avoid the evolution process to initialize repeatedly. The experimental results show that the proposed algorithm can apply to extract the colon tissue. Meanwhile, the segmentation accuracy can reach 94.5%.