A 3-D liver tumor level set method based on narrowband gradient modification

Liver tumor segmentation is a hot issue in current medical image processing research, fast and accurate liver tumor segmentation method for abdominal CT sequences is the basis for liver lesions diagnose. The evolving curve in traditional level set algorithm often stopped in local gradient minimal regions or false edges while dealing with low contrast or borderline blurred CT images. In order to solve this problem, a level set method based on narrowband gradient modification is proposed in this paper. First, morphological gradient transformation was performed on the image to get the gradient image. Then it set up function relationship between structural elements radius and gradient value, and modified the image gradient. Using level set method to segment the image according to gradient information. Finally using the result as the initial contour of the adjacent CT sequence, built narrowband with a special width, split the liver tumor within the narrowband, repeat the process until getting all the results of the entire slices, then 3-D reconstructed.