A Robust Liver Segmentation in CT-images Using 3D Level-Set Developed with the Edge and the Region Information

Abstract

CT-images have been used widely in hospitals around the world. The segmentation of liver from CT-images is important, because it can help medical doctors to have a clear view of the liver with rendering tools. The segmentation's result is also useful for radiotherapy. However, liver segmentation is a challenging task because of the liver's geometrical structure and position and because of the similarity between the liver and its nearby organs about the intensity of voxels. In this paper, we propose a method to segment the liver from CT-images by modeling the segmentation with a proposed level-set method on 3D-space. In combination with the proposed 3D level-set methods, we propose to combine the edge information with the region information into the level-set's energy function. The experimental results are compared with manual segmentation performed by clinical experts and with recently developed methods for liver segmentation. Our proposed method can perform the segmentation more accurate in comparison with the others. It also can produce a surface that is smoother than one resulted from the other methods in the comparison.