For the sensitivity of the portal vein data to classical liver functional segmentation method, a liver segment method based on hierarchical vascular tree combining with the Couinaud theory and portal vein distribution characteristics is proposed. Firstly, liver and vessels are extracted from the abdominal CT image by image segmentation and skeletonization methods. Secondly, secondary subtree set was determined through statistical analysis on average radius of vascular branches, so as to divide the secondary subtree set into several different classes by k-means++ clustering algorithm according to their own blood-supply area. Thirdly, a nearest neighbor segment approximation algorithm was used to segment the liver into parts. Finally, the internal anatomical structure of liver and its vascular system was demonstrated using three-dimensional visualization technology, and then making annotations on liver segments to extract clinical interest information. Experimental result shows that the method can obtain good results when vascular tree contains plenty branches and complex structure. Furthermore, for considering the impact of major secondary branches, the final liver segment distribution and attribute results are in line with the Couinaud liver segment theory.