A graph description and analysis framework of liver vascular system

Accurate description and analysis of liver vascular system is critical in diagnosis and treatment of liver diseases. Even though there are plenty of researches put their efforts on the acquisition of optimal visual information in liver vascular system, quite a few studies focus on the quantitatively geometrical and structural analysis on them. In this paper, we proposed a graph description and analysis framework for liver vascular structure representation. First of all, the vessels were enhanced through Hessian eigenvalue calculation in order to be conveniently segmented from the liver image. Secondly, a three dimensional thinning based skeletonization method was applied on the vessel image to obtain the topological structure of the liver vessels. The skeleton result was then represented as a graph, followed by the geometrical attributes computation. Finally, the graph was efficiently pruned to form a directed acyclic vascular tree without redundant and irrelevant branches based on a flow conservation approach. The experiments on liver CT datasets show that the proposed framework can effectively obtain the liver vascular structure and its quantitative description.