Abstract
In this thesis, we present a system for liver segmentation and vessel structure extraction from 3D Computer Tomography (CT) images with the corresponding 3D model reconstruction. For clinical application, we want the system can be controlled or modified by the radiologists or doctors, so we purpose a semi-automatic method to segment the liver area which is based on a region growing technique combined with 3D random walker algorithm for interactive segmentation refinement. The experimental results show higher accuracy in the segmentation results by using the proposed algorithm. After obtaining the liver segmentation result, we use a modified tubular filter to extract the vessel structure which is more robust than the original one. Using some special characteristics of the human vessel structure, we achieve more accurate vessel extraction for 3D model reconstruction.