Research On Segmentation And Visualization Of Liver And Intrahepatic Vessels

With continuous development of computer-aided diagnosis techniques, the medical image post-processing workstation has already became one of indispensable technical methods toward routine physical examination, pre-operative assessment and intraoperative assistance of liver diseases. For complexity and variability of liver organs, basic visualization and analysis tools in the workstation are difficult to meet the demands of diagnosis and surgery. Hence the virtual liver surgery planning system based on image segmentation and 3D visualization techniques has attracted considerable attention recently; however, for complicated clinical applications, there are still many challenges in its accuracy, stability, efficiency and interactive simplicity of segmentation, which are required further studies to improve.The system mentioned above consists of several procedures such as liver segmentation, tumor segmentation, vessel segmentation, vessel analysis and classification, liver segments estimation, target visualization, virtual surgery and parameters calculation. Among them, the segmentation and visualization of liver and intrahepatic vessels are most crucial parts and are even the basis of others.This paper focuses on research of the problems of liver segmentation and intrahepatic vessels segmentation. At the beginning, based on the study of two kinds of classical liver segmentation algorithms, an improved method has been proposed. This method employs 3D region growing process as basis with both Gaussian nonlinear coupling function and Canny edge detection as constrains, and combines some image post-processing algorithms such as holes filling, largest connected domain judging and segmentation contour correcting, thus the proposed approach can improve robustness of liver segmentation and attain a balance between the system stability and the process efficiency. Once the body of liver is acquired, the segmentation methods for intrahepatic veins inside liver then are researched at the following stage. This process composes of several steps such as pre-processing, multi-scale vessels enhancement based on Hessian matrix eigenvalues, optimal threshold vessel segmentation, manual interaction and post-processing. As a result, the portal and hepatic veins can be extracted. In addition, some further researches about the skeleton extraction and the graph theory modeling of vessels have been done, which can transfer vascular images to Vertex-Edge graphs and acquire topological structure of vessel branches to provide structure models for the next stage:division of portal and hepatic veins. Finally, based on the volume rendering method involving OpenGL texture mapping and GLSL fragment shader, liver and intrahepatic vessels can be differentiated by means of multiple texturing techniques. And then, the cut tools of either arbitrary planar or curved planar have been developed to implement 3D visualization and interaction of the segmented objects.The above procedures are accomplished on VC++ platform and are measured by the clinical data. The results are shown as follows:comparing with the classical algorithms, the proposed method of liver segmentation outperforms in segmentation quality and simple interaction in a moderate execution time; the segmentation method of intrahepatic vessels can obtain the main vessels of portal and hepatic veins; the procedures of either vascular skeleton extraction or graph theory modeling work on the main branches of portal vein successfully; 3D visualization with the gray-opacity transfer function and the interactive tools can display the spatial and internal information of liver and intrahepatic vessels intuitively, which will be helpful for the virtual liver surgery.