Research On Liver CT Images Segmentation Method Based On Graph Cuts Theory

With the widely use of new medical imaging methods like CT and so on. At present, medical Image Process and Analysis has become one of the fastest development and the most significant achievements in medical technology. Now liver CT scan is the necessary method for the preoperative imaging examination of liver cancer. How to analyze the gat imaging data has become a crucial problem. Therefore segmentation technology is the key technology in subsequent image processing and analysis.Firstly, this paper introduces the purpose and significance, the research background and the present situation of liver CT image segmentation in brief. And then having a review of graph cuts theory that applying to image segmentation. Some liver image segmentation methods are introduced briefly in this paper. And then choosing graph cuts theory to segment liver.Secondly, the basic theory and the basic framework to accomplish image segmentation of traditional graph cuts are introduced. In consideration that traditional graph cuts needs construct a lot of graph nodes, so adding the concept of superpixel into graph cuts theory. Using superpixel method can reduce the computational cost effectively and strengthen local consistency. Lazy snapping is a typical application which is based on superpixel. Lazy snapping uses watershed transform to get small regions, the small regions is used to instead original image pixel. But the region consistency is not good enough by the pre-segmentation of watershed in the image. And watershed will also cause serious over-segmentation phenomena. So through the compare and analysis of various superpixel segmentation methods, we choose apply SLIC (simple liner iteration cluster) to graph cuts. We put forward graph cuts based on SLIC to implement the liver CT image segmentation quickly and accurately. Through the compare and analysis of experiments, the result is more accurate and the computation time more short. The effectiveness and accuracy are proved through the experimental comparison and differences evaluation.Then, through analyzing the segmentation results and considering the boundaries of different organizations are fuzzy in liver CT image. We put up with a two-phase approach. The first phase provides a quick and rough segmentation result using graph cuts. Then the second phase refines the segmentation boundaries using matting. Matting has a better edge thinning effect for fuzzy boundaries. Through comparing, we can draw a conclusion that the segmentation results of liver are more accurately.At last, it sums up the innovations, the shortage and the work in the future.