Research About Method Of Kidney MRI Image Segmentation Based On Graph Cut And Level Set

In recent years, the renal transplant has become an option for end-stage renal disease patients because of its higher performance in lifetime and survival quality than dialysis treatment. Utilizing MRI technology could identify various non-surgical causes that contribute to renal graft dysfunction. However, influenced by similar gray level of adjacent organs, partial volume effect and injected contrast, it becomes a very difficult task to segment kidney magnetic resonance images. In addition that different MRI Scanners generate images with different image features, changes of renal shape and position strengthen the difficulties of identifying and segmenting kidney magnetic resonance images.This paper reviews the research background and significance of kidney magnetic resonance image segmentation. The state of art of graph cuts and the level set method in the world has been summarized, and theoretical knowledge of graph cuts is introduced in this dissertation. By studying graph cuts and level set method, a complex model which combines these two methods is proposed. The basic principles of the complex model are shown as follow: to begin with, after applying improved min cut/max flow algorithm to pre-segment the non-concave boundary of the target region, the approximate boundary contour of the region could be achieved. Subsequently, we add the global item and information of adjacent elements into the energy function, regard the contour obtained by graph cut process as the initial contour of the improved variation level set model, then utilize the level set model to cut the concave boundary of the target region. After finite steps of iterations, the final contour converges to the objective boundary quickly and accurately. As a result, the accurate segmentation of the objective kidney magnetic resonance image is achieved. Experimental results testify that the complex model proposed in this paper could not only segment the image with intensity inhomogeneity but also decrease the steps of iterations dramatically. Compared with previous other models, the complex model show better robustness.The method proposed in this paper performs well when segmenting kidney magnetic resonance images with low resolution ratio, low contrast and intensity inhomogeneity. Thus, this method could assist experts to analyze the reinforced characteristics of the tumor correctly, and play a significant role in detection and qualitative analysis of kidney cancer. Findings of this dissertation provide theoretical bases of kidney cancer researches for the related Hospital.