Research And Application On 3D Medical Image Segmentation

As the prevalence of chronic kidney disease is increasing worldwide,it is essential for the early diagnosis of kidney diseases and timely effective treatment.The related clinical practice and research have shown that most of kidney diseased are associated with the increase or decrease of kidney volume and the change of kidney surface in some degree.The automatic kidney segmentation is a challenging issue in medical image segmentation,because there are a lot of interference factors during the kidney segmentation process,such as the unclear boundary between the kidney and other adjacent organs,imaging artifacts,image noise and the pathological deformation of kidneys,and so on.Besides,kidney segmentation process should be without human intervention.Based on the widely comparison of various image segmentation methods,this paper proposed an automatic kidney segmentation algorithm combining the Exemplar-SVM(E-SVM)and level set method,which can detect kidneys from medical image slides and segment the kidney cortex using the location information.First,we employ E-SVM to locate the kidney.Because HOG descriptors are able to represent its characteristic of geometry and topology,it is easy to distinguish the kidneys from other similar targets,and ensure the robustness of the following segmentation of kidney cortex.Second,the level set method can be utilized to partition the cortex part accurately.The experimental results show that our proposed algorithm can realize the accurate kidney segmentation based on CT and MIR enhanced images,and for the unenhanced images,it outperforms other mainstreaming methods.