Abdominal CT Image Segmentation Based On Adaptive Shape-Constrained Graph Cuts Algorithm

With the development of medical imaging technology,the medical image segmentation has become more and more significant in the field of clinical medicine.The diagnosis of many diseases in the clinic requires doctors to quantitatively analyze the morphological structure of related tissues and organs,and the premise of analyze these tissues and organs is to segment them accurately from the medical images.The traditional segmentation task relies mainly on the manual process by the physician.This segmentation method requires a lot of time and energy,and the accuracy of segmentation depends largely on the knowledge and the experience of the physician.Therefore,the segmentation results sometimes are subjective.In this paper,the adaptive shape-constrained Graph Cuts algorithm is applied to three-dimensional abdominal CT image to automatically segment multiple targets which include liver,kidney and spleen,and overcome the defects of manual segmentation.The traditional Graph Cuts algorithm mainly uses the gray information of the image.However,the abdominal CT image contains more tissues and organs,and the target contour is more blurred,so it is easy to appear over-segmentation and under-segmentation.In order to make up for the deficiency of the traditional Graph Cuts algorithm,we makes full use of the shape prior information of the target organ when segmenting the image,and adds a penalty term to the energy function to constrain the shape of the target to be segmented,thus improving the segmentation accuracy of the image.The research content of this paper mainly includes the following aspects: the target organs in the CT image were segmented by the method based on Multi-atlas registration,and contour of the Initial segmentation was taken as the shape prior of the Graph Cuts algorithm.In the process of segmentation which based on Multi-atlas registration,the probability map of the target organ is obtained by fusing the deformed label image with weight voting algorithm.In order to improve the segmentation accuracy,the sign distance was calculated according to the initial contour,and it is taken as the penalty term to constrained the energy function of traditional Graph Cuts.The weight coefficient of the penalty term was selected by the target probability map adaptively,and the fusion of the shape prior was implemented accurately.In order to verify the effectiveness of the improved segmentation algorithm,we uses the adaptive shape-constrained Graph Cuts algorithm to segment the liver,kidney and spleen from the abdominal CT image(3D-IRCADb-01).The experimental results show that the combination of shape priors can overcome the over-segmentation and under-segmentation problems which caused by the traditional Graph Cuts algorithm.The adaptive shape prior can further improve the segmentation accuracy of the algorithm.When segment the CT images of different patients,the improved algorithm can achieve better segmentation with less fluctuation in segmentation accuracy and better robustness.