Research On Segmentation Methods Of Medical Ultrasound Image Based On Level Set

At present, ultrasonic imaging has wide applications in clinical diagnosis and treatment mainly for its real-time, security and repeatability, high sensitivity and low cost. Therefore, to efficiently and accurately segment medical ultrasound images has important significance for promoting the application of ultrasonic imaging technologyBased on the situation of medical ultrasound images segmentation, this paper has made a research on the medical ultrasound image segmentation method, and the main work is summarized as the followings:(1) This paper makes a systematic analysis on the application of image segmentation, detailed introduction on the characteristics of the medical ultrasound images, medical ultrasound image segmentation difficulties and existing segmentation methods, basing on which the importance and realistic significance of this research is proposed. It has introduced some related algorithm theory, including the level set method, mathematical morphology, C-V model algorithm, kernel fuzzy c-means clustering algorithm. Through the analysis of the principles and characteristics of these algorithms, we propose that the main research direction is to improve the speed function of the level set curve evolution, all of which is well prepared for the statement of the following improved algorithms.(2) In this paper, we propose a method of combining the morphological algorithm with the level set for medical ultrasound image segmentation. The morphology gradient field can be used as a speed function of the level set curve evolution to get a segmentation image. The simulation results show that the method can get the more ideal smooth boundaries, when it is used for complex medical ultrasound images segmentation, in which the target area gray is uneven caused by speckle noise.(3)In order to improve the speed of the traditional level set image segmentation method, in this paper, we propose a level set method of medical ultrasound image segmentation based on the improved Chan-Vese (C-V) model. We have also analyzed the selection of initial closed curve in the process of the level set evolution. Through the simulation, it works out the reasonable initial closed curve can speed up image segmentation. The improved method is more effective with fewer time cost than the traditional Chan-Vese model segmentation. (4) The influence of the parameter selection of kernel fuzzy C-means clustering (KFCM) on medical ultrasound image segmentation result is studied. The kernel fuzzy clustering results are combined with the level set curve evolution to get a rough boundary, and then the gradient vector flow is used for the curve evolution to get a refinement boundary. The experiment proves that this method can effectively overcome the difficulties of medical ultrasonic image segmentation, and getting a truer image boundary-...