Magnetic Resonance Image Segmentation Of Active Contour Model Based On Regional Information

Magnetic resonance imaging (MRI) has the advantages of less harmful, high resolution and seldom affected, which has already become a hot topic in the community of medical images analysis. A large number of clinical diagnoses showed that:with the help of advanced medical imaging technology, qualitative and quantitative analysis of medical images plays an important role in standardizing the diagnosis and treatment of cardiovascular disease and controlling the patients' condition. However, because of medical imaging equipment limitations, it leads to the images which we obtained include of noise, weak boundary and bias fields etc. This makes the traditional medical image segmentation method out of operation. The active contour model based on the variation theory has been better applied in image segmentation technology, which has many advantages, such as the support of rigorous mathematical theory.In this background, the paper studies the active contour model based on variational method under complex scenes, which is sensitive to the noise, less accurate, and so on. This paper mainly researches the following two aspects:(1) A novel couple model of left ventricle MR image segmentation and bias field estimation based on regional information (WLIF) was proposed. Combing of bias field estimation method, and using the structure information of the image to construct non-local weights, this model keeps the slender topology structure of the image better. It reduces the impact of noise and recovers the image bias field at the same time, so as to realize the accurate image segmentation of intensity inhomogeneity.(2) To solve the problems of multi-objective medical image segmentation, this chapter puts WILF model generalized in multiple phase image segmentation, and proposes a multiple phase active contour model for brain MR images segmentation. The multiphase level set function is used to fit local intensities, to ensure the smoothness of the segmentation results well. What's more, the non-local information is introduced to keep the brain MR image details better. The experiment results show that this method can fast and efficiently segment the white matter, gray matter and the CSF.