The Application Of Active Contour Model In MRI Segmentation

Cardio Vascular Disease(CVD) is one of the most dangerous disease of the time. Qualitative and quantitative analysis of heart movements by Cardiac Imaging has a great help on CVD diagnosis. In all kinds of Cardiac Imaging technique, we pay much attention on Magnetic Resonance Imaging (MRI) with tag linear because it has many advantages such as large deal of imaging parameter, no invasion and so on. Precise segmentation on heart MRI can provide a great deal of information for further analysis of heart movement which is instrumental in research on heart anatomical structure, quantitative analysis of heart disease pathologic and categorical, and making efficient diagnosis. On the while, two-dimensional sequence segmentation on heart MRI is the essential step in building model of heart movements in three-dimensional. As its importance, there are large deals of research concerned on the segmentation on heart MRI both here and abroad.Segmentation based on variational method and the optimization of variational functional has been applied into complex image segmentation successfully. Based on the problem of medical image segmentation, this paper introduced the theory of vatiational method, and efficient numerical algorithms are discussed.Active Contour Model is based on the theory of variational function and variational method., it has great effecting in much area of image segmentation. Active Contour Models is also very important tool in medical image segmentation. Many people have researched on the application of Active Contour Model onto cardiac MRI segmentation and heart reactivity. In general, Active Contour Models include parametric active contour model and geometric active model. Kass proposed parametric active contour model. It detect image edge by minimize the energy function of spline near the region of interest (ROI). Parametric model has successfully applied in medical image segmentation, but it easy to fall into local minima and can not deal with topological changes. Many people tried to improve parametric model. This paper introduced some typical work about it.Aiming at the disability of parametric active contour models, people proposed Level Set Methed. In Level Set method, the active contour can deal with topological changes naturally. Classical Level Set method only used image gradient information in cardiac MRI segmentation. In can hardly get nice result because of strong structural noise, image weak edge and tag line. To solve these problems, this paper integrates region information of image with gradient information, and constructs a new velocity function based on K-means clustering. This new velocity function has better antinoise capability and can deal with images which have strong noises, weak edges and low contrast. The experiments on the tagged left ventricle Magnetic Resonance Images showed the effective of this method.Nowadays, almost all active contour models need human initialization. People must initialize a origin curve near ROI. Especially in parametric active contour models, theorigin curve strongly need to be put near ROI, and this curve can not include any noise. This paper proposes a new automatic segmentation framework of deformable models for the segmentation of left ventricle in cardiac MRI. At first improves proposed a multi-indexes Fuzzy C-means clustering (FCM), then detected the region of left ventricle in cardiac MRI automatically and constructed a original curve based on the improved FCM, at last improved the classic velocity of Level Set method. In this framework, the whole segmentation process is automatic. The experiments on the tagged left ventricle Magnetic Resonance Images showed the effective and robust of this method, and easy to implement.Parametric method and geometric active contour models both are local method, when they are used in complex medical image segmentation, they can not avoid same limitation. Recently, people pay more and more attention on Mumford-shah variational function model for image segmentation and recover. Mumford-Shah model has much advantage such as global optimization, and can get image edge and recovered image in same time. This paper also introduce Mumford-Shah model for medical image segmentation.