Level Set Method Based On New Image Force In MR Image Segmentation

Now Cardio Vascular Disease(CVD) becomes fatal illness threating health ofhuman beings, ten million people in the world died of CVD every year. Magneticresonance imaging (MRI) is paid much attention to with many advantages such aslarge deal of imaging parameter, no invasion, high spatial resolution, seldom effectedby the motions of objection and so on. Precise segmentation on heart MR images canprovide a great deal of information for further analysis of heart movement which isinstrumental in research on heart anatomical structure, quantitative analysis of heartdisease pathologic and categorical, and making efficient diagnosis. Meanwhile,two-dimensional sequence segmentation on heart MR images is the essential step inbuilding model of heart movements in three-dimensional.The level set method based on geometric deformable model, which translates theproblem of evolution of 2-D(3-D) close curve(surface) into the evolution of level setfunction in the space with higher dimension, obtains the advantage in handling thetopology changing of the shape. As an accurate and steady algorithm, the level setmethod has a wide application.In this paper, we introduce the level set mothod. The traditional level setmethod, using the local marginal information of the image, can get good result whenit is used to deal with synthesis image or image with great differences of itsbackground and objects. However, it is difficult to obtain a perfect result when it isused for MR images which has strong noises, weak edges and tag linear. To solvethese problems, this paper integrates clustering information of image withinformation of the region of interesting, and constructs a new velocity function basedon this new image force. This new velocity function has better antinoise capabilityand can deal with images which have strong noises, weak edges and low contrast. Inthe new model, the level set curve moves under the function of region and boundaryinformation. The experiments on the tagged left ventricle MR images showed theeffective of this method.