Research On The Segmentation Methods And The Implementation Of The Brain MRI Image

Brain MRI image segmentation is the foundation and the key to normal tissues and pathological organization for3D reconstruction, quantitative analysis and also is the bottleneck of clinical applications. The accuracy of the segmentation is very important for the doctor in the orientation of the lesions organization, the measurement of the shape and size, the judgement of the real disease and making the right diagnosis plan.In this paper, according to the characteristics of brain MRI image, with Matlab for development tools, we look for the ideal algorithm applicable to the segmentation of the brain MRI image. In the former research, the region growing and the level set method is commonly used in the brain MRI image. According to the basic principles of the methods of the region growing and the level set, we design programme s to achieve the segmentation of brain MRI image. Through the analysis of experimental results, we point out that the existing defects of the two methods. Namely, the region growing method severely depends on the setting of the threshold and the selection of the seed, which requires the operator has some prior knowledge; The level set method is not only closely related to the initial parameters and the position of the initial contour, but also for the deformity object, especially for those who have sunken object boundary, not able to convergence to the edge of the actual object well. In addition, for the targets having fuzzy edges, it is likely to continue expanding (or contract), and causes the leakage of the edge.Because of these shortages, we use Ostu which also is the classic image segmentation algorithm. This method is based on the gray histogram of the image, and is most ideal to the image whose target and background has great contrast. Because MRI (magnetic resonance imaging) system inevitablely contains noise, moreover, in this topic, we use low magnetic field equipment to acquire brain MRI images. These make the definition and exquisite of images not very ideal and the twin peaks of gray histogram not obvious. So we improved the Ostu method. Firstly we have two artificial interactions, which makes the twin peaks obvious. Namely, the target and background present a greater contrast. On this basis, then, we apply the method of Ostu to segment image. Compared with the traditional methods (that is, region growing and the level set method), this method is not only simple, and whether in segmentation quality or running time, it has obvious advantages.And we use this method for the segmentation of a series of brain MRI images, realizing three-dimensional(3D) reconstruction of part of the brain, and further proving the feasibility of this method.