The Segmentation Of MR Image Based On DICOM Format

The research of segmentation has been an important field in medical image analysis. The advancement of the segmentation played active development of clinical medical study. For example, image segmentation can get the size of the focal region. Then the doctor use the information to get the patient's condition in detail. At the same time, with the ability of computer, image segmentation has active application in computer assisted surgery. So the development and prospect of this skill is wide.At first, this thesis introduces the basis principle and method. At the same time, this thesis indicates the meaning and condition of segmentation. Because of DCM format, whether the image can be accurately displayed or not is the important premise of image segmentation. The research must analyze the standard of DICOM then display the medical image by VC++ or VC++ with DCMTK. In this thesis, the object of study is MR brain image. In the process of analyzing these professional medical images, grasping and understanding the principle of image is the first step. Next, using the knowledge can acquire the appropriate image to analyze the lesions. Meanwhile, the object of MR brain images is the part of brain that removing skin and bones. It can be acquired by mathematics morphology. Then it can be prepared for segmentation. In this thesis, for MR brain image, image segmentation has two different applications. Segmentting MR brain images into different tissue classes is the first one, such as gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF). Extracting the focal region of interesting (ROI) from other tissues is the second one.This can assist physicians to make right diagnosis. So this thesis promotes different algorithms to solve these applications.Fuzzy C-mean (FCM) clustering algorithm can be used in image segmentation to solve the first application. Through using, fuzzy membership matrix and clustering center, the distance that between the present pixel and clustering center can be acquired to achieve the aim. But the number of iterations is big. In order to solve this problem, the thesis promotes the Euclidean distance to replace the Gaussian distance 基于DICOM文件格式的MM图像方法研究that between the present pixel and clustering center to improve the ability that reduce the number of iterations. Then the algorithm only takes account to the present pixel but ignore the spatial information. In order to solve this problem, the thesis gives that restrict the present Fuzzy membership matrix to update the membership matrix of the pixel. This method makes the result perfect.For the second application, the thesis summarizes relational theory about evolving curve and analyzes the principle of Shakes model and C-V model. In this thesis, these images that contain lesions are the sequence of infarction. At first, the study give explanation of the pathology. Then the thesis gives the process of derivation and analyzes the application of extracting the focal region by Level Set method. Through getting the number of pixels in the focal region and reading tags to get the thickness to computer the volume of the region. Then according to above method, the lesions of every image in the sequence can be computed. Then, the thesis can get the volume of the focal region. This method is easier than 3D reconstruction.