| The main purpose of the medical image segmentation is to distinguish between different objects in a given and to label each pixel with its underlying class. Image segmentation is very important to image preprocessing and patter recognition, such as feature quantification, image registration 3D reconstruction and etc. In this thesis, the brain magnetic resonance(MR)image segmentation is researched. There are two purposes for the segmentation Of MR brain images. The first one is to segment MR brain images into different tissue classes, especially gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF), which is crucial to the image registration,3D reconstruction and medical image visualization. The second one is to extract the focal region of interesting (ROI) from other tissues in order to assist physicians in making right diagnosis, and working out the therapeutic strategy.In general an ideal MR image is assumed to be piecewise constant, unfortunately, the property is destroyed by electron and structured noises, intensive inhomogeneity and partial volume effect (PVE), etc. Because of the individual differences in the tissue anatomy and the slow calculating speed and inaccuracy, the current segmentation algorithms fail to satisfy the need of clinical practice. Fuzzy C-means (FCM) algorithm has been widely used in brain MR image segmentation. However, the conventional FCM algorithm is used for image segmentation, no spatial information is taken into account, leading to the FCM algorithm get the unexpected results of segmentation when dealing with the images corrupted by noises. In the thesis, an improved model of FCM algorithm is proposed. The performance of the algorithm is remarkably superior to the conventional ones in terms of accuracy.Medical images behave fuzziness duo to PVE artifacts and the uncertainty in some focal regions. The idea of using membership function associated with fuzzy-set theory to represent partial volume proportions of each "pure" tissue has been a quite popular and widely used model, in which Fuzzy c-means (FCM) clustering algorithm is the well-established approach to the implementation of the image segmentation. However, the conventional FCM fails to incorporate the spatial information of the image leading to aberrant consequences in the case of dealing with low signal-to-noise ratio (SNR) MR images.Multiple sclerosis (MS) is an autoimmune and inflammatory demyelization disease of the central nervous system (CNS). It is a common disease of young adults. In people affected by MS, patches of damage called lesions appear in seemingly random areas of the CNS. Lesions appear in the central nervous system:encephalon, especially the white matter, spinal cord and optic nerves. Usually lesions are due to a demyelinization with a replacement of cerebra-spinal fluid instead of myelin. Fluid attenuated inversion recovery (FLAIR) pulse sequences could suppress the signal from CSF and produce very heavy T2 weighting. The marked reduction in flow artifact from CSF and the high T2 weighting enabled MS anatomical detail to be seen within the brain stem and produced high lesion contrast in areas close to CSF.According to the diagnostic criteria for MS, experienced clinicians can diagnosis disease distribution and brain lesions by FLAIR MR image. In clinical, doctors generally portray boundary of hundreds of human brain slice image by hand according to the imaging features and anatomical properties of MS. Then they conceived lesion and its surrounding tissues in 3D structure and their spatial relations, and use it as the basis for formulation of treatment plans. This method is time-consuming and laborious. Segmentation results depend on doctor's anatomical knowledge and experience, and are difficult to reproduce. Because MS causes a long time, in order to assess the effects of treatment, we need to analysis MR images of the same patient in different periods, which makes the workload very load. So, how to use mathematical method and computers segment MS lesions automatically or semi-automatically has become a very meaningful research.At first, we provide an all overview of image segmentation methods. We describe the wide variety of medical image segmentation methods and applications. The thesis is devoted to general study of medical image segmentation, including the theory, the classification and the method of segmentation.In chapter 3, we present many improved methods of the FCM algorithm in recent years in the lecture. There are generally classed into three kinds:the first one, the constraints on membership function is changed, the second one, the term of spatial information is introduced, the third one, the kernel method is introduced. Finally, the typical ones of these algorithms are analyzed and appraised simply.In chapter 4, an automatic segmentation algorithm of MS lesions for MR FLAIR images is presented based on FCM by using a priori knowledge of characteristics of MS lesions and anatomical structures. The connectedness employs fuzzy relationship to describe the analogy of two neighboring pixels. FCM is a clustering algorithm based on gray level and each pixel is independent from each other in clustering process. It doesn't consider the impact between adjacent pixels and spatial information, so it can't receive an ideal segmentation when image overlaps noise in image segmentation. Considering that the adjacent pixels of the same classification have a similar probability in the same brain tissue, we value the pixels in a 8 neighborhoods as a date set, and filter the date to reduce the impact of noise on the clustering accuracy. The testing results using clinical MR FLAIR brain images demonstrate that the performance of the proposed algorithm is significantly improved over the traditional FCM clustering algorithms. This unsupervised algorithm can be used in clinical practice with adequate calculating speed. |
PDF Full Text Request