Research On Human Brain DTI Image Segmentation Algorithm Based On K - Medoids Clustering And Its Fiber Tracking

Diffusion tensor imaging(DTI) is the only imaging modality that can non-invasively get the anisotropic diffusion characteristics of the water molecules in biological tissue. In addition, it can provide more subtle structure information. In DTI, voxels of the same tissue area contain the same diffusion characteristic information and voxels of different tissue area contain different diffusion characteristic information. Based on the difference of the characteristics, a certain DTI image segmentation method is used to split out the lesion site of biological tissue to assist the early diagnosis of clinical diseases and subsequent treatment such as primary alzheimer’s disease, cerebral ischemia, stroke and schizophrenia.DTI image segmentation algorithms described above generally have shortcomings of high computational complexity, low accuracy and cannot effectively extract the fine structure information. The studies are done in this paper as following:1) To solve the problem of losing detailed edge information inherent in traditional DTI image segmentation methods, some new morphological gradient tensor parameters are proposed. Firstly, new tensor similarity morphological gradients are defined based on the eight neighborhoods; Secondly, new tensor morphological gradient anisotropies are put forward; At last, the tag based watershed algorithm is applied in DTI image segmentation, that is, morphology techniques are adopted to operate opening and closing operations on morphological gradient images to accomplish segmentation and achieve a more satisfactory segmentation results.2) In order to resolve the problems of sensitivity to noise, inefficiency, low accuracy and over-segmentation existing in traditional watershed segmentation algorithm, a new DTI brain image segmentation algorithm based on k-medoids clustering is proposed. Firstly, the human brain DTI image morphological gradient diffusion anisotropy parameters are calculated to transform the diffusion tensor data into scalar maps; then the maximum and minimum distance algorithm is used to choose the initial cluster medoids, while k-medoids clustering algorithm is used to complete morphological gradient image preprocessing; and finally watershed segmentation algorithm is applied to finish the more detailed and more accurate segmentation of target objects.3) By measuring the diffusion tensors of water molecules in brain, DTI technique can well tract the base structure of the nerve fibers. To study and analyze the target segmentation area more clearly, Toolkit Diffusion and Track Vis software are used in this paper to carry out fiber tracking and visualization of the experimental data, and to provide more accurate information for the treatment of the diseases.Diffusion tensor imaging is a non-invasive imaging technique developed on the basis of Resonance Imaging Magnetic(MRI), which is widely used in medical clinic. Therefore, the research on DTI image processing, analysis and fiber tracking is helpful to the further understanding of the human brain’s internal structure and working mode, and to solve the mystery of human wisdom.