导　　师： 吕庆文

学科专业： 081203

授予学位： 硕士

作　　者： ;

机构地区： 南方医科大学

摘　　要： 弥散张量成像/(Diffusion Tensor-Magnetic Resonance Imaging,DT-MRI/)是近年来提出的一种新的磁共振成像/(MRI/)方法。分子的弥散运动即是分子的随机位移,但其位移非常小以致在常规MR成像技术中无法显示。弥散成像技术最早出现于20世纪80年代中期,通过双极磁场梯度脉冲对水分子的弥散运动效应进行编码以得到弥散加权磁共振/(Diffusion Weighted-Magnetic ResonanceImaging,DW-MRI/)图像。随后出现了弥散张量成像/(DT-MRI/),它根据不同分子弥散特性的不同,利用多个方向/(至少6个方向/)的弥散加权图像计算出每一个体素的弥散张量数据。利用DT-MRI可以提取分子弥散的各向异性特征,使得全面充分研究活体组织微细结构成为可能。DT-MRI最先成功应用于颅脑的神经纤维束追踪,它也在中风、多发性硬化症、精神分裂症等神经疾病的诊断上发挥着重要作用。因为弥散数据包含了被扫描颅脑组织内部结构固有的物理信息,DT-MRI能够提供有关颅脑组织结构和几何构形的独特数据,这是我们目前可以观察和研究活体脑白质微细结构的唯一途径,而常规MRI无法实现这一点,随之不久前出现了基于DT-MRI的图像分割。 目前DT-MRI的图像分割大多是应用纤维束追踪算法结合主弥散方向数据方法以实现分割,而这类方法在遇到纤维束交叉点时可能出现被邻近纤维误导致分割错误的情形。本文发现,基于马尔可夫场/(MRF/)模型的DT-MRI图像分割算法能够充分利用图像的空间相关信息,能够实现对低信噪比DT-MRI图像进行分割。 因为弥散张量场的特征值和特征向量具有旋转不变性,因此大多数研究者使用这类数据用于DT-MRI图像的分割。然而这类数据仅包含了体素坐标和主方向上弥散的物理信息,且对噪声敏感,稳定性差。因此本文选用弥散张量矩阵代替上述数据作为DT-MRI待分割图像的数据集。 本� Diffusion Tensor /(DT/) Magnetic Resonance Imaging /(MRI/) is a recent MR imaging modality. Diffusion refers to the random translational motion of molecules which is too trivial to be detected by conventional MRI techniques. The basic principles of diffusion MRI were introduced in the mid-1980s, when the the diffusion weighted images /(DW-MRI/) were acquired encoding molecular diffusion effects in the NMR signal via bipolar magnetic field gradient pulses. Thereafter emerged diffusion tensor imaging /(DT-MRI/), which utilize the diffusion tensor of each voxel calculated from DW-MRI images of multiple directions /(at least 6 directions/) according to various characteristics of molecules' diffusion. With DT-MRI, diffusion anisotropy effects can be fully extracts, characterized, and exploited, providing even more exquisite details on tissue microstructure. The most advanced application is certainly that of fiber tracking in the brain. DT-MRI has also played a large part in the diagnosis of a variety of diseases including stroke, multiple sclerosis, schizophrenia, dyslexia, and etc. Since diffusion data contains intrinsic physical information about the internal structure of the tissue being scanned, DT-MRI can provide unique clues to the structure and geometric organization of tissues beyond the bounds of conventional MRI, which is actually the only means to observe and research the micro-structure of brain white matter and gray matter. So emerged tissue segmentation based on DT-MRI not long ago. Research of the DT-MRI segmentation mostly focused on the fiber tracking algorithms based on the principal diffusion direction of diffusion, which have been proven suboptimal for the reason that the fiber tracking algorithm failed in tending to be misled by other fiber bundles in close proximity to the one being traced at nerve fiber crossings. In this paper, the Markov random field /(MRF/) model was proposed for the segmentation of DT-MRI images, which takes into account a priori knowledge of the spatial correlations o

关 键 词： 弥散张量 磁共振成像 可视化 图像分割 马尔可夫场

领　　域： [自动化与计算机技术] [自动化与计算机技术]