颅脑磁共振图像分割技术研究

发布时间：2018-03-08 09:13

本文选题：MRI　切入点：图像分割　出处：《南昌航空大学》2017年硕士论文　论文类型：学位论文

【摘要】：当今社会,随着科技的发展,经济的增长,人们越来越注重自身的健康问题,颅脑作为人体的中枢神经系统,对脑部疾病的检测和治疗更是重中之重。建立精准的颅脑三维模型有利于脑部外科手术、远程医疗等医疗手段的实施、辅助脑部医疗诊断以及传授脑部医学知识。颅脑核磁共振(MRI)图像的各个组织精准分割是核磁共振(MRI)颅脑图像三维模型重建、定量分析和可视化的基础,且颅脑MRI图像分割可以提取颅脑图像中特殊组织的定量信息,可以解决颅脑医学图像中样本数据大,后续分析工作强度高等问题,这已经成为当今研究热点。但是磁共振脑图像是一种复杂的医学图像,由于脑组织之间互相混叠在一起没有清晰的边界,不同个体之间的差异性较大,再加上在成像过程中磁场的不均匀性、部分容积效应以及噪声的影响造成的图像内在的不确定性,且诸如脑脊液、颅骨和头部空腔在MRI图像中的灰度又非常近似,使分割问题显得更为复杂和困难。为了在无损情况下获得精确的人体头部三维模型,需要从颅脑的MRI序列图像中分割出头皮、颅骨、灰质、白质和脑脊液等五种主要的组织。针对这一问题,本文提出了一种基于FCMVBM相结合的整体分割方法。该方法首先通过BET算法从原始的MRI图像中摘取出颅脑区域,然后通过FCM算法对得到的颅脑区域进行细分得到灰质、白质和脑脊液,继而再通过VBM分割算法分割出颅骨、头皮和脑部空腔等非脑组织区域,最后对分割得到的各个组织进行平滑和形态学处理,最终成功分割出了所需的五种组织。通过与K均值聚类算法及形态学分割方法对比发现本文的分割算法在组织边缘梯度更高的情况下具有更低的形态失真度。
[Abstract]:In today's society, with the development of science and technology and the growth of economy, people pay more and more attention to their own health problems. The brain is the central nervous system of the human body. The detection and treatment of brain diseases is of the utmost importance. The establishment of accurate three-dimensional brain models is conducive to the implementation of medical means such as brain surgery, telemedicine, etc. It is the basis of 3D model reconstruction, quantitative analysis and visualization of MRI MRI images. The segmentation of craniocerebral MRI image can extract the quantitative information of special tissue in craniocerebral image and solve the problems of large sample data and high intensity of follow-up analysis in craniocerebral medical image. But magnetic resonance imaging is a kind of complex medical image, because there is no clear boundary between brain tissue and each other, the difference between different individuals is great. In addition, the inhomogeneity of magnetic field in the imaging process, partial volumetric effect and the effect of noise cause the inherent uncertainty of the image, and the gray levels of the cerebrospinal fluid, skull and head cavity in the MRI image are very similar. In order to obtain accurate three-dimensional model of human head without damage, we need to segment scalp, skull, gray matter from MRI sequence images of brain. There are five main tissues in white matter and cerebrospinal fluid. In order to solve this problem, a global segmentation method based on FCMVBM is proposed in this paper. Firstly, the brain region is extracted from the original MRI image by BET algorithm. Then the gray matter, white matter and cerebrospinal fluid are subdivided into the brain area by FCM algorithm, and then the non-brain tissue areas such as skull, scalp and brain cavity are segmented by VBM segmentation algorithm. Finally, each tissue is smoothed and morphologically processed. By comparing with K-means clustering algorithm and morphological segmentation method, it is found that the proposed segmentation algorithm has lower morphologic distortion when the edge gradient of tissue is higher.
【学位授予单位】：南昌航空大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R445.2;TP391.41

【参考文献】