功能脑网络的构建分析及在抑郁症中应用研究

发布时间：2018-03-31 10:02

本文选题：重度抑郁症　切入点：复杂网络　出处：《太原理工大学》2013年硕士论文

【摘要】：在竞争压力巨大的当今社会,抑郁症发病率居高不下。而目前在医学上还没有检查或者化验可以作为抑郁症诊断的依据,对抑郁症的诊断还是以临床诊断为主,这种诊断方式很大程度上是依赖于医师的临床经验的,诊断准确率偏低,因此,如何提高精神脑疾病诊断的准确率是目前医师所关心的一个重要问题。目前国内外越来越多的研究人员将复杂网络理论应用到脑认知研究中,将一个整体的大脑系统看成一个复杂网络,并利用越来越先进的技术手段采集人脑数据,利用所采集的大脑数据来构建功能脑网络,运用复杂网络理论对功能脑网络进行分析研究,取得了令人兴奋的实验结果。在以往的研究中已经证明:无论是抑郁症患者还是正常人的功能脑网络都明显具有小世界属性并且抑郁症患者与正常人脑网络的拓扑结构是有些差异的。因此,我们就可以使用复杂网络理论对抑郁症患者和正常人的功能脑网络的拓扑结构进行研究。本文首先利用抑郁症患者与正常人的功能核磁共振数据在一个连续的阈值空间内构建了所有被试的功能脑网络;再运用复杂网络基础理论知识分析了所有被试功能脑网络的网络属性,例如特征路径长度、聚合系数,度等;然后,对正常组和抑郁组功能脑网络的全局属性和局部属性做统计检验,找出正常组和抑郁组功能脑网络差异最为显著的网络属性,以提取出可以从不同维度来表示抑郁症患者功能脑网络的属性值;采用差异最为显著的网络属性的不同组合作为分类特征,利用支持向量机分类算法对所有被试进行分类,再设定阈值,根据统计显著性提取不同数量的节点属性与全局属性组合作为分类特征,利用决策树、贝叶斯、支持向量机、人工神经网络四种分类算法,对抑郁症患者和正常人进行分类研究。通过分类研究,结果发现:若利用网络属性的不同组合作为分类特征,将全局属性和局部属性组合作为分类特征所得到的分类准确率最高。如果使用不同数量的网络属性作为分类特征,运用不同的分类算法,支持向量机和神经网络算法在阈值P为0.05下,所建的模型的分类模型的准确率较高,分别达82.78%及81.36%,本文的研究对于抑郁症的辅助诊断具有一定的应用价值。
[Abstract]:In today's society, where the competitive pressure is enormous, the incidence of depression remains high. At present, there is no medical examination or laboratory test that can be used as the basis for the diagnosis of depression. The diagnosis of depression is mainly based on clinical diagnosis. This diagnostic method depends largely on the physician's clinical experience, and the diagnostic accuracy is low, so, How to improve the diagnostic accuracy of mental brain diseases is an important issue that doctors are concerned about. At present, more and more researchers at home and abroad apply complex network theory to brain cognitive research. Think of a whole brain system as a complex network, and use more and more advanced technology to collect human brain data, using the collected brain data to build a functional brain network. Using the complex network theory to analyze the functional brain network, In previous studies, it has been shown that the functional brain networks of both depression and normal people have obvious small-world properties, and that depression patients and normal human brain networks extend. The structure of the flutter is somewhat different. We can use complex network theory to study the topological structure of functional brain network in patients with depression and normal people. In this paper, we first use the functional nuclear magnetic resonance data of depression patients and normal people at a continuous threshold. The functional brain network of all the subjects was constructed in the value space. Then using the basic theory of complex network, we analyze the network properties of all the functional brain networks, such as characteristic path length, aggregation coefficient, degree and so on. The global and local attributes of functional brain network in normal group and depression group were statistically tested to find out the most significant difference between normal group and depression group. In order to extract the attribute values that can represent the functional brain network of depression patients from different dimensions; using different combinations of the most significant network attributes as classification features, support vector machine classification algorithm was used to classify all subjects. Then set the threshold, extract different numbers of node attributes and global attributes as classification features according to statistical significance, and use four classification algorithms: decision tree, Bayesian, support vector machine and artificial neural network. Classification of depression patients and normal people. Through the classification study, it is found that: if different combinations of network attributes are used as classification features, The combination of global attributes and local attributes is the most accurate classification feature. If different network attributes are used as classification features, different classification algorithms are used. Support vector machine (SVM) and neural network (NN) algorithm have a high accuracy of 82.78% and 81.36%, respectively, when the threshold P is 0.05. The research in this paper has a certain application value for the assistant diagnosis of depression.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：R749.4;O157.5

【相似文献】