抑郁症脑网络异常拓扑属性分类研究

发布时间：2018-03-28 10:26

本文选题：脑网络　切入点：功能磁共振成像　出处：《太原理工大学》2013年硕士论文

【摘要】：抑郁症是在生理、心理以及社会环境等因素影响下,人的大脑功能失调,导致认知、情感、意志等精神活动出现不同程度障碍的疾病。目前抑郁症的诊断,多是根据病人所表现的外在症状来诊断病人是否患病。脑网络作为一种广泛应用于识别异常拓扑属性的工具,为抑郁症的诊断提供了一个新的视角。然而,静息状态脑功能网络的拓扑属性能否应用于抑郁症的分类仍然未知。进一步的研究需要找到最合适的特征选择方法,建立分类模型,为抑郁症患者的计算机辅助诊断提供支持。主要工作如下：本次研究中,我们采集了38例未尝试任何药物的首发抑郁症患者和28例正常对照组的静息状态功能磁共振成像数据,构建静息状态脑功能网络。利用图论的方法计算脑网络的聚合系数,最短路径长度,节点的度,节点中间中心度和节点效率,把这些典型的脑网络拓扑属性作为分类特征,使用特征选择的方法对其进行筛选。用五种不同的分类算法构建分类器,分类器包括支持向量机,神经网络和决策树等。用分类特征在统计意义上的重要性作为阈值来划分特征,并对包含不等特征数量的分类器的性能进行评估。实验得到在28个特征(P0.05)下,基于径向基函数的支持向量机算法和神经网络算法得到最高平均预测率(分别为84.36%和80.70%)。结果表明抑郁症和异常脑功能网络拓扑属性有关,并且其统计意义成功的用于分类算法中的特征选择中。建立的分类模型对抑郁症的诊断具有一定的参考价值。
[Abstract]:Depression is a kind of disease which is affected by physiological, psychological and social environment, which leads to different degrees of disorder in mental activities such as cognition, emotion, will and so on. At present, the diagnosis of depression is based on the external symptoms of the patients. The brain network is widely used as a tool to identify abnormal topological properties. This provides a new perspective for the diagnosis of depression. However, whether the topological attributes of resting brain function network can be applied to the classification of depression is still unknown. Establish classification model to provide support for computer-aided diagnosis of depression patients. The main work is as follows:. In this study, we collected resting functional magnetic resonance imaging data from 38 first-episode depression patients and 28 normal controls to construct a resting brain functional network. In this paper, the aggregation coefficient, the shortest path length, the node degree, the center degree and the node efficiency of the brain network are calculated by using the graph theory method. These typical topological attributes of the brain network are taken as the classification features. The method of feature selection is used to screen it. Five different classification algorithms are used to construct the classifier, which includes support vector machine, neural network and decision tree. The performance of classifier with unequal number of features is evaluated. The maximum average predictive rate (84.36% and 80.70%) of the support vector machine (SVM) algorithm based on radial basis function (RBF) and neural network algorithm was obtained under 28 features (P0.05). The results show that depression is related to the topological properties of abnormal brain functional networks. And its statistical significance is successfully used in feature selection in classification algorithm. The established classification model has a certain reference value for the 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;TP311.13

【参考文献】