静息态功能脑网络核心节点评价方法及其在抑郁症分类上的应用

摘要

识别大脑功能网络的核心节点对于脑科学与脑疾病研究有重要的指导意义.目前,研究者普遍运用度中心性和k-core 分解法来度量网络的核心节点,然而度中心性只考虑节点自身的邻居个数而忽略了其在网络中的位置.k-core 分解法只考虑节点在网络中的位置而忽略了其自身的特性.本文综合考虑节点的度值及其在网络中的位置,提出了一种基于度值和节点位置相结合的核心节点评价方法.对正常被试大脑功能网络进行蓄意攻击仿真实验表明:与度中心性和 k-core 分解法相比,对采用新方法识别出的核心节点进行蓄意攻击后,网络的全局效率下降幅度最大;其次,依据文中提出的中心性指标,找到抑郁症患者和正常被试之间具有显著差异的脑区,并将这些脑区的中心性指标作为分类特征进行分类,使得分类的准确率提高了7%.%Identifying the hub of the brain function network has important guiding significance for the research of brain science and brain diseases.At present,degree centrality and k-core de-composition method are used to measure the hub of the network.However,degree centrality can only take into account the number of neighbors of the node,regardless of its location in the net-work,while k-core decomposition only measures the position of the nodes in the network and neglects its characteristic.In this paper,we proposed a method of evaluating the hub based on the degree value and node location by combining the degree of node and its position in the net-work.Through malicious attacking the hub nodes of brain network,the results show that the network is most seriously damaged when compared with degree centrality and k-core decomposi-tion method.Then,the hub was used as classification feature to identify major depressive disor-der patients from normal controls.The classification accuracy was improved by 7%.