基于节点相似度的功能脑网络建模方法

Method of brain functional network modeling based on nodes similarity

为比较3类链路预测指标(共同邻居、节点的度以及将二者相结合)中哪类指标更适用于功能脑网络建模,从5种脑网络全局与局部属性的角度分析建模效果,提出一种评价模型网络与真实网络整体相似度的指标"E值"。采用将脑网络结构与功能特性相结合的方法进行建模,结构特性为解剖距离,功能特性为节点相似度。实验结果表明,不同的链路预测指标对网络属性的拟合程度各不相同,从整体拟合度E值来看,共同邻居最好,共同邻居与度相结合次之,度最差。

To explore which of the three kinds of link prediction indexes including common neighbor,degree,common neighbor combined with degree can fit the real brain network better,the modeling results were analyzed from the perspective of five global and local brain network properties.In addition,an E value was proposed to evaluate the overall similarity between the model net-works and the real networks.Brain functional network modeling was based on anatomical distance and node similarity.The re-sults show that,different link prediction indexes can fit the properties differently.In terms of the overall similarity,common neighbor is the best index followed by common neighbor combined with degree,and degree is the worst.