基于多变量格兰杰因果关系的运动想象因效网络构建

运动想象神经活动规律的探索为脑损伤肢体瘫痪患者康复训练新方法研发等提供理论指导。基于格兰杰因果关系构建的因效网络是分析运动想象神经活动状态的重要工具，但是格兰杰因果关系只能反映两个变量之间的相互作用，而一个简单的运动想象过程也需要多个神经节点参与，针对该问题，本研究引入可反映一个集群中多个变量之间相互作用的多变量格兰杰因果分析，优化运动想象因效网络构建方法。针对4位受试者，利用多变量及传统格兰杰因果关系，分别构建同一人两种不同运动想象模式的因效网络，并提取网络特征进行运动想象模式分类。结果表明，基于多变量格兰杰因效网络进行4位受试者运动想象模式分类的正确率分别为90.4%、88.8%、91.1%、90.3%，基于格兰杰因效网络的正确率为88.5%、89.3%、90.2%、89.7%。与传统格兰杰因果关系相比，基于多变量格兰杰因果关系构建因效网络，能更准确地反映运动想象神经活动特征状态。

Construction of motion imagination causal network based on multivariable Granger causality

Abstract: The exploration of the rules of neural activity in motion imagination can provide theoretical guidance for the development of new methods of rehabilitation training for patients with cerebral injuries and acroparalysis. The causal network constructed based on Granger causality is an important tool for analyzing the state of neural activity in motion imagination. However, Granger causality can only reflect the interaction between two variables, and a simple motion imagination process requires multiple neural nodes to participate. To solve this problem, a multivariable Granger causal analysis that can reflect the interaction among multiple variables in a cluster is introduced to optimize the construction method of motion imagination causal network. For the 4 subjects, multivariate and traditional Granger causality relationships are used to construct two different motion imaging patterns of the same subject, and the network characteristics are extracted to classify the motion imaginary patterns. The results show that the accuracy rate of motion imaging pattern classification in 4 subjects is 90.4%, 88.8%, 91.1%, and 90.3% in multivariable Granger causal network, as compared with 88.5%, 89.3%, 90.2%, 89.7% in traditional Granger causal network. Compared with the traditional Granger causality, causal network based on multivariable Granger causality can more accurately reflect the characteristics of neural activity in motion imagination.