Analyzing Brain Connectivity in the Mutual Regulation of Emotion–Movement Using Bidirectional Granger Causality

Abstract

Body language and movement are important media of emotional expression. There is an interactive physiological relationship between emotion and movement. Thus, we hypothesize that the emotional cortex interacts with the motor cortex during the mutual regulation of emotion and movement. And this interaction can be revealed by brain connectivity analysis based on electroencephalogram (EEG) signal processing. We proposed a brain connectivity analysis method: bidirectional long short-term memory Granger causality (bi-LSTM-GC). The theoretical basis of the proposed method was Granger causality estimation using a bidirectional LSTM recurrent neural network (RNN) for solving nonlinear parameters. Then, we compared the accuracy of the bi-LSTM-GC with other unidirectional connectivity methods. The results demonstrated that the information interaction existed among multiple brain regions (EEG 10-20 system) in the paradigm of emotion–movement regulation. The detected directional dependencies in EEG signals were mainly distributed from the frontal to the central region and from the prefrontal to the central–parietal.