Using Visibility Graph to Analyze Brain Connectivity

Recognition of mental activities in brain-computer interface systems based on motor imagery has attracted the attention of many researchers. A visibility graph is a powerful method for analyzing the function and connectivity of different areas of the brain. The aim of this study is to improve and develop the visibility graph method for analyzing brain behavior and detecting motor imagery.

First, brain signals including four motor imagery classes of left-handed, right-handed, foot, and tongue were transformed into three types of visibility graphs, and important features of these graphs were extracted. Then, to reduce features, the method of analysis of variance was used. To classify the motor imagery classes, the support vector machine was used. In most investigations, graph degree distribution has been used to extract information and graph weighting. In the present study, amplitude difference distribution has been used so shorter time series are required. To analyze the function and connectivity of different areas of the brain and to obtain the direction of information flow, a new method called weighted horizontal visibility graph-transfer entropy has been proposed.

Increasing the kappa value compared to other studies showed that a weighted horizontal visibility graph is a suitable method for processing brain signals based on motor imagery. A comparison of brain graphs and the direction of information flow in the four classes of motor imagery showed a significant difference between them.

Temporal networks provide a better understanding of brain dynamics in brain-computer interface systems based on motor imagery.