EEG network research based on graph spectrum analysis
                                                              XU Shanzhi, HU Hai, JI Linhong, WANG Peng
(1.Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China;2.Department of Precision Instrument, Tsinghua University, Beijing 100084, China)

Abstract: In order to solve the problem of functional brain connectivity analysis, the approach in the graph spectrum domain was applied in the electroencephalograph (EEG) network research. The rhythm extraction, the weight matrix and Laplacian matrix of the correlations between different channels were studied. Then the decomposed elementary matrixes of Laplacian matrix were further analyzed, in consistence with the EEG network topology structures of different orders by binary processing. The method of EEG network research based on graph spectrum analysis was proposed. The normal and epileptic EEG signal was verified by the proposed method. The experimental result shows that the proposed method can extract the EEG network topology structures of different variability patterns under both conditions. At the same time, the proposed method can realize the classification of normal and epileptic regions effectively. The area under the receiver operating curve is 0.834, which performs better than the extensively used methods.

Key words: electroencephalograph (EEG) network; topology structure; singular spectrum analysis（SSA）; graph spectrum analysis