Analysis Of EEG-based Effective Network Of Schizophrenia

Presently, schizophrenia(SZ) has been the forth disease, effecting the brain health of adolescents and adults. Investigating the brain mechanisms of normal people and the patient with SZ is of great practical and academic value. In order to study the interactions among the network nodes, increasing number of electroencephalogram(EEG) analysis methods which based on the medical imaging technique have been proposed by the researchers. There are plenty of techniques for EEG signals analysis, such as partial correlation, mutual coherence, cross-correlation etc. However, these methods can only analyze the functional networks, but not suitable for describing the interactions of causality among the different network nodes. Nevertheless, the partial directed coherence(PDC) analysis, which based on the Granger causality(GC), can successfully solve this problem. In this paper, we apply the PDC into the effective networks analysis between the normal people and the patients of positive SZ and negative SZ.Firstly, EEG signals of 32 electrodes among different subjects were collected, at the same time, the biological artifacts were diminished by means of independent component analysis(ICA) employing function runica from EEGLAB. Secondly, selecting an appropriate model order for each subject in order to built the multivariate autoregressive model(MVAR), and extracting the characteristics of all EEG signals among different rhythms(delta 0.5~3 Hz, theta 4~7 Hz, alpha 8~12 Hz, beta 13~30 Hz). Then, exploring the causal relationships between the given paired nodes by use of PDC, and determining a threshold to build the effective brain networks, which includes 32 network nodes. Finally, the analysis of effective brain networks for healthy people, the subjects of positive SZ and negative SZ have been shown in three aspects: the hubs of network nodes, the local and global efficiency and the information flows respectively.Our study shows that: 1. Some lesions of functional connectivity have taken place in the SZ patients’ brain network when compared with the healthy people. 2. The cerebral activities of different subjects will result in different consequence. The hubs of network nodes in the healthy people mainly focus on the frontal lobe, temporal lobe and central lobe of the left hemisphere in the brain. While the positive SZ’s hubs of nodes are in the frontal lobe and temporal lobe of the right hemisphere, and the network hubs nodes of negative SZ mainly focus on the occipital lobe and parietal lobe. 3. The network properties for the same group subjects in different rhythms showed a similar properties.4. The local and global efficiency of normal people showed lower values when compared with the SZ subjects. In conclusion, the study has a potential value in the diagnosis and the treatment of schizophrenia.