Analysis Of Effective Brain Network Based On Granger Causality

Schizophrenia is a severe mental illness.Some patients will appear to lose initiative and cognitive function,which may contribute to a final recession and mental disability,eventually lead to heavy burden to the family and society.Yet some patients can be cured after treatment.Based on the previous works,this thesis analyzes the effective connection of brain network by Granger causality,studies the difference of the network attributes between normal controls,schizophrenic patients and their uneffected relatives,and looks for the objective quantitative indicators to help doctors find illness in time.Based on the related theory,this thesis mainly includes the following aspects.First,using direct directed transfer function(dDTF)to solve the Granger causality of three groups' magnetoencephalography(MEG)signals,the effective brain network of three groups based on the adjacency matrix of Granger causality is constructed.The results show that the patients have stronger effective connection in the frontal lobe and have weaker effective connection in the temporal lobe.These abnormal connections may explain the patient's mental disorders,even auditory hallucinations and visual hallucinations and other symptoms.It is worth noting that the uneffected relatives have weaker effective connection in the left central region,which may be a recessive symptom during the family genetic process.Second,the thesis researches the global topological properties of the complex brain network,such as the network topological entropy,average clustering coefficient and global network efficiency,and analyzes the difference of barin network under different brain rhythms.The results show that there is no significant difference in the global network properties among the three groups,while there are significant differences in dDTF values under different brain rhythms.The global characteristics of the network in normal controls and patients under different rhythms are similar.The global measurements of the network of patients' relatives are generally higher than that of the other two groups in the delta band,while in the alpha and beta bands are generally lower than those of the other two groups.Third,the thesis analyzes the local topological properties of the complex brain network,including the degree of centrality,betweenness centrality,closeness centrality and core numbers of nodes.The results show that the brain network of the normal is the densest and there are more core nodes and connections,while the brain networks of the patients and their uneffected relatives are relatively sparse.The patients have stronger effective connection in the frontal lobe and have weaker effective connection in the temporal lobe compared with normal controls.One node of the patients and their uneffected relatives loses the effective connection with the rest of the nodes.The central frontal area has the highest betweenness,which is the core position in the brain network.The patients and their uneffected relatives have higher betweenness in left occipital than normal controls.The closeness centrality of most nodes of the uneffected relatives is significantly higher than that of the normal and patients.At last,different thresholds are selected to study the similarities and differences between the brain hierarchical networks of the three groups.The results show that,under the high and medium threshold,the information flow in the same brain area is more than that between the brain areas,which indicates that the brain is more in line with the global sparse hierarchical modular networks proposed by Hilgetag et al.Through the analysis of network topological entropy,it is found that the network topological entropy of the three groups is the highest under the medium threshold,and the lowest under the low threshold.Under the medium and high thresholds,the network topological entropy of the normal is the highest,followed by the patient,and the uneffected relatives is the lowest.