Functional Connectivity Analysis Of Brain Networks In Patients With Paranoid-schizophrenia Based On Graph Measurements And Graph Filtering

Paranoid-schizophrenia is usually thought to be caused by abnormalities in the brain functional connectivity networks.In this paper,the functional connectivity network of paranoid schizophrenics are analyzed based on complex network measurements and graph signal processing.Through the calculation of the strength of functional connectivity networks and the analysis of network topology,the abnormal connections between brain networks of paranoid schizophrenics and those of normal people are researched,in order to reveal the pathological mechanism of paranoid schizophrenics from the aspect of graph signal processing.The main contributions of this paper are as follows:1.Based on complex network measurements to analyze the topological structure of functional networks,this paper divides brain network into 32 functional subnetworks,calculates the local and global characteristics of brain networks by graph measurements,and analyzes the connectivity of subnetworks with large differences from two aspects of connectivity diversity and connectivity strength.The experimental results show that paranoid schizophrenics have higher global effificiency,local effificiency and clustering coeffificient,and smaller betweenness centrality,average path length and small-worldness.Paranoid schizophrenics have more connectivity diversity and stronger connectivity strength.2.Based on the analysis of functional network connections by GSP,the brain network is further subdivided into 132 brain regions,and the total variation of brain signals are observed by using the eigenvector of graph Laplacian matrices,graph low pass and graph high pass filters are used to observe the atlas and the graph filtered brain signal.Using the preprocessed region of interest to calculate the Pearson correlation coefficient to obtain the weighted adjacency matrix,then three groups of connectivity density thresholds were set to obtain binary adjacency matrices.On the total variation of the graph Laplacian eigenvector,these three cases are different in low,medium and high frequency.On the graph filtering,these three cases have significant difference after the graph low-pass filtering,that is to say,low-frequency signals are easier to distinguish the two groups of people.