Brain Network Analysis And Application Based On Multimodal Information Fusion

In recent years,with the increase of economic conditions,people’s pressures of daily life is also increasing.The consequent result is that the incidence of various mental diseases is increasing,which has also brought about an impact on the development of the entire society.Therefore,we urgently need to learn about the effects of these diseases on the human brain,and we need to diagnose and treat the disease earlier so that we can get prompt treatment.The development of nuclear magnetic resonance image technology makes brain network analysis more and more popular.The development of artificial intelligence also makes the diagnosis of mental diseases more objective and accuracy,which allows us to open many mysteries about the human brain as soon as possible.Based on the multimodal MRI data of depression and autism,this paper will analyze the brain network using a variety of analysis methods.According to the difference of data and analysis methods,this paper mainly consists of three aspects:(1)Using morphological analysis methods and topological analysis methods of structural networks,we would analyze the magnetic resonance image data of depression,what’s more,we also would discuss and analyze the different conclusions generated by different methods in current related research.In this paper,we found that the brain gray matter in patients with depression is abnormal,and the volume of gray matter in the brain regions such as the anterior leaflet and central posterior gyrus in patients with depression is found to be abnormal with VBM.By constructing the structural network and analyzing the network attributes,what is found that the small-world properties of the depressive network have declined,and the properties in the local network have also been abnormal.In addition,the method of brittleness analysis of brain networks was first proposed in the analysis of structural networks.It was found that depression patients are more likely to experience network collapse when they are badly upset.(2)Based on the multimodal data of autism patients,the volume of gray matter in the temporal lobe,frontal lobe,and anterior central gyrus of autism patients was significantly abnormal with the VBM analysis method.Through the construction of a structural network,the network topology properties were analyzed and the results showed that the global efficiency of autism patients was higher than that of the normal control group(p<0.05).In addition,we analyzed the changes in ReHo and ALFF for functional magnetic resonance image data in patients with autism,and found that the default network activity in autistic patients was significantly reduced.In the analysis of functional networks,it was found that the node attributes of autism patients were abnormal,and finally the correlation coefficient matrix was used to analyze the relationship between the brain structure network and the functional network.As a result,it was found that the morphological differences of autism could lead to disorder of the functional network,but the anomalies in the functional network are not obvious in the structural network.(3)In this paper,the machine learning classification model and feature selection method are applied in the diagnosis of mental diseases.In this paper,we diagnoses the structural image of depression with an accuracy of 72.86%;We also would use embedded feature selection methods and PCA to select dimensions,and LASSO was used for model training.The classification accuracy rate reached 82.32% and 81.27%.