Research And Application Of Brain Mechanism Based On Multi Feature Fusion

Economic globalization promotes the rapid development of economy and technology in various countries.When people enjoy the convenient life with fast-paced life,the cost of life and the pressure of life are increasing day by day.Various mental diseases come one after another,destroying the fragile physical and mental health of human beings.Depression,as the top mental illness,the patients will suffer from persistent depression,inferiority,self-mutilation and even suicide,which brings great burden on life and medical care to patients,families and society.Depression is calculated in advance to become the world’s largest disease by 2030.Therefore,we urgently need to find out the pathogenesis of depression and other mental diseases,and timely diagnosis and treatment of patients.Mental illness is generally related to abnormal brain structure and function.Fortunately,the rapid development of medical technology and computer technology provides more possibilities for further study of brain mechanism.Based on the structure and functional MRI data of depression,combined with graph theory,statistics,artificial intelligence and complex network knowledge,this paper analyzes the brain network structure and topological attributes of patients.The main work of this paper is as follows:(1)Based on the brain MRI data of depression patients and healthy subjects,the structure network and function network of the two groups of subjects were constructed by Pearson correlation coefficient.By comparing the global attributes of the networks,it is found that the clustering coefficients of the structure network and the function network of depression patients have declined to a certain extent,and their small world attributes have also declined,and there is a trend of evolution to the random network.Compared with the normal subjects,assortativity,modularity,hierarchy and synchronization of depression patients decreased to some extent.Therefore,the normal subjects’ brain functional separation and integration ability is better,and they have better logical thinking and task processing ability.In addition,after studying the local properties of depression,we found that thalamus,insula,middle frontal gyrus and other brain areas were abnormal,such as middle frontal gyrus,which may lead to slow thinking and low work efficiency of patients with depression.After calculating the gray matter volume of the two groups of subjects,we found that the gray matter volume of thalamus,hippocampus and other brain areas of depression patients changed.And it leads to the phenomenon of "slow thinking" in patients with depression.(2)Based on the time series of fMRI data,we construct two groups of hypernetworks,define and extract three kinds of clustering coefficients of hyper-networks,and study the interaction between brain regions.In addition,by K-S test statistics found that depression patients’ insula,hippocampus and other brain areas appear abnormal.Some of the abnormal brain areas found coincide with the abnormal brain areas found in the structural network and functional network,indicating that the brain structure and function of depression patients have indeed changed.Based on the correlation coefficient matrix of the network,the relationship between the networks is studied.The results show that the network connection of the structural network provides a certain physical basis for the functional connection.In addition,the aging of the brain was also studied.It was found that the volume of gray matter in the frontal lobe,occipital lobe,temporal lobe and limbic system of the elderly shrank,and the small world also showed a downward trend.This is related to the memory decline,slow response,logical thinking ability decline and other symptoms of the elderly.(3)In this paper,we studied the brain activity signals of two groups of subjects,mainly extracted and analyzed the ALFF,fALFF and ReHo signal values.According to the t-test of two samples,the abnormal brain areas were found in the insula,anterior cuneiform lobe,frontal lobe,cingulate gyrus and thalamus of depression patients.Abnormalities in these brain areas lead to the disorder of emotion regulation,the disorder of empathy,and the decline of self-cognition in patients with depression.(4)Based on the above research results,this paper screened out the characteristics with significant differences,and used NB,NN,lightGBM algorithm to build a depression auxiliary diagnosis model.In addition,PCA,LDA,AE and RFE are used to reduce the feature dimension.Finally,the features are fused,and NB,NN,lightGBM and their integrated models are trained respectively.Compared with the classification results of each model,the integrated model has the best effect of assistant diagnosis,with the classification accuracy of 88.72%,accuracy of 86.21% and specificity of 89.42%.