A Study Of Depression Recognition Based On EEG Functional Network And Microstates

Depression is a psychiatric disorder with widespread affective disorders.It has already had a wide-ranging impact on individuals,families,and society.The affected people's daily life is severely hindered,and they may also have suicidal tendencies.At the same time,depression has caused a serious health and economic burden on society,but the diagnosis of depression is still used in a more traditional way,resulting in low recognition accuracy of depression and problems with obvious individual characteristics.Accurate diagnosis of depression can effectively prevent the development of depressive symptoms and other psychological or physical diseases.With the development of computer-aided diagnosis,EEG signal characteristics are considered to be an effective biomarker that can be used to identify patients with depression.The search for biomarkers of depression is still the current research frontier in the field of computers and biology.Compared with the task state,the operation of obtaining the resting state EEG data is more convenient.However,the current research in this field generally has a small sample size,diversified data analysis methods,poor uniformity of results,and low persuasion.Based on this,this article mainly uses two methods,micro-state and brain network,to analyze the resting state EEG data of 27 depressed patients and 28 normal control groups,explore depression-related biomarkers,and study the brain function of depressed patients abnormal.(1)In the analysis of micro-states,the EEG data of the two groups of subjects were used as quasi-micro-states at all peaks of the GFP signal calculated by the effective brain electroencephalogram of the whole brain,and then the clustering algorithm was used,Cluster the quasi-microstates into 4 target microstates and use them as microstate templates,then use the 4 microstate templates to fit the EEG signals of each subject,and count the duration of different microstates.Attributes such as time,number of occurrences,and coverage,as well as the transition probabilities between microstates,are analyzed.Finally,use information theory and other methods to analyze the microstate sequence and make comparisons between groups.The clustering results of microstates are basically consistent with previous studies.At the same time,the statistical analysis of the duration,number of occurrences and coverage of the four micro-states found that micro-states C and D have significant differences,while micro-states B to C and The probability of transition from D to A is significantly different.Through information theory analysis,it is found that the mutual information of the four microstates with different delays is also significantly different.These differences can provide effective evidence for the diagnosis of computer-assisted depression.(2)In the study of brain functional networks,in order to make full use of the information such as the amplitude and phase of EEG signals,correlation,coherence,and phase lag index(PLI)were used in this study to construct resting brain functional networks.Through the statistical test of the brain function network of the two groups of people,it was found that the results of different functional connectivity methods are different.Among the correlation and coherence networks,the difference between depression and normal function is less,and the difference between PLI network is different.There are many and widely distributed in the left hemisphere,showing an asymmetric structure of left and right brain.In order to further explore the differences in brain function networks,by calculating the characteristic path of the network,clustering coefficients and small world attributes,it was found that there was no significant difference between the groups at the global level of the network;at the node level,the E114 electrode and coherence of the correlation network The characteristic paths on the E40 electrode of the network are significantly different.In order to explore the feasibility of studying brain function network in depression diagnosis engineering,this study used support vector machine(SVM),nearest neighbor classifier(KNN),decision tree(DT)and naive Bayes(NB),etc.The classifier performs two classifications on the three brain function networks respectively.The classification results show that the combination of SVM and coherent networks achieves the best classification results,and the classification accuracy rate can reach 90%.This result shows that the use of resting state the brain function network of EEG can be used as a biomarker for depression diagnosis,which provides an effective basis for the engineering realization of depression diagnosis.At the same time,the classification mode of the three functional networks shows that the main difference between depression and normality is located in the forehead.This discovery provides the possibility of single-channel electrode detection technology.Finally,combining the research results of the two parts of micro-state and brain function network,we can find that the resting state EEG signals of depressed people carry relevant information about depressive symptoms,which can be used in academic research for further depression-related research In engineering,it can be used as a biomarker for the effective detection of depression.However,the current way of extracting this information is more complicated,and more similar research is needed to expand the research results and analyze the research results more deeply.