Research On Recognition Of Depression And Bipolar Disorder Based On EEG

Major depressive disorder（MDD）and bipolar disorder（BD）are common emotional disorders,which beneficial to personal life,family harmony as well as social order and stable development,and they are seriously hindered the daily life of patients.Current scientific research on mental disorders from the perspective of electroencephalogram（EEG）,iconography and molecular biology,and computers are widely used in the diagnosis of brain diseases as a tool of intelligent processing and control.EEG signals contain a large amount of physiological information.The related studies mostly focused on the time domain and frequency domain information of EEG signals.Therefore,this study concentrate on the phase characteristics of timefrequency domain and brain topographic map.The detailed arrangement in the paper is as follows:First of all,this paper briefly expounds the basic concept of EEG,and makes clear the basic principle of studying brain diseases from the EEG signal that reflects brain activity.Then,the relevant methods and technologies of EEG signal analysis are introduced in detail.The theory of brain network analysis and micro-state analysis method are introduced emphatically,and the principle and necessity of using these two methods are clarified.Finally,the acquisition process and pre-processing operation of EEG signal are introduced.Secondly,a method for depression and bipolar disorder recognition based on phase synchronization feature fusion came up at this study.From the perspective of time and frequency domain,the extraction of EEG signals respectively phase lay index（PLI）,phase locking value（PLV）and weighted phase delay index（WPLI）as characteristics and their mutual fusion,and uses the different machine learning classifiers in alpha,beta,theta and delta frequency classifying depression and bipolar disorder as well as visualization.The outcomes of experiments indicate that the mixed feature has a better effect than the single feature,and the mixed feature PLV＿PLI＿WPLI has the best performance in beta band.Visualization of brain network connections with different features showed that the main brain regions with different characteristics were distributed in the frontal lobe,temporal lobe and parietal lobe.Finally,from the perspective of brain topographical map,this paper conducted the study on micro-state identification of MDD and BD patients with eyes open,eyes closed and combine eyes open with eyes closed.Through inter-group and intra-group statistical analysis of the parameters of different micro-states in HC,MDD and BD under the condition of open and closed eyes,it was proved that the occurrence was an excellent feature to distinguish MDD,BD and HC.In view of this,we further studied the transition probability of occurrence between different states of the three groups of subjects when their eyes were open and closed,and found that micro-state C often appeared in MDD group,while micro-state A in BD group was more likely to convert to micro-state B.In addition,K-ELMC algorithm was used to identify and compare different micro-state parameters of open,closed and mixed states.It was found that,for MDD,BD and HC classification tasks,the closed eye dataset achieved better effect.Meanwhile,the occurrence identification effect of micro-state indicators is better than others.