Research On Diagnosis Of Depression Based On Eeg

Major Depression Disorder is one of the most common mental illnesses in the world,which not only seriously disrupts the lives of patients and their families,but also has a huge impact on economic development and social stability.Therefore,the diagnosis of Major Depression Disorder and the differentiation of the degree of depression have received more and more attention from all walks of life.However,at present,the diagnosis of depression is mainly based on depression scale,doctor consultation and clinical observation.The diagnosis results will be affected by the subjective factors of the patient and the professional level of the doctor,and have a great degree of subjectivity.Using neuroimaging technology to assist diagnosis can improve the objectivity and accuracy of diagnosis results.Among them,Electroencephalogram(EEG)is widely used in the study of brain diseases due to its millisecond-level temporal resolution and its ability to reflect the activity of brain neurons.In addition,based on the relationship between depression and the prefrontal area of the brain,this thesis selects the EEG data of the three channels of the prefrontal lobe(Fp1,Fpz and Fp2)as the research object.Two depression diagnosis models were established for single-channel and three-channel EEG data: 18 CNN and a dual-branch fusion model.For the single-channel EEG data of the prefrontal lobe,this thesis proposes to use the time-frequency map of the EEG signal as an effective input for the model.The time-frequency transform selected is the wavelet transform based on the complex Morlet wavelet,and the 18 CNN model is built based on the convolutional neural network,which realizes the two-classification of healthy and moderate depression.The experimental results show that compared with other input forms of models,image classification models,wavelet basis functions and time-frequency transforms,this method shows obvious advantages in accuracy,precision,recall and F1-score.Furthermore,the experimental results show that the classification results based on Fpz EEG data are significantly better than Fp1 and Fp2.For the three-channel EEG of the prefrontal lobe,this thesis designs a dual-branch fusion model from the perspective of making up for the loss of information when the EEG signal is converted into a wavelet time-frequency diagram,which consists of a picture feature extraction branch and a time series feature extraction branch.The time series feature extraction branch embeds the channel attention designed in this thesis,which is based on the Discrete Fourier Transform Coefficients Attention(DFTC-ATT)layer.DFTC-ATT is an extension of Global Average Pooling(GAP)from the perspective of DFT frequency.The results of the comparative experiments show that the dual-branch fusion model is better than other classification models in the tasks of healthy and moderate depression,healthy and mild depression,mild depression and moderate depression and healthy and mild depression and moderate depression.Furthermore,by comparing the two-branch fusion model when GAP is replaced by the attention coefficient vector in DFTC-ATT,it is verified that adding DFT frequency components on the basis of GAP helps to improve the classification performance.The research in this thesis provides new ideas and insights for the diagnosis of depression based on the three-lead EEG of the prefrontal lobe,and the designed model is beneficial to the promotion of depression detection.