Analysis And Classification Of Depression Based On EMS And EEG

Depression is not only the main cause of disability in the world,but also the main cause of global disease burden.However,due to the complexity of the pathogenesis of depression,the relevant pathogenesis is still in the exploration stage.In addition,because the diagnosis of depression is mainly based on the traditional psychiatric interview,its subjectivity makes the early diagnosis still challenging.Previous studies have confirmed the cognitive dysfunction of depression patients in dealing with emotional conflict.In fact,researches on depression mainly focus on negative attention bias in emotional information processing and conflict control obstacles in executive function.However,as a result of the interaction between the two,emotional conflict disorder is more complex.Therefore,the neural mechanism of brain detecting and solving emotional conflict is still unknown to a large extent.In order to further explore the differences between depression patients and normal people in dealing with emotional conflict mechanism,based on the face word Stroop paradigm,this paper collected behavioral and physiological signals of MDDs and HCs,and analyzed the differences between them from two aspects of attention deployment mode and brain function integration.In the end,we use the collected physiological signal data to classify depression from the aspects of eye movement single mode,EEG brain network global attribute single mode and the combination of two modes on the feature layer.The main work of this paper is as follows:(1)Based on the face word Stroop task,the behavioral data and eye movement signals of 49 MDDs and 50 HCs were collected.In behavioral data,the ACC of MDDs was significantly lower than that of HCs.In the aspect of eye movement signal,this paper evaluates the directional response by the latency of initial fixation,checks the initial maintenance component by the first fixation duration,and measures the later attention maintenance component by the total fixation duration.Results there was no difference between the two in the directional response stage,but at the beginning and the end of the maintenance stage,the negative attentional bias and the lack of positive emotional protective bias were found in MDDs.(2)Similarly,EEG signals of 34 MDDs and 38 HCs were collected,and the adjacency matrix of brain functional network was constructed by calculating the Coh and PLV.After the connection matrix was binarized by selecting the threshold value,the better performing brain network construction method(PLV)was selected for further calculation.The measure of complex network based on graph theory analysis method is presented.It is found that the brain function network of HCs is more regular than that of MDDs in ? band under the condition of happy congruence.Under the condition of sad congruence,the brain network topology of HCs is more robust.And the HCs' small world attribute is stronger.For local network attributes,the results showed that there were differences in the LC regions under the condition of inconsistent stimulus or conflict.In the happy mood,the difference is located in the left hemisphere,while in the sad mood,the difference is mainly concentrated in the right hemisphere.(3)Finally,based on the eye movement and EEG data collected in the above two parts,this paper classifies depression using machine learning method under different stimulation conditions of Stroop task.The results show that the accuracy of single-mode eye movement signal classification is 70.83%(one method is reserved for verification).In the feature layer,based on CFS + best first feature selection strategy and statistically significant difference feature selection strategy,through linear fusion of the global attributes of brain function network graph theory,the classification accuracy has been improved under different stimulation conditions,up to 72.22%.The results of single electrode classification using local measurement of brain function network show that the accuracy of E19,E20,E27,E28 electrode classification is more than 70%,which has obvious advantages in distinguishing MDDs and HCs.The results showed that the attention deployment patterns of patients with MDDs and HCs under emotional conflict tasks were significantly different in the later stage,and their brain function networks were significantly different under the condition of happy consistent stimulation.The brain networks of HCs were more random and had stronger small world attributes.Finally,the results of depression classification provided by this study can provide an effective basis for early diagnosis of depression.