Source Localization Analysis And Classification Algorithm Study Based On ERP Data In Depression

Negative cognitive bias in depression is one of the directions that the academic community has been studying.Studies have shown that negative cognitive bias may reflect the abnormal response of depression with negative emotional stimuli,and may also reflect lack of functioning of cognitive control,which can't suppress the interference caused by task-irrelevant emotional stimuli.In recent years,with rapid development of bioinformatics and computer technology,source localization and electroencephalography(EEG)detection technology have provided new research ideas for brain research and mental disease detection.Combining EEG and source location techniques,our study firstly explore the characteristics of abnormal attentional bias in major depression.Afterword,we used face-word Stroop task to examine how attentional bias of influences the neural process of major depressive disorder(MDD)during emotional conflict.The different characteristics of the neural mechanism between MDD and controls in negative attentional bias and emotional conflict processing were compared.Finally,we used data mining to investigate whether the component of ERPs contains useful features to identify depressed patients.The main research is as follows:1.Electroencephalography data was collected in 17 patients with MDD and 17 controls during a dot-probe task.Combined with behavioral,event-related potential(ERP)and standardized low-resolution electromagnetic scanning technology(s LORETA),the processing mechanism of negative attentional bias was discussed.The results indicated that patients with MDDs had shorter reaction time(RTs)and enhanced amplitudes of P100 and P300 for valid sad trials compared with invalid sad trials.2.Our study used a face-word Stroop task involving emotional faces while recording EEG in 20 MDD and 20 controls.And then ERP components P300 and N450 were extracted and the corresponding brain sources were reconstructed using the sLORETA.Behaviorally,subjects with MDDs manifested significantly increased Stroop effect when examining the RT difference between happy incongruent trials and happy congruent trials,compared with controls.ERP results exhibited that MDDs were characterized by attenuated difference between P300 amplitude to sad congruent stimuli and sad incongruent stimuli,as electrophysiological evidence of impaired conflict processing in subjects with MDD.The sLORETA results showed that MDD patients had higher current density in rostral anterior cingulate cortex(rostral ACC,BA24?32?33)within N450 time window in response to happy incongruent trials than happy congruent stimuli.This may imply that negative emotional distraction lead to greater conflict in MDDs and therefore require more cognitive resources to overcome it.Moreover,healthy subjects had stronger activity in right inferior frontal gyrus(rIFG,BA47)region in response to incongruent stimuli than congruent stimuli,revealing successful inhibition of emotional distraction in healthy subjects,which was absent in MDDs.3.For ERP data during the dot-probe task,four classification algorithms were adopted to identify MDDs from controls,including k-nearest neighbor(kNN),C4.5,Sequential Minimal Optimization(SMO)and Logistic Regression(LR).And Correlated Feature Selection(CFS),ReliefF and GainRatio were applied for features selection.Data mining results suggested that CFS was the best feature selection algorithm,especially for the P300 induced by valid sad trials,classification accuracy of combination CFS with any one of the classifiers reached more than 85%,which KNN(k=3)reached higher classification accuracy up to 94%.Through the above studies,we found that the affective processing bias in MDD begins in the early stages of perceptual processing and continues at later cognitive stages.Depressive patients generally pay more attention to negative information and difficult to disengage from negative emotions.Secondly,compared with controls,impaired inhibition of task-irrelevant emotional stimuli in MDDs was demonstrated through novel behavioral and neurophysiological evidence.Finally,data mining results demonstrated that using P300 component duing dot-probe task,combined with CFS and kNN algorithm can be good detection for patients with MDDs.These findings not only increase our understanding of the mechanism of negative attention bias and emotional conflict processing,but also provide an effective method for identifying depression patients.