Abnormal Gamma Band Ossicliations Related To Suicide Risk Under Emotional Faces Using Magnetoencephalogram

Suicide risk is one of the most common disease faced by patients with depression,the severity of suicide risk varies in different patients and is difficult to be evaluated in clinical.Therefore,it is of greatly necessary to find biomarkers related to suicide risk in depression and predict the potential risk level.Previous studies have confirmed that abnormal gamma band oscillations are closely related to impairment of emotional and cognitive functions in depressed patients and may lead to suicide risk.Abnormalities in gamma band are critical for the study of the neuroimaging biomarkers related to suicide risk.Due to high temporal and spatial resolution of Magnetoencephalography,abnormal brain patterns in the gamma band can be found.Based on Magnetoencephalography,the paper aimed to explore biomarkers associated with suicide risk in depression under different emotional facial stimuli and to construct a predictive model of the severity of suicide risk as a clinical diagnosis decision support.It mainly contains:(1)Under different emotional modes,time frequency analysis methods were used to explore interesting time period and frequency band associated with depression disease.Comparing time-frequency value in sensor level between healthy controls and depressed patients using cluster based permutation test,the significant differences were found in high gamma band(55Hz-75Hz),middle and late stage of visual stimulation task.Therefore,the timefrequency domain related to depression disease was determined and can be used as the interested time-frequency domain for further study.(2)Abnormal connectivity pairs under maximal correlation window closely associated with suicide risk can be regarded as neuroimaging biomarkers in this field.Comparing source power under happy facial stimuli in whole brain between healthy controls and depressed patients,we found some regions of interest connected to suicide risk and divided these regions into 3 functional networks: attention network,default mode network,salience network.Then maximal connectivity window algorithm was applied to determine the exact window where the maximum connectivity appeared for different networks.Within this this time window,we compared differences between high suicide risk patients,low risk patients and healthy controls,and then obtained significant connectivity pairs survived FDR correction and post-hoc test.It was found that the functional connectivity pairs in the middle stage of visual stimuli including left posterior cingulate cortex,right orbitofrontal cortex and functional connectivity pairs in the late stage of visual stimuli including left inferior frontal gyrus and right amygdala may lead to potential suicide risk.(3)Under sad facial stimulation,semi-supervise framework was applied to find neuroimaging biomarkers related to suicide risk and predict the potential suicide risk of patients with only suicidal ideation.By comparing the significant functional connectivity between patients with suicide attempts and patients without suicidal ideation in central executive network,default mode network,salience network,connectivity pairs survived FDR correction were found as input features of model.Then patients with only suicidal ideation were clustered into 2 subgroups.Permutation test was used to observe impaired neural patterns among comparisons between all disease groups and healthy controls,resulting in gradual impaired tendencies varied by suicide risk.Model results were consistent with Nurses’ Global Assessment of Suicide Risk Scales evaluated by clinicians,which verified the reliability of model.Therefore,abnormal activities in gamma band within central executive network and default mode network can be regarded as biomarkers associated with the severity of suicide risk and semi-supervised framework can be considered as an auxiliary tool for evaluating suicide risk.In summary,our study has explored abnormal brain activities in gamma band associated with suicide risk in depression.These abnormalities may cause impairment in emotional and cognitive functions leading to suicide risk.In addition,semi-supervised learning framework can be used to identify potential suicide risk in depressed patients with only suicidal ideation.