A Study Of Depression Identification And Source Localization Based On EEG Data

Depression, also known as major depressive disorder, the subjects who had depression sufferes from a long depressed mood, loss of interest and motivation. People who have depression tend to be younger and the disorder is easy to recur. The World Mental Health Survey estimated that depression will constitute the second largest burden of disease by the year 2020. However, there is no clear quantitative method to diagnosis depressive disorder, and could not find the incidence reason, therefore, it is imperative to find a clear approach to identify depression.EEG (Electroencephalography, EEG) is a noninvasive, safe detecting methods, with high temporal resolution, simple operation and low cost characteristics, usually used to check brain function, diagnose brain dysfunction in clinical, especially widely used for epilepsy. This paper selected 16 channels from the EEG data of 10 mild depressed subjects and 10 normal subjects. The high frequency noise, low frequency drift were removed by a band-pass filter and the ocular artifacts were also eliminated. Then linear and nonlinear characteristics were extracted, we use four kinds of classifiers to classify, analyze and evaluate the classification accuracy of mild depression and normal in each channel; we also use source location methods to analysis the abnormal brain area in mild depression, to provide doctors a supplementary assistance for medical treatment.C4.5, bayesian network, logistic regression and KNN are adopted to evaluate the classification accuracy in consideration of low computational complexity and good performance on both linear and nonlinear EEG signal features. Standardized low-resolution brain electromagnetic tomography (sLORETA) is a brain source localization method, which used to compute the location in brain source. Compared with other EEG neuroimaging methods, sLORETA has zero localization error, which makes a good credibility on brain source localization solution. The source location result showed the activities increased in the temporal lobe area when mild depressed subjects viewing negative facial pictures; the classification of single channel results exhibited a higher classification accuracy in the electrode position which is located in the right hemisphere of the brain, can effectively identified subjects with mild depression.The innovation of this article is:(1) analyzed the EEG signals of mild depression individuals and explored the abnormal brain areas of mild depression, which will have a certain contribution to prevent and treat depression; (2) used different methods to analysis EEG signals from different angles, not only distinguished the mild depression and normal individuals, but also found the abnormal brain regions may appeared in mild depressed subjects on neuroimaging; (3) combined the linear and nonlinear features of EEG signals and analyzed with four methods; (4) we classified the EEG features for each single channel, it’s easy to use fewer electrode positions to diagnose the depression.