An Automatic Scheme with Diagnostic Index for Identification of Normal and Depression EEG Signals

Detection of depression utilizing electroencephalography (EEG) signals is one of the major challenges in neural engineering applications. This study introduces a novel automated computerized depression detection method using EEG signals. In proposed design, firstly, EEG signals are decomposed into 10 empirically chosen intrinsic mode functions (IMFs) with the aid of variational mode decomposition (VMD). Secondly, the fluctuation index (FI) of IMFs is computed as the discrimination features. Finally, these features are fed into cascade forward neural network and feed-forward neural network classifiers which achieved better classification accuracy, sensitivity, and specificity from the right brain hemisphere in a 10-fold cross-validation strategy in comparison with available literature. In this study, we also propose a new depression diagnostic index (DDI) using the FI of IMFs in the VMD domain. This integrated index would assist in a quicker and more objective identification of normal and depression EEG signals. Both the proposed computerized framework and the DDI can help health workers, large enterprises and product developers to build a real-time system.