Dynamical pattern of successive bits to predict the outcomes of the SSRI and rTMS depression therapies using EEG signals.

Depression is a mental disorder that can lead to suicide and self-harm if left untreated. Predicting the outcomes of depression therapy is one of the most challenging tasks for psychiatrists and neurologists. Selective Serotonin Reuptake Inhibitors (SSRIs) and repetitive Transcranial Magnetic Stimulation (rTMS) are two well-established therapies for depression, commonly prescribed as first-line treatments for different levels of depression. However, the response rate is about 50%, which is not significantly high. In this paper, we propose a novel technique called the dynamical pattern of successive bits (DPSB) to extract distinctive features from Electroencephalography (EEG) signals to predict the outcome of different depression therapies and suggest the best course of treatment. Our novel feature extraction technique is built on the nonlinear fusion of variations in successive bits. Two features, amplitude (A) and phase (ϕ), are extracted from EEG signals using the DPSB technique. These features are utilized along with a feedforward neural network architecture to predict the outcome of depression therapy. The results show that our accuracy levels are 99.40% and 99.59% in predicting the outcomes of SSRI and rTMS therapies, respectively. The model is trained and tested using a 10-fold cross-validation technique to avoid bias in the results. The findings indicate that the temporal lobe exhibits a better capability for predicting the outcomes of depression therapies. Furthermore, the T3 channel shows the best performance in predicting these outcomes. Two biomarkers, A and ϕ, are proposed using the DPSB technique for the T3 channel to illustrate the brain's complex behaviors in a 2D plane. The proposed model is embedded in software that could assist neurologists in selecting the best course of treatment for depressed patients.