Chaotic Prime Number Patterns for EEG Feature Extraction

This code implements the CPPN model for EEG feature extraction. Full theoretical details and validation are provided in the article.

Depression is a prevalent mental health disorder that, if left untreated, can lead to severe consequences, including self-harm and suicide. One of the primary clinical challenges in treating depression is selecting the most effective therapy for each patient. Selective Serotonin Reuptake Inhibitors (SSRIs) and repetitive Transcranial Magnetic Stimulation (rTMS) are two commonly prescribed treatments; however, both exhibit variable response rates of approximately 50\%. This study proposes a novel computer-aided decision (CAD) system designed to predict therapeutic outcomes in depressed patients using pre-treatment electroencephalogram (EEG) signals. The system introduces a new time-domain feature extraction method called Chaotic Pattern of Prime Numbers (CPPN) which captures the nonlinear and nonstationary characteristics of EEG signals. EEG data are first denoised using multiscale principal component analysis (MSPCA), followed by CPPN-based feature extraction across all channels. A neighborhood component analysis (NCA) technique is then used to select the most informative features, which are classified using a single-layer feedforward neural network (FFNN). The model is validated using a 10-fold cross-validation strategy across two datasets: the publicly available Mumtaz database (for SSRI therapy) and a clinical Atieh Hospital dataset (for rTMS therapy). The proposed approach achieves classification accuracies of 97.41\% for SSRI and 99.31\% for rTMS, highlighting its potential as a clinical decision support tool for personalized depression treatment planning. Notably, the most predictive EEG channels include Fp1, Fp2, F8, O1, O2, Pz, and A1A2 for SSRI therapy, corresponding to the frontal, occipital, parietal, and temporal lobes; and Fp2, Fz, P3, Cz, C3, C4, and Pz for rTMS therapy, reflecting activity from the frontal, central, and parietal lobes. These regions are critically involved in emotional regulation, cognitive control, visual processing, and psychomotor function, all of which are known to be altered in depression and modulated by SSRI and rTMS treatments.}