EEG based environment classification during cognitive task of multiple sclerosis patients

Multiple sclerosis (MS) is a neurodegenerative central nervous system disease in which the tissues in the brain, cerebellum, brain stem, and spinal cord are damaged as a result of the immune system disorder. The aim of this study is to classify the environment from the EEG signals recorded during the cognitive task in the computer and virtual reality environment of MS patients and healthy volunteers. Multilayer perceptron (MLP), k-nearest neighbors algorithm (kNN), and Support Vector Machine (SVM) classifiers' performances are compared using EEG signals during a cognitive task of 11 MS patients and 28 healthy volunteers. EEG signals of volunteers are separated into alpha, beta, gamma, delta, and theta subbands with Wavelet Daubechies (db2). Spectral and statistical features of the subbands are extracted. The most important features are determined by the Recursive Feature Elimination (RFE) algorithm. Training and testing data are separated by Leave-One-Out Cross-Validation. While the best environment classification for healthy volunteers is 91.07% accuracy with the SVM classifier, the best classification performance for volunteers with MS is 95.45% accuracy with the kNN classifier.