Classification of multiple sclerosis (MS) and healthy electroensefalography signals

Multiple Sclerosis (MS) is a chronic, inflammatory demyelinating disease of the central nervous system that is thought to be autoimmune. Early diagnosis of MS is of great importance for the treatment and course of the disease. Magnetic resonance imaging (MRI), cerebrospinal fluid (CSF) and evoked potentials are used in the diagnosis of MS. These methods are invasive or expensive, making it difficult to diagnose. In this respect, electroencephalography (EEG) analysis may be a method that will contribute to the pre-diagnosis of the disease due to the effect of cerebral cortex of the brain as a result of demyelination and axonal damage seen in MS. This is because EEG measures the electrical potential generated by the synchronized activity of thousands of neurons. Based on this information, in this study, EEG signals obtained from healthy individuals diagnosed with MS in Neurology Clinic of Başkent University Ankara Hospital were analyzed and the differences that could distinguish these two groups were determined. Based on these differences, it is aimed to develop a method that can classify MS and healthy individuals with high accuracy by machine learning approaches. In the first stage of the study, synchronization analysis was performed between inter-hemispheric and intra-hemispheric channel pairs for EEG signals taken in eyes closed resting position. As a result of the coherence analysis, the areas of the frequency ranges corresponding to the EEG subbands (delta, theta, alpha, beta and gamma) under the normalized coherence spectrum curve and the mutual information values of the channel pairs were determined as features. In the second stage of the study, continuous wavelet transform method was applied to the EEG signals received in the case of photic stimulation at 5 Hz, 10 Hz, 15 Hz, 20 Hz and 25 Hz frequencies. As a result of this analysis, wavelet transform coefficients corresponding to these stimulation regions were calculated respectively. Sums, maximums, standard deviations and minimums of absolute wavelet coefficients corresponding to frequency ranges "1-4 Hz" and "4-13 Hz" in each stimulation regions were determined as features. Using the features determined for the eye closed and photic stimulation EEG signals, classification studies were performed with the nearest neighbor, support vector machine, decision trees and ensemble methods. At the end of the study, a decision support system was developed for eye-closed EEG and photic stimulation EEG signals that can distinguish MS and healthy at high accuracy rate.