The Comparison of Different Dimension Reduction and Classification Methods in Electroencephalogram Signals

Objective: Electroencephalogram (EEG) signals have been broadly utilized for the diagnosis of epilepsy. Expert physicians must monitor long-term EEG signals that is sometimes difficult and time consuming process for epilepsy diagnosis. In this study, classification performances of support vector machine (SVM) and linear discriminant analysis (LDA), which are widely used in computer supported epilepsy diagnosis, were compared by using wavelet-based features of extracted from EEG signals which were derived in either normal or inter-ictal periods. In addition, principal component analysis (PCA) and independent component analysis (ICA) were used to determine the effects of dimension reduction on classification success. Materials and Methods: The EEG data were sampled from the EEG laboratory of the Department of Neurology and Clinical Neurophysiology in Adnan Menderes University. Study was approved by Local Ethics Committee with protocol number 2016/873. Ten patients with epilepsy and 10 normal were the study group. EEG signals of patients with epilepsy were contains only seizure free- epochs. EEG signals were first decomposed into frequency sub-bands by using discrete wavelet transform (DWT) and then some statistical features were calculated from those to classify it's as normal or epileptic. Results: In classification of the EEG signals, it's as normal or epileptic, we achieved 88.9 0 /o accuracy rate using SVM with radial basis function (RBF) kernel without dimension reduction. Conclusion: Results showed that SVM was a powerful tool in classifying EEG signals if it's normal or epileptic.