An accurate system to distinguish between normal and abnormal electroencephalogram records with epileptic seizure free intervals

The number of people affected by epilepsy is growing. Therefore, the design of accurate automated systems for detection and classification of electroencephalogram (EEG) signals of epileptic patients is a great aid in the diagnosis process. The purpose of this study is to present an accurate and fast automated diagnosis system to distinguish between normal and abnormal EEG records with seizure free intervals. The system is based on generalized Hurst exponent estimates at different scales used to characterize EEG records, and subsequently on a support vector machine classifier with different kernels to be employed for classification purpose. Statistical tests such as t-test, F-test, Kruskal-Wallis test and Kolmogorov-Smirnov test all show that multifractal based features are significantly different across normal EEG records and those with seizure free intervals. Finally, classification experiments following ten-fold and leave one out method cross-validation techniques yielded to 100% accuracy with low time processing cost.