Nonlinear feature comparision of EEG using Correlation Dimension and Approximate Entropy

Abstract

The automated seizure detection in EEG is significant for epilepsy monitoring, diagnosis and rehabilitation. In this study, we evaluated the differences between epileptic EEG and normal EEG by computing some nonlinear features. Correlation Dimension (CD) and Approximate Entropy (ApEn) were calculated for one hundred segments of epileptic EEG and one hundred segments of normal EEG. A comparison is made between epileptic and normal EEG in those nonlinear parameters. Results show that the mean value of CD is 2.64 for epileptic EEG, and 5.22 for normal EEG. The mean value of ApEn is 0.64 for epileptic EEG, and 0.95 for normal EEG. Both CD and ApEn of epileptic EEG are generally lower than that of normal EEG, and there are statistically significant differences between those nonlinear features of epileptic and normal EEG signals. This indicates the degree of complexity of epileptic EEG signals is lower than that of normal EEG signals, and the nonlinear parameters such as CD and ApEn could be helpful for distinguishing epileptic EEG and normal EEG.