Stability analysis of epileptic EEG signals

Macroscopic neurophysiologic models, supposed to produce EEG signals, suggest that the epileptic seizures are due to a limit of stability of the system. A technique based on multivariate autoregressive model (MVAR) and stability analysis by means of the eigenvalues of the estimated model is presented in this paper. This method has been used to analyse stability of EEG signals of epileptic patients, especially during epileptic seizure. We computed the maximum of the module of the eigenvalues on short term intervals by using a sliding window in the EEG signals. Results show that these values remain close to 1 and nearly constant during the epileptic seizure and occasionally thereafter, especially in the case of secondary generalized seizures. A new feature (IndexS) has been introduced in the purpose of emphasizing this phenomenon. The number of seizures identified correctly by this index was 11 among a total of 17 seizures.