Epileptic seizure prediction using phase synchronization based on bivariate empirical mode decomposition

Objective: Epilepsy is a common neurological disorder with unpredictability. An effective algorithm for seizure prediction is important for the patients with refractory epilepsy.

Methods: We proposed a seizure prediction method based on the phase synchronization information of neuronal electrical activities. Firstly, the instantaneous phase of the intracranial electroencephalograph (EEG) recordings was detected by the combination of bivariate empirical mode decomposition (BEMD) and Hilbert transformation. Then, the phase information was used to calculate the mean phase coherence (MPC) as a measure of phase coupling strength between different channels of EEG recordings. In the end, the preictal changes of MPC time courses were used to raise the seizure alarms. We compared the proposed method with other existing methods to further investigate its effectiveness.

Results: Both the increase and the decrease of phase synchronization were found prior to seizure onset. Our results indicated that the proposed method had the best performance among three predictors.

Conclusions: The proposed algorithm can effectively extract the phase synchrony changes prior to the seizure onset and contribute to the application of the seizure prediction.

Significance: Phase synchronization analysis based on the BEMD method may be a useful algorithm for clinical application in epileptic prediction.