Temporal lobe seizure prediction based on a complex Gaussian wavelet

Abstract

Objective: Abnormal synchronisation change is closely associated with the process of seizure generation. The immediate and accurate detection of the changes in synchronisation may offer advantages in seizure prediction. Thus, we develop a phase synchronisation detection method for this purpose.

Methods: An analysis of phase synchronisation based on the complex Gaussian wavelet transform (PSW) was conducted to detect synchronised phases of long-lasting scalp electroencephalograph (EEG) recordings from eight epilepsy patients with intractable temporal lobe epilepsy. Four assessment indicators, namely sensitivity, maximum false prediction rate, seizure occurrence period and seizure prediction horizon were used to assess and compare PSW with the analysis of phase synchronisation, based on the Hilbert transform (PSH) and a random predictor Poisson process.

Results: An obvious decrease was found upon phase synchronisation prior to visual detection of electroencephalograph seizure onset, which was consistent with the EEG mechanism in the ictal events. The results suggest that PSW is the most effective among the three prediction methods.

Conclusions: The results confirm that the analysis of phase synchronisation based on the complex Gaussian wavelet transform can be used for seizure prediction.

Significance: Phase synchronisation analysis may be a useful algorithm for clinical application in epileptic prediction.