Abstracts

Rationale: Intracranial EEG not only clinically guides neurosurgeon to resect epileptogenic brain cortex in epilepsy surgery, but also experimentally provides investigators with information of normal or abnormal brain function processing. Although many signal processing techniques have been used to analyze the ictal activity or establish algorithms to estimate the seizure onset, traditional Fourier-based or wavelet-based analysis methods are not adequate to characterize the non-stationary dynamics of seizure development and propagation because Fourier and wavelet based analysis requires signal to be linear and stationary in time, and composed of sinusoidal waves of constant amplitude and period. To resolve the difficulties related to nonstationary behavior, an innovative approach, namely Hilbert Huang Transform (HHT) has been applied to physiological studies. HHT decomposes complex signals into different Intrinsic Mode Functions (IMFs). Each IMF represents a certain frequency-amplitude modulation at a specific time scale for different physiological processes. In this report, with HHT, we can explore the dynamic change of instantaneous frequency and energy of seizure activity in subdural electrocorticogram (ECoG), rather than the global frequency and energy defined by the Fourier spectral analysis. Methods: Chronic ECoG recordings from consecutive patients with subdural grids and/or strips were analyzed for this study. Patients with extratemporal (neocortical) lobe epilepsy and pharmacoresistance to antiepileptic drugs conducted intracranial EEG recording for epilepsy surgery. The seizure activity in ECoG was identified by an epileptologist. Each seizure-related EEG activity (one-minute preictal, ictal and one-minute postictal) was extracted and analyzed by HHT. Results: The complex non-stationary seizure signals were derived adaptively by the empirical mode decomposition sifting procedure basically. The instantaneous frequencies were computed from derivatives of the phase functions of the Hilbert transform of the basis function. Then the ictal EEG activity was presented in the time-frequency space. Two frequency domains were clearly demonstrated at the time-frequency spectrum. The low frequency activity oscillated at 3-5 Hz with features of slight frequency increase at seizure onset and slows down gradually. The other frequency activity oscillated at 10-20 Hz with features of no decelerating during seizure and turning to bursting generation before seizure ceased. The dynamic process of these 2 frequency domains at the neocortical seizure was presented consistently in each patient. Conclusions: The dynamic process of the distinct instantaneous frequencies suggested the consequence of cortical excitation and inhibition in the seizure. The HHT showed the power and effectiveness in analysis of the itcal EEG activity.