Abstracts

Rationale: Ictal EEG events often exhibit phase differences and offsets in their initiation times from electrode to electrode. These differences can be difficult to discern when EEG traces recorded at different electrodes are compared with each other visually. Detecting these differences can be clinically relevant, such as when deciding whether a seemingly generalized epileptiform discharge is truly generalized, as opposed to being focal with secondary bilateral synchrony. We present a novel technique to better visualize phase and time differences of epileptiform discharges using the Stockwell transform, which is a Fourier-based time-frequency analysis method.Methods: We analyzed EEG from three patients, each of whom had one prolonged (10-20 seconds) burst of generalized spike and wave discharges with varying clinical symptoms, but not generalized convulsions. Conceptually, the Stockwell transform is a hybrid of the short-time Fourier transform with the Morlet wavelet transform, in that its “wavelet” has a scalable Morlet-like amplitude envelope that translates in time, but a non-translating Fourier sinusoid to provide the oscillations. The resulting Stockwell amplitude and phase spectra depend on both time and frequency, instead of only frequency, as with the Fourier transform. Our estimates of inter-electrode phase difference are obtained by simply comparing Stockwell phase spectra of data traces from different electrodes. The time offsets are then obtained by calculating frequency-domain gradients of these phase differences; this is analogous to the link between frequency “phase ramping”, and time translation, in ordinary Fourier analysis.Results: We observed that electrode-to-electrode phase differences and time offsets of the spikes evolved over time for all ictal discharges. These phase differences also depended upon both frequency and electrode position. Specifically, immediately after seizure onset, the estimates of phase shift and time translation were unstable, but they assumed a more organized appearance 1-2 seconds after ictal onset. For the remainder of the seizure, the discharges were not truly generalized. Instead, one or two electrodes (typically central or parietal) displayed the earliest change for a given discharge. Furthermore, the precise electrode that displayed the first electrographic change for a discharge, changed at least once the course of the seizure. Conclusions: We have shown that the phase relationship between spikes during prolonged bursts of generalized spike and wave discharges can be visualized using the Stockwell transform. The results show that generalized ictal events are complex phenomenon that are not truly generalized in nature and that their electrophysiological characteristics change over time for a given ictal event.