Despiking SEEG signals reveals dynamics of gamma band preictal activity

Interictal epileptiform discharges, or 'interictal spikes', are the hallmark of epilepsy. Still, there is growing evidence that oscillatory activity - whether in the gamma band (30-120 Hz) or at higher frequencies is another important marker of hyperexcitable tissues. A major difficulty arises from the fact that interictal spikes and oscillations overlap in the frequency domain. This hampers the correct delineation of the cortex producing pathological oscillations by simple filtering. Here, we propose a nonlinear technique for fitting the spike waveform in order to remove it, resulting in a 'despiked' signal. This strategy was first applied to simulated data inspired from real stereo-electroencephalographic (SEEG) signals, then to real data. We show that despiking leads to a better space-time-frequency analysis of the oscillatory part of the signal. Thus, in the real SEEG signals, the spatio-temporal maps show a buildup of gamma oscillations during the preictal period in the despiked signals, whereas in the original signals this activity is masked by spikes. Despiking is thus a promising venue for a better characterization of oscillatory activity in electrophysiology of epilepsy.