An adaptive strategy for detection of high-frequency oscillations in surface EEG/MEG recordings in an epilepsy patient

Introduction: High frequency oscillations (HFOs) have recently received attention as possible biomarkers for epileptic activity in both intracranial as well as surface recordings. HFOs have been used to investigate the epileptogenicity of individual brain regions and a predictive role of HFOs with respect to the clinical outcome of resective surgery has been postulated. While some of the studies published to date rely on the occurence of spikes in intracranial recordings as triggers for time-locked averaging of EEG or MEG to reveal HFOs, the present study assessed whether HFOs can be detected in surface EEG and MEG only.

Methods: We concurrently recorded EEG (36 channels, 10/20 system, sampled at 1 kHz) and MEG (BTI 256 channels, sampled at 1 kHz) in a 49 year-old patient presenting with focal non-lesional epilepsy with mostly complex partial seizures since birth. EEG and MEG data were first assessed visually by a skilled neurophysiologist with respect to general physiological and pathophysiological patterns. Spikes occurring in either EEG or MEG were used for spike time-locked averaging of the data in an attempt to reveal spike-associated HFOs. In a second line of analysis, EEG and MEG data were high-pass filtered and submitted to a continuous wavelet-transform with short, high-frequency Morlet wavelets tuned to the spectral characteristics of the HFOs observed in the spike-time-locked averages. The occurrence of HFOs in the wavelet-transformed data was then contrasted with the HFOs detected via spike-time-locking with respect to spatial patterns and overall sensitivity of the detection method

Results: Spikes occurring in EEG and/or MEG were reliably associated with peri-spike HFOs. Utilising the spectral characteristics of these HFOs for an informed wavelet transform facilitated the detection of additional HFO occurrances and thus improved the sensitivity of the analysis.

Conclusions: The strategy proposed here could be a promising starting point for the development of an adaptive spike- and HFO-detection routine independent of intracranial recordings. Given the possible pathophysiological relevance of HFOs as markers for epileptogenicity and/or predictors for post-surgical outcome, such an adaptive analysis could yield valuable information in diagnostic or pre-surgical settings.