HP20: Fusion of EEG and MEG for the localization of high frequency oscillations in focal epilepsy

Abstract

High frequency oscillations (HFOs) may help to define the epileptogenic zone, the cortical area that needs to be removed to achieve seizure freedom after epilepsy surgery. HFOs can be detected with EEG and MEG. It is hypothesized that the combination of both (‘fusion’) could provide more information than the sum of the individual modalities. Simultaneous resting-state MEG and surface EEG data was gathered from 10 epilepsy patients who underwent subsequent invasive EEG. Beamformer virtual electrodes (VEs) for EEG, MEG and fusion data were reconstructed at the locations of invasive electrodes. Fusion was improved by signal-to-noise ratio (SNR) transformation, multivariate source prelocalization and matching the number of MEG and EEG channels. Ripples were detected with an automatic HFO detector in beamformed MEG, EEG and fusion VEs and fast ripples were detected in invasive EEG, and visually checked. In all patients, ripples were found in fusion VEs that were not found in EEG or MEG. Ripple rates were 7.7/min for EEG, 1.7/min for MEG and 3.6/min for fusion data. Taking invasive EEG as gold standard, sensitivity was 34.2% for EEG, 17.0% for MEG and 22.1% for fusion. Specificity was 70.2%, 84.0% and 80.5% for EEG, MEG, and fusion, respectively. We showed that combined EEG and MEG source reconstruction for use of ripple detection provides additional information not found in either modality alone. The high specificity indicates that noninvasive ripple locations are concordant with invasive fast ripple locations. Fusion and HFO detection can be further improved to better predict the HFO zone.