Abstracts

Rationale: Given the potential of fast oscillations (FOs) as biomarkers in epilepsy, it is highly desirable to identify and analyse them with non-invasive techniques. Here, we present a practical approach for the detection of FOs in the high gamma (40-80 Hz) and ripple (80-160 Hz) bands in MEG signals from patients with focal epilepsy, and for the localization of the cortical sources generating these events.Methods: We analyzed a cohort of 15 patients in whom scalp FOs had been identified with EEG (Neurology 2011;77:524-31; Epilepsy Res 2013;106:345-56) and who also underwent a 60 minutes MEG recording session in a CTF system equipped with 275 axial gradiometers. We applied a MEG automatic FO detector based on a scalp EEG detector (Clin Neurophysiol, 2012;123:670-80). The detector was set to work with high sensitivity, leading to a large number of false positives. The MEG channel with highest rate of pre-detections was identified in each patient, and up to 50 automatically detected events in each band were reviewed by two neurophysiologists. Only the events that both raters considered real FOs were analyzed further (adopted criteria derived from EEG studies consisted of at least 4 oscillations standing out of the background). Source localization of the individual events was performed with the wMEM (wavelet-based Maximum Entropy on the Mean) method (IEEE Trans Biomed Eng, 2014;61:2350-64), which decomposes the signal in a discrete wavelet basis. Source localization was performed only for the time-frequency component exhibiting the highest amplitude during each FO, within either the high gamma or the ripple band. The energy distribution of the results of single FOs source localization was averaged to obtain a single cortical activation map per patient.Results: The agreement between raters was 85% for both frequency bands, and the specificity of the automatic detector was close to 10%. Overall, 12 out of 15 patients had FOs (2 only high gamma oscillations, 2 only ripples, and 8 both). The number of detected oscillations per patients was 1 to 17 for gamma and 1 to 24 for ripple. In 9/10 patients the source map of the high gamma oscillations was concordant with the presumed focus. For ripples, concordance was found in 6/10 patients. In 3 out of the remaining 4 patients there was only one detected event, which could lead to unreliable source localization results.Conclusions: Despite the unavailability of ideal sleep recordings, the limited duration of recordings, and the increased possibility of artifacts, we demonstrated that it is possible to record FOs non-invasively with MEG. The combination of an automatic detector with high sensitivity and validation by human raters is a practical approach, appropriate for the high number of channels and low FO rates encountered in MEG. We also showed for the first time that it is possible to localize the cortical generators of this oscillatory MEG activity, revealing non-invasively the focal sources of these potential biomarkers of the epileptic focus. Acknowledgements: Study supported by CIHR (MOP-93614), NSERC, FRQS, CECR, AES and the Savoy Foundation.