Abstracts

Rationale: Single dipole model (SDM) is a standard technique of magnetoencephalography (MEG) for estimating equivalent current dipole (ECD) of interictal spikes. Synthetic aperture magnetometry (SAM) is an adaptive spatial filtering algorithm for MEG data. The spike discharge yields a large positive kurtosis (steepness). SAM kurtosis (SAM(g2)) provides source locations of intracranial discharges with excess kurtosis. SAM(g2) results have been reported concordant with the single clustered ECDs. To analyze multiple spike foci in children with tuberous sclerosis complex (TSC), we developed the systematic approach of SAM(g2) clustering analysis to provide multiple epileptic zones comparing ECDs.Methods: We analyzed MEG in five children with intractable epilepsy secondary to TSC. We used whole head gradiometer Omega system (151 channels, VSM MedTech Ltd., Port Coquitlam, BC, Canada) with simultaneous EEG recorded from 19 electrodes. We recorded at least 15 times of 2 min data set. For SDM analysis, we visually identified MEG spikes using a band pass filter of 10-70 Hz to attenuate slow waves. We applied a single moving dipole analysis with a single-shell, whole head spherical model. We defined the single ECD for each spike, from the earliest phase of spike discharges. Using clustering analysis, we localized “clustered dipoles” which contains more than 6 ECDs within 1.7cm distance between ECDs. For SAM(g2) clustering analysis, we selected five datasets with interictal spikes without artifacts in each patient. We set 5mm resolution voxel on whole brain MRI. SAM(g2) calculated kurtosis value at each voxel using a band pass filter of 20-70Hz. In each data set, we selected active voxels with local peak kurtosis higher than half of maximum value. We overlaid all selected active voxels of 5 datasets on patient’s MRI. Using clustering analysis, we localized “clustered voxels” which contain more than 3 voxels within 1.7cm distance between voxels. We compared clustered voxels with clustered ECDs. Results: SDM localized two clustered ECDs in four patients and one single clustered ECDs in one patient. SAM(g2) clustering analysis localized 12 clustered voxels in 5 patients. Eleven clustered voxels overlapped with zones of nine clustered ECDs in all five patients. One clustered voxels overlapped with scattered ECDs by SDM. Conclusions: SAM(g2) clustering analysis succeeded in localizing the multiple independent epileptic foci in children with TSC. The multiple independent clustered active voxels mainly overlapped the clustered ECDs. This systematic analysis using peak kurtosis value and clustering analysis for SAM(g2) provided consistent active clustered voxels with valuable spikes in the epileptic zones by the SDM analysis.