Abstracts

Rationale: Magnetoencephalography (MEG) has been typically used to localize epileptic activity by modeling interictal activity as equivalent current dipoles (ECDs). Synthetic aperture magnetometry (SAM) is a recently developed adaptive spatial filtering algorithm for MEG that provides some advantages over the ECD approach. SAM-kurtosis (also known as SAM(g2)) additionally provides automated temporal detection of spike sources using excess kurtosis value (steepness of epileptic spike on virtual sensors). To evaluate the efficacy of the SAM(g2) method, we applied SAM(g2) to children with intractable epilepsy secondary to tuberous sclerosis complex (TSC) and compared localizations with ECDs. Methods: We studied 13 TSC children (7 female) with age ranging 13 months to 18 years (mean 7.3 years). We analyzed video EEG, MRI and MEG. We applied a single ECD model to localize ECD clusters. The SAM kurtosis value was calculated at each SAM(g2) virtual voxel in the patient’s MRI defined brain volume. We defined “epileptic voxel of SAM(g2) (evSAM(g2))” as those with local peak kurtosis values higher than half of the maximum. Clustering of ECDs and grouping of evSAM(g2) were containing more than six ECDs and three evSAM(g2) within 1cm of each other. We then compared both ECD clusters and evSAM(g2) groups with the surgical resection area and seizure outcome. Results: Seizures started from 6 weeks to 8 years of age(median, 6 months), and became intractable secondary to multiple tubers in all cases. Ictal onset on scalp video EEG was lateralized in eight patients (62%). MEG showed multiple ECD clusters in seven patients (54%). SAM(g2) showed multiple groups of epileptic voxels in eight patients (62%). Co-localization of grouped evSAM(g2) with ECD clusters ranged from 20 to 100 % with a mean of 82%. Eight patients underwent resective surgery of single (one) and multiple lobes (seven), with six achieving seizure freedom. In eight patients who underwent surgery, seven patients had the resection area covering ECD clusters and grouped evSAM(g2). The remaining one patient had the resection area partially including ECD cluster and grouped evSAM(g2). Six of the seven patients became seizure free. Conclusions: The combination of grouping SAM(g2) and clustering ECD analyses succeeded in localizing the complex epileptic zones in TSC children with intractable epilepsy secondary to multiple cortical tubers. For subset of TSC children who present with early onset and non-lateralized seizures, MEG using SAM(g2) and ECD might identify suitable candidates for resective surgery to control seizures.