Abstracts

Rationale: The goal for surgical treatment of medically intractable epilepsy is removal of sufficient epileptogenic cortex to stop seizures without incurring functional impairment. Convergence of findings from multimodality presurgical evaluations has been used to plan grid placement. Within clinical neurophysiology, EEG and magnetoencephalography (MEG) are non-invasive techniques with fine temporal resolution. Visually identified interictal and ictal discharges in both are often analyzed with single or multiple equivalent current dipole (ECD) models. More recently spatial filter algorithms and current density reconstructions have been used to localize epileptic activity. Intracranial ECoG has been regarded as the ‘gold standard’. However, surface findings of grids may not well represent deeper cortical sources. Algorithms traditionally used for analysing extracranial signals may improve localization of deep sources from intracranial surface ECoG signals. Methods: Interictal and/or ictal onset of extracranial MEG/EEG and intracranial EEG were recorded for 20 patients who underwent surgical treatment for intractable epilepsy. These data were analyzed with ECD models and current density reconstructions including standardized low resolution brain electromagnetic tomography (sLORETA), minimum norm (MN), scan methods with multiple signal classification (MUSIC) and synthetic aperture magnetometry with excess kurtosis statistic (SAM(g₂)) using Curry6 software and CTF MEG software. The MEG data were acquired using a 275-channels CTF MEG at 300Hz and/or 4 kHz sampling rate. At least 30 minutes recording for each patient were analyzed with SAM(g₂). In addition, independent component analysis (ICA) was performed for separating spikes from 4 kHz sampling rate MEG data so as to have sufficient data samples with a narrow time range at high signal to noise ratio (SNR). SNR values were also compared among those methods. The peaks of sLORETA, minimum norm volume, SAM(g₂) volume, and ECDs were measured and compared with ICA and SNR. The localization results were co-registered to patients’ MRI scans and 3D reconstructed images and then compared to localization of the epileptogenic zone by subdural grid array. Results: All methods had good agreement for MEG signal spikes with good SNR. The localization for MEG/EEG recordings were close to the seizure onset location as determined by ECoG in some but not all cases. One particularly interesting comparison was between the cortical grid surface voltage contour/current density maps and deep source localizations, using multiple source localization methods. Conclusions: This study demonstrates how algorithms borrowed from non-invasive analyses could be applied to ECoG recordings to provide localization of focal ictal/interictal discharges from epileptogenic zones deep in gyral folds, which visual inspection of the surface grid signals alone might not be able to represent accurately.