Related Experiment Video
Updated: May 29, 2025
Author Spotlight: Advancing Pediatric Epilepsy Surgery in Children Through Novel Biomarkers and Enhanced Localization
Published on: September 20, 2024
Accuracy of SEEG Source Localization: A Pilot Study Using Corticocortical Evoked Potentials
Benjamin C Cox1,2, Rachel J Smith3, Ismail Mohamed4
1Department of Neurology, University of Alabama at Birmingham, Birmingham, AL.
Introduction:
EEG source localization is an established technique for localizing scalp EEG in medically refractory epilepsy but has not been adequately studied with intracranial EEG (iEEG). Differences in sensor location and spatial sampling may affect the accuracy of EEG source localization with iEEG. Corticocortical evoked potentials can be used to evaluate EEG source localization algorithms for iEEG given the known source location.
Methods:
We recorded 205 sets of corticocortical evoked potentials using low-frequency single-pulse electrical stimulation in four patients with iEEG. Averaged corticocortical evoked potentials were analyzed using 11 distributed source algorithms and compared using the Wilcoxon signed-rank test ( P < 0.05). We measured the localization error from stimulated electrodes and the spatial dispersion of each solution.
Results:
Minimum norm, standard low-resolution electromagnetic tomography (sLORETA), LP Norm, sLORETA-weighted accurate minimum norm (SWARM), exact LORETA (eLORETA), standardized weighted LORETA (swLORETA), and standardized shrinking LORETA-FOCUSS (ssLOFO) had the least localization error (13.3-15.7 mm) and were superior to focal underdetermined system solver (FOCUSS), logistic autoregressive average (LAURA, and LORETA, 17.9-21.7, P < 0.001). The FOCUSS solution had the smallest spatial dispersion (7.4 mm), followed by minimum norm, L1 norm, LP norm, and SWARM (20.8-28.3 mm). Gray matter stimulations had less localization error than white matter (median differences 3.1-6.1 mm) across all algorithms except SWARM, LORETA, and logistic autoregressive average. A multivariate linear regression showed that distance from the source to sensors and gray/white matter stimulation had a significant effect on localization error for some algorithms but not SWARM, minimum norm, focal underdetermined system solver, logistic autoregressive average, and LORETA.
Conclusions:
Our study demonstrated that minimum norm, L1 norm, LP norm, and SWARM localize iEEG corticocortical evoked potentials well with lower localization error and spatial dispersion. Larger studies are needed to confirm these findings.
