Colgate University (New York)
Samantha Frank, Dr. Sara Inati, Dr. William Theodore
Background: 65 million people worldwide are affected by epilepsy.20-40% of these patients suffer from refractory epilepsy, meaning that seizures persist even after pharmacological interventions, and these patients may be candidates for surgical resection. Magnetoencephalography(MEG) is one imaging method that may be used to localize the origin of the patient’s seizures presurgically. As opposed to EEG, which measures the electrical activity of the brain, MEG measures the correspondingmagnetic field generated by brain activity. Many MEG analysis tools have been developed, but most centers rely exclusively on dipole modeling for source localization. Other source localization tools have not been extensively tested for use in patients with epilepsy. The goal of this preliminary study was to compare 3 source localization techniques using 2 different software packages to assess their concordance with each other and with surgical seizure focus localization.
Methods: We compared the localization results of source localization using the ECD technique inboth CTF and CURRY, 2 beamforming techniques (CTF SAM and CURRY vector beamformer), and a current density model (CURRY SWARM) with surgical seizure focus localization in 14 patients who remained seizure free at one year post-resection (Engel Class 1 or 2).
Results: We found that all methods had similar concordance with surgical localization (approximately 64% of patients for each technique). The CURRY vector beamformeranalysis produced the highest number of results localizing only the area of surgical resection (57% of patients). When comparing the two software packages, we found that the dipole localization results were concurrent in 10/14 patients; beamformer localizationswere only concurrent in 7/14 patients. The current density method only showed concurrent localizing results in 2/14 patients.
Discussion: Although traditional epilepsy MEG source localization relies exclusively on ECD methods, our preliminary analysis shows that other techniques may be relatively similar in clinical localizing ability. Using multiple analysis methods when looking for interictal spikes in epilepsy may increase the yield of the test, and beamformers can provide additional or confirmatory localizing information. Additionally, all methods had similar rates of non- or mis-localizing results. ECD methods are more uniform across the two tested software packages, likely because they rely on direct visual analysis of raw data and have fewer differences in technical implementation. The current density method that was tested provided little localizing information. In the future, additional patients will be analyzed to obtain more robust results, and additional localization methods may be evaluated in a similar fashion. In addition, patients with poor surgical outcomes may yield additional information about false positive results.
Last updated November 26, 2013