Abstract
Correlating Magnetoencephalography to Stereo-Electroencephalography in Patients Undergoing Epilepsy Surgery
Murakami H., Wang ZI, Marashly A, Krishnan B, Prayson RA, Kakisake Y, Mosher JC, Bulacio J, Gonzalez-Martinez JA, Bingaman WE, Najm IM, Burgess RC, Alexopoulos AV. Brain 2016. doi:10.1093/brain/aww215. Epub ahead of print, August 26, 2016.
Magnetoencephalography and stereo-electroencephalography are often necessary in the course of the non-invasive and invasive presurgical evaluation of challenging patients with medically intractable focal epilepsies. In this study, we aim to examine the significance of magnetoencephalography dipole clusters and their relationship to stereo-electroencephalography findings, area of surgical resection, and seizure outcome. We also aim to define the positive and negative predictors based on magnetoencephalography dipole cluster characteristics pertaining to seizure-freedom. Included in this retrospective study were a consecutive series of 50 patients who underwent magnetoencephalography and stereo-electroencephalography at the Cleveland Clinic Epilepsy Center. Interictal magnetoencephalography localization was performed using a single equivalent current dipole model. Magnetoencephalography dipole clusters were classified based on tightness and orientation criteria. Magnetoencephalography dipole clusters, stereo-electroencephalography findings and area of resection were reconstructed and examined in the same space using the patient's own magnetic resonance imaging scan. Seizure outcomes at 1 year post-operative were dichotomized into seizure-free or not seizure-free. We found that patients in whom the magnetoencephalography clusters were completely resected had a much higher chance of seizure-freedom compared to the partial and no resection groups (P = 0.007). Furthermore, patients had a significantly higher chance of being seizure-free when stereo-electroencephalography completely sampled the area identified by magnetoencephalography as compared to those with incomplete or no sampling of magnetoencephalography results (P = 0.012). Partial concordance between magnetoencephalography and interictal or ictal stereo-electroencephalography was associated with a much lower chance of seizure freedom as compared to the concordant group (P = 0.0075). Patients with one single tight cluster on magnetoencephalography were more likely to become seizure-free compared to patients with a tight cluster plus scatter (P = 0.0049) or patients with loose clusters (P = 0.018). Patients whose magnetoencephalography clusters had a stable orientation perpendicular to the nearest major sulcus had a better chance of seizure-freedom as compared to other orientations (P = 0.042). Our data demonstrate that stereo-electroencephalography exploration and subsequent resection are more likely to succeed, when guided by positive magnetoencephalography findings. As a corollary, magnetoencephalography clusters should not be ignored when planning the stereo-electroencephalography strategy. Magnetoencephalography tight cluster and stable orientation are positive predictors for a good seizure outcome after resective surgery, whereas the presence of scattered sources diminishes the probability of favourable outcomes. The concordance pattern between magnetoencephalography and stereo-electroencephalography is a strong argument in favour of incorporating localization with non-invasive tools into the process of presurgical evaluation before actual placement of electrodes.
Commentary
Both magnetoencephalography (MEG) and stereo-EEG (sEEG) have been in use for a number of years and are now considered the workhorses of presurgical evaluation of patients with treatment-resistant epilepsies. As a reminder, MEG noninvasively measures magnetic fields produced by electric charges moving from one point to another with the directionality of the current flow determined by the right-hand rule (1). Of importance are several features of the MEG recordings—the strength of the signal is inversely proportional to the distance from the detector (i.e., deeper sources are more difficult to detect), signal detection is best from sources tangential to the skull with sources that are purely orthogonal (basically not detectable); of importance is the fact there may be multiple electrical sources of epileptiform discharges in the brain that may be clinically important—these and other features, such as dipole clustering, are important for the correlation between MEG and sEEG and for surgical outcomes as examined in the manuscript by Murakami et al. The results of MEG, when epileptiform discharges are detected, have an overall fairly high chance of influencing the decision regarding placement of intracranial electrodes during phase II video/EEG evaluation (2) and eventually detecting the ictal onset zone (3). On the other hand, sEEG is one of the methods of invasive monitoring performed to localize the ictal onset zone and determine the feasibility of subsequent curative resection that has been utilized in Europe for more than 50 years and is gradually gaining momentum in the United States (4, 5). This method of intracranial monitoring was developed to address the limitations of other invasive methods, such as grid-based subdural monitoring that include surgical morbidity and relative difficulty with evaluating deep brain, bilateral, or large structures. Although there is a place for both methods of invasive seizure monitoring, the grids have an advantage when it comes to superficial seizure foci and when cortical mapping is needed; the advantage of the sEEG monitoring lies in its ability to assess deep or bilateral brain structures. This interest and surge in the utilization of sEEG in the last few years has prompted various studies of its relative safety and utility in patients with treatment-resistant epilepsies and discussions at various fora including the recent “Battle Royale II: StereoEEG versus Subdural Electrodes” Special Interest Group meeting at the 2016 Annual Meeting of the American Epilepsy Society. However, since large prospective and randomized studies comparing grid-based and depth-based (sEEG) evaluations are not available, we continue to depend on our training and experience to make the best educated guesses regarding the type of evaluation and electrode localization, as recently reflected in the Q-PULSE survey published in the May/June issue of Currents (Koubeissi 2016;163:206–208).
In their relatively sizeable, albeit retrospective, study (N = 54 patients with MEG and sEEG data), Murakami et al. analyzed multiple features of the MEG data, including clusters’ tightness (tight, loose, scattered) and orientation (stable vs. variable), the completeness of resection based on the sEEG evaluation—compete (>70% of the MEG cluster resected) versus partial versus discordant and the surgical outcomes. Since insertion of the sEEG electrodes was dictated by the results of the entire presurgical evaluation and not just the MEG results, there was various overlap between the MEG dipole and sEEG electrode insertion—ranging from complete overlap to complete lack of overlap. The outcomes depended on the completeness of the MEG dipole resection and whether the sEEG sampled completely the MEG dipole with partial concordance between MEG and sEEG, resulting in significantly less favorable surgical outcomes. Thus, in a retrospective way, these authors were able to demonstrate that the resection is more likely to result in seizure-free outcomes when guided by the results of sEEG, which suggest that sampling of the MEG dipole is crucial during the invasive monitoring for the long-term outcome. While these data are retrospective, they argue not only for performing MEG in all patients who undergo evaluation for surgical treatment of extratemporal lobe epilepsy but also for using the MEG results as a guidance for sEEG electrode implantation. This is overall in agreement with numerous previous studies that offered similar, or the same, suggestions based on the evaluation of the concordance between MEG and grid-based evaluations.
Several clinical and translational research lessons can be derived from this study:
The MEG cluster needs to be sampled extensively whether the seizure semiology agrees with the location of the cluster or not; inadequately sampling the cluster may result in a less favorable outcome. The sEEG exploration of the MEG cluster resulted in better understanding of the meaning of the cluster as likely the initial component of an evolving discharge prior to its spread, with a tight cluster indicating small ictal onset zone and a scattered cluster indicating larger ictal zone with distant areas interconnected, signifying potentially the need for larger volume of resection. Patients in whom the sEEG did not sample or the resection did not include the MEG cluster had generally poor outcomes. This indicates that, in some cases, the interpretation of the sEEG findings is not as straightforward as we may believe, and additional validation methods may need to be in place before resection is planned in such patients. Connectivity analyses and machine learning algorithms may be able to help in the near future. Finally, probably the most important conclusion that can be made based on this retrospective study is that prospective and randomized evaluations of all noninvasive and invasive modalities utilized in the presurgical evaluation of patients with epilepsy are necessary before we determine which methods are useful or not for this process.
It is incumbent upon us—and we owe it to our patients—to develop and test the best possible algorithms for epilepsy treatment, including surgical interventions. But until such studies are completed, we will have to depend on retrospective studies, idiosyncrasies of our training and experiences, and expert opinions to make the best possible decisions for our patients. Clearly, both of the discussed techniques—MEG and sEEG—should be part of modern epileptologists’ training and pre-resection testing armamentarium.