Address correspondence and reprint requests to Dr. M. Oishi at Department of Neurosurgery, National Nishi-Niigata Central Hospital, 1-14-1 Masago, Niigata 950-2085, Japan. E-mail: firstname.lastname@example.org
Summary: Purpose: To test the sensitivity of extracranial magnetoencephalography (MEG) for epileptic spikes in different cerebral sites.
Methods: We simultaneously recorded MEG and electrocorticography (ECoG) by using subdural electrodes with 1-cm interelectrode distances for one patient with lateral frontal epilepsy and one patient with basal temporal epilepsy. We analyzed MEG spikes associated with ECoG spikes and compared the maximal amplitude and number of electrodes involved. We estimated and evaluated the locations and moments of the equivalent current dipoles (ECDs) of MEG spikes.
Results: In patient 1, MEG detected 100 (53%) of 188 ECoG lateral frontal spikes, including 31 (46%) of 67 spikes that activated three subdural electrodes. MEG spike amplitudes correlated with ECoG spike amplitudes and the number of electrodes activated (p < 0.01). ECDs were perpendicular to the superior frontal sulcus. In patient 2, MEG detected 31 (26%) of 121 ECoG basal temporal spikes, but none that activated only three subdural electrodes. ECDs were localized in the entorhinal and parahippocampal gyri, oriented perpendicular to those basal temporal cortical surfaces. The ECD strength was 136.6 ± 71.5 nAm in the frontal region, but 274.5 ± 150.6 nAm in the temporal region (p < 0.01).
Conclusions: When lateral frontal ECoG spikes extend >3 cm2 across the fissure, MEG can detect >50%, correlating with spatial activation and voltage. In the basal temporal region, MEG requires higher-amplitude discharges over a more extensive area. MEG shows a significantly higher sensitivity to lateral convexity epileptic discharges than to discharges in isolated deep basal temporal regions.
Magnetoencephalography (MEG) measures the extracranial magnetic fields generated by intraneuronal electric currents with superconducting quantum interference devices (1). Extracranial magnetic fields result from intracranial tangential currents, such as neuronal activity, in the fissural cortex, which makes up two thirds of the surface of the human brain (2). During MEG analysis, magnetic field recordings are fitted to an equivalent current dipole (ECD) model to localize sources of intracranial activity, such as epileptic spikes; the spike source locations are then overlaid onto magnetic resonance (MR) images of corresponding areas of the brain. Because magnetic fields are relatively unaffected by the different electrical conductivities of the brain, cerebral spinal fluid, skull, and skin, MEG can accurately localize the source of intraneuronal electric currents that contribute to extracranial magnetic fields (3).
Electroencephalography (EEG) dipole recordings delineate both radial and tangential currents (4). However, the electrical fields, as measured by EEG, are affected by the conductivities of different tissues.
MEG has clinical application for patients with partial epilepsy. Neurosurgeons use advanced multisensor helmet-shaped, whole-head neuromagnetometers as presurgical evaluation tools for functional mapping and epileptic spike–source localization. Studies that compare intracranial EEG recordings and surgical outcomes have confirmed the accuracy of spike-source localizations by MEG (5–11). However, few reports compare simultaneously recorded MEG and electrocorticography (ECoG) results (8,11,12). In addition, no study has compared simultaneous MEG and ECoG recordings of lateral frontal and deep temporal epileptic spikes.
We hypothesize that the spatial extent, depth, and orientation of aligned epileptic discharges influence MEG spike-source localization. The aim of this study was to clarify the sensitivity and accuracy of MEG spike-source localizations by correlating them with simultaneous ECoG recordings by using permanently implanted subdural electrodes in epilepsy patients.
MATERIALS AND METHODS
We studied simultaneous MEG and ECoG recordings from two patients with localization-related epilepsies refractory to currently available antiepileptic medications (AEDs). Patient 1 was a 35-year-old man who had undergone a lesionectomy for cavernous angioma in the right middle frontal gyrus when 11 years old. After the first surgery, partial motor seizures developed, consisting of tonic posturing of the left face and arm with or without secondary generalization; these became medically intractable. Patient 2 was a 22-year-old man who had mild atrophy of the left temporal lobe and complex partial seizures since age 10 years. Each patient gave informed consent for the study.
Simultaneous recordings of ECoG and MEG
For long-term video-EEG recording, we implanted subdural grids and strips for 2 weeks to localize the epileptic zone for cortical excision. Subdural grids and strips consisted of platinum electrodes (Nihon Kohden, Tokyo, Japan) and stainless steel electrodes (Unique Medical, Tokyo, Japan). The interelectrode distance was ∼1 cm (center to center). To compare simultaneous MEG and ECoG, we selected for each patient the 16 electrodes from which interictal spikes were frequently recorded. In patient 1, the selected electrodes were over the right frontal area; in patient 2, the electrodes were under the left basal temporal area (Fig. 1). We chose a single remote electrode outside of the interictal zone for reference. We simultaneously recorded ECoG and MEG at the end of the standard video-EEG recording period.
We recorded MEG with a helmet-shaped neuromagnetometer (Vectorview; 4D-Neuroimaging, Helsinki, Finland) in a magnetically shielded room. This device uses 204 planar-type, first-order gradiometers. Before recording, we digitized for data reference the positions of three anatomic fiduciary points (nasion and bilateral preauricular) and four indicator coils on the scalp with a three-dimensional electromagnetic digitizer (Polhemus, Colchester, VT, U.S.A.).
ECoG and MEG data were sampled at 300 Hz with a bandpass filter between 0.03 and 130 Hz. We collected simultaneous MEG and ECoG recordings for 20 and 15 min for patients 1 and 2, respectively.
We analyzed ECoG and MEG recordings with a bandpass filter between 3 and 45 Hz. We selected interictal spikes on ECoG and then searched for the corresponding spikes on MEG. We analyzed the amplitude of ECoG spikes at the electrode with the highest amplitude, the number of subdural electrodes activated by the spikes, and the amplitude of MEG spikes at the MEG sensor location with the highest amplitude. To measure the spatial extent of epileptic discharges, we quantified each electrode with spikes as 1 cm2. For each MEG spike, we calculated the single ECD source at the initial spike peak with a spherical model (3). We accepted as valid ECD spike sources with goodness-of-fit (GOF) >75%. Of 204 MEG sensors, we used data from the 40 neighboring sensors that encompassed both extremes of the magnetic flux distribution associated with each event. Typically, <40 sensors did not adequately cover the field, and >40 sensors introduced extraneous data from outside the magnetic flux pattern. We evaluated the dipole moments (strengths and orientation) of the MEG spike sources and overlaid the spike sources, with respect to the three anatomic fiduciary points, onto MR images of the patient's head. The patient underwent MRI with a 1.5-Tesla system (MAGNEX Epios15; Shimadzu, Kyoto, Japan) on the same day that MEG data were collected. We used simple regression analysis, t test, and two-way analysis of variance for statistical analysis.
In patient 1, video-EEG with subdural electrodes localized the ictal onset and interictal zone in the right superior frontal gyrus. The patient has had minor facial twitching without generalization for 10 months since subpial gyrectomy in the right superior frontal region. In patient 2, video-EEG with subdural electrodes localized the ictal-onset zone in the left basal temporal region, not including the hippocampus, and the interictal spikes in the basal temporal region. This patient has been seizure free for 13 months since corticectomy of the left basal temporal region.
Table 1 describes the relations among MEG and ECoG spikes, the number of electrodes activated by the ECoG spikes, and the number of spikes detected by MEG. In patient 1, MEG detected 100 (53%) of 188 lateral frontal ECoG spikes (Fig. 1). Ninety-seven percent of the detected MEG spikes had an association with ECoG spikes that activated three or more subdural electrodes. If four or more electrodes were activated, MEG detected virtually all ECoG spikes. In patient 2, MEG detected 31 (26%) of 121 basal temporal ECoG spikes. All (100%) of the detected MEG spikes had an association with ECoG spikes that activated four or more subdural electrodes. If four, five, or six electrodes were activated, the detection rate of MEG was ≤50%. If seven or more electrodes were activated, the MEG detection rate was 100%(Table 1).
Table 1. Relationships among MEG spikes, ECoG spikes, and number of subdural electrodes activated by the ECoG spikes
MEG spikes/ECoG spikes (%)
No. of subdural electrodes
MEG, magnetoencephalography; ECoG, electrocorticography.
For lateral frontal ECoG spikes, MEG detected 46% of spikes that activated three subdural electrodes, whereas for basal temporal ECoG spikes that activated three subdural electrodes, MEG detected no spikes. For ECoG spikes that activated four, five, or six electrodes, the MEG detection rate was 88% for lateral frontal spikes, but only 35% for basal temporal spikes. MEG spikes correlated to ECoG spikes in the frontal region more than the basal temporal region, with a statistical significance (p < 0.05).
As shown in Fig. 2A, the amplitudes of MEG spikes (taken at the recording location of the maximal amplitude) for lateral frontal ECoG spikes (mean ± SD, 245.2 ± 81.4 fT/cm) were larger than those for basal temporal ECoG spikes (168.1 ± 45.7 fT/cm), even though the average amplitudes of the lateral frontal ECoG spikes (taken from the subdural electrode with the largest amplitude) were smaller (434 ± 127.1 μV) than those for the basal temporal spikes (749.5 ± 158.6 μV). Larger MEG spike amplitudes correlated with larger ECoG amplitudes in the lateral frontal region, with a statistical significance of r = 0.418; p < 0.01). There was no significant correlation for basal temporal spikes.
Figure 2B shows the relation between MEG spike amplitude and the number of ECoG electrodes activated. For lateral frontal ECoG spikes, there was a significant correlation (r = 0.482; p < 0.01), with larger-amplitude MEG spikes associated with ECoG spikes that activated more electrodes. For basal temporal spikes, however, there was not a significant correlation.
ECD spike-source estimation
For patient 1, 84 (84%) of 100 MEG spikes had ECD sources with GOF >75%, and for patient 2, 20 (68%) of 31, consistent with the relative MEG amplitudes for the two patients. For patient 1, the ECD strength was 136.6± 71.5 nAm but 274.5 ± 150.6 nAm for patient 2 (p < 0.01).
ECD spike-source locations and orientations
Figure 3 illustrates ECD spike sources that met the GOF criterion superimposed onto each patient's MR images. Patient 1 had a cluster of ECD spike sources in the right superior and middle frontal gyri. The main spike sources were perpendicular to the superior frontal sulcus and tangential to the brain and scalp surfaces (Fig. 3, top). The MEG spike sources aligned on the fissural cortex were concordant with the interictal spike zone on ECoG. In patient 2, 14 of 20 ECDs were localized in the entorhinal and parahippocampal gyri, with the dipole moments oriented perpendicular to those basal temporal cortical surfaces and tangential to the scalp surface (Fig. 3, bottom). The remaining six were scattered in the temporal lobe (not in Fig. 3).
MEG detected and localized the current sources of epileptiform discharges in the cerebral convexity and basal temporal area in two patients. The areas delineated were from a smaller area than implicated by scalp EEG and were confirmed by simultaneous ECoG recordings and successful surgical elimination of seizures. In patient 1, who had frontal lobe epilepsy, MEG detected 53% of simultaneously recorded ECoG spikes and 97% of those extending over three or more electrodes. Center-to-center electrode distance was ∼1 cm; therefore the epileptiform discharges extending over a 3-cm2 area produced a strong enough extracranial magnetic field to be recorded by MEG. In the basal temporal region of patient 2, epileptiform discharges needed to extend over ≥4 cm2 for MEG to detect the projected extracranial magnetic field. As a result, MEG spikes significantly correlated to ECoG spikes more often in the frontal region than in the basal region.
The dipole source orientation of epileptiform discharges significantly affected the sensitivity of MEG. In the lateral frontal cortex of patient 1, MEG accurately localized tangentially oriented epileptiform discharges in the walls of the superior frontal sulcus that were recorded on ECoG. In patient 2, whose basal temporal epileptiform discharges produced tangential currents to the scalp but not the hippocampus, MEG detected the discharges in the basal temporal cortex.
These results are consistent with and expand on previous studies. Mikuni et al. (12) recorded simultaneous MEG and ECoG in two patients with temporal lobe epilepsy. They found that ≥4 cm2 of inferior temporal cortex was required for MEG spike-source localization. Hari (13) estimated that a 1- to 2.5-cm2 area needed to be activated for MEG detection. She based her results on the correlation of a single postsynaptic potential in a neuron and the neuronal density in the somatosensory cortex. For EEG, discharges on ≥6 cm2 of gyral cortex must be similarly active to produce a scalp-recordable electrical field (6,14).
The convex cortical surface of the brain, including the lateral frontal area, is close to the MEG detecting coil. Because magnetic fields decrease rapidly with increasing distance, this short distance allows MEG to localize weaker epileptic discharges accurately from a smaller area than for the deep basal temporal area. Thus even though the basal temporal ECoG spikes in patient 2 were of higher amplitude than those from the lateral frontal area of patient 1, the MEG signals were smaller. A possible contributing factor is that planar sensor coils have a low sensitivity for deep magnetic sources (15). The calculated source strengths, however, were consistent with the relative sizes of the epileptic discharges from the two areas. Thus epileptic discharges from deep tissues must have higher amplitudes and wider distribution for MEG detection because the magnetic fields originating from deep areas and projecting to the extracranial detectors are significantly reduced. The location of a single ECD, however, can be misleading if the sources activate a wide area of cortex and are in deep tissues (16).
Large areas including the hippocampus and neocortex in the temporal lobe can be simultaneously or sequentially activated via association fibers or by neocortical propagation alone (17). Dipole source modeling of EEG also has been used for temporal lobe epilepsy (18). However, closed-circuit neurons in the hippocampus and fast-conducting volume currents to the temporal neocortex have affected the acceptance of single EEG dipoles or a limited number of dipoles as adequate models of the generators of the interictal temporal lobe spike discharges (5,17,19). Dipole sources calculated for superficial neocortical epileptic discharges, which extended ≤20 cm2, were usually located deeper than the cortex (19), as predicted for an extended source. The MEG spike-source localizations for basal temporal spikes were not so precisely correlated with the ECoG spike localization as were those in frontal epilepsy, probably because the ECoG had an extensive field of spikes in the basal temporal cortex. In patient 2, we found MEG spike sources around the entorhinal and parahippocampal cortices, corresponding to the widely distributed basal temporal ECoG spikes. Because the hippocampus was not involved in the disease of patient 2, he underwent cortical excision alone in the basal temporal epileptic region and became seizure free. In another patient with focal basal temporal lobe epilepsy, MEG did localize the spike sources of the basal temporal epileptiform discharges (20). The orientations of these interictal spikes were radial to the temporal base and tangential to the scalp surface and permitted MEG spike source localization because the hippocampus was not involved.
When epileptic spikes in the convex cortical surface of the brain extend >3 cm2 across the fissure, MEG spikes can be detected with a probability of >50% and correlate with the spatial extent and amplitude of spikes recorded simultaneously by ECoG. In the basal temporal region, MEG spike-source localization requires epileptiform discharges of higher amplitude and more extensive distribution, because the magnetic field in this area declines significantly. MEG shows a significantly higher sensitivity to epileptic discharges in lateral convexity than to discharges in isolated deep basal temporal regions.
Acknowledgment: We thank Kuniko Tsuchiya and Tohru Kondo for technical assistances, Dr. Ayako Ochi for editorial comment, and Carol Squires for editorial assistance.