Ictal magnetic source imaging as a localizing tool in partial epilepsy

Human subjects.

Between February 1993 and June 1996, 20 patients with intractable complex partial seizures who were candidates for epilepsy surgery in the University of California Los Angeles (UCLA) Seizure Disorder Center underwent MEG study at Scripps Clinic and Research Foundation (SCRF), La Jolla, CA. In all cases, the epileptogenic zone was insufficiently localized with noninvasive studies, including scalp-sphenoidal video EEG, MRI, and FDG-PET. All the patients were being considered for further invasive video EEG monitoring.

The 20 patients comprised a subgroup of 60 patients studied with MEG during this period using a standard interictal MEG protocol.14 The distinguishing characteristic of the patients in the group was frequent or predictable seizures. A recording protocol optimized to capture seizures was utilized for the group. In order to increase the probability of an ictal episode during the recording session, the patient’s antiepileptic drugs were partially tapered a day prior to the study and the patients were sleep deprived for the whole night before the MEG study. One patient (Patient 1) had sufficiently frequent seizures to be studied fully medicated.

All patients were accompanied to the MEG study site by a responsible adult family member, and a physician/epileptologist attended the recording session. All studies were approved by the Human Research Committees of UCLA Medical Center and SCRF and written informed consent was obtained before each study.

MEG studies.

MEG studies were performed using a large array single-probe 37-channel (Patients 1 and 2) or dual-probe 74-channel (Patients 3 to 7) biomagnetometer (Biomagnetic Technologies, Inc., San Diego, CA). The biomagnetometer probe or probes were housed within a magnetically shielded room (MSR) (3 × 4 × 2.5 m). Each probe covered a recording area 15 cm in diameter over the scalp. One 37-channel probe was suspended from a ceiling-mounted gantry and the other was mounted on a positioning device that rested on the floor of the MSR. Both probes could be adjusted sufficiently to cover all areas of the scalp. The patient rested on an adjustable bed with the probe faces lightly in contact with the scalp on roughly opposite sides of the head. The probe positions were selected for each patient based on prior knowledge of the patient’s EEG and other clinical studies to optimize recordings. When possible, multiple probe recording positions were utilized to ensure full coverage.

Conventional gold scalp EEG electrodes were also attached and a simultaneous recording was taken with a 21-channel bipolar array.

Because the patients were likely to have a seizure, a staff member remained in the MSR with each patient during recordings. When a seizure occurred, the staff member immediately retracted the upper probe, allowing the patient complete freedom of movement. A closed circuit television focused on the patient and audio monitor also allowed the physician and operators outside the MSR to continuously monitor the patient.

All patients were drowsy and fell asleep during the recording sessions. The patients were allowed to remain in light sleep unless it was necessary to reposition the probes or the patient’s head, or if sleep to wake transition was felt necessary to obtain ictal data.

During recording, the accompanying physician monitored real-time displays of the EEG and MEG waveforms. Epochs of MEG and EEG data were collected when the physician detected epileptiform activity on the real-time displays and activated a trigger. By using a buffer memory, epochs containing 5 to 60 seconds of pre-trigger data and 1 to 5 seconds of post-trigger data were collected with a sample rate of 300 Hz and a bandpass of 0.1 to 200 Hz. Whereas ictal activity was of primary interest, samples of interictal activity were also obtained. The duration of studies ranged from 2 to 10 hours, with a typical study requiring 3 to 4 hours; patients who did not develop a seizure were monitored for 8 hours. This time included EEG electrode placement, informed consent, and electrode removal. Typically about 50% of the time was spent in actual data acquisition.

MEG data analysis.

The MEG and EEG data were digitally filtered with a bandpass of 3 to 70 Hz for analysis. The data epochs were visually examined, utilizing both the MEG and EEG waveforms, and segments that contained epileptiform activity and were free of motion artifact were marked.

A single equivalent current dipole (ECD) model was used to calculate the sources of the activity underlying the MEG field patterns during the selected data segments. The ECD model assumes that the recorded magnetic field can be mathematically treated as though it was produced by a very simple current configuration, consisting of a current source and sink separated by a very short insulator. Although simple, the model produces accurate results whenever the magnetic field is generated predominantly by activity arising from a focal neuronal population.14 A search algorithm was used to determine the location, orientation, and strength of a dipolar source that best reproduces the measured magnetic field.15 In the calculations, the head was modeled as a sphere with a radius that best fit the local skull curvature at the probe position. The skull shape was derived from a three-dimensional digitization of the surface of the patient’s scalp prior to the recording session. This “local sphere model” is almost equivalent to the use of a “realistic head model” in the accuracy of localization (1 to 3 mm). In the most unfavorable case, a simple homogeneous sphere model results in errors of only 3 to 8 mm for a dipolar source in a realistic skull phantom.16

The MEG field pattern at each time point in the marked segments of the digitized MEG record was examined and subjected to three criteria. When the root-mean-square amplitude of the field exceeded a criterion value of 400 femtoTesla, an ECD solution was calculated. The solution was accepted as a potentially valid source localization if the correlation between the forward-calculated magnetic field and the measured magnetic field exceeded a goodness of fit criterion of 0.98, and the strength of the dipolar source was less than 400 nanoAmpere meters (nAm). These criteria ensure that the activity had sufficient signal strength relative to background and brain magnetic noise, that the dipole model was appropriate, and that the strength of the calculated dipole source was physiologically reasonable. These criteria have been empirically derived from a series of independent studies at Scripps Clinic, University of Tokyo, and the VAMC Albuquerque in the years 1990 to 1992 and have been previously applied and cited.17-19⇓⇓ Distinct epileptic spikes without temporal overlap with other spikes typically meet such criteria. Complex patterns of temporally overlapping activity produce complicated field patterns and do not have high-correlation ECD solutions. Broadly distributed patterns expected of spatially extended current sources do not have high-correlation ECD solutions and typically have excessively high source strengths. Finally, the MEG and EEG waveforms associated with the time periods that met the three criteria were visually inspected to ensure that they reflected an epileptiform spike or sharp wave. If all numerical and visual criteria were satisfied, the localization of the source at the time point with the highest correlation for each spike or sharp wave was taken to represent the event. This time point usually corresponded closely with the peak amplitude of the event. For most patients, multiple dipoles satisfied the described criteria and were used for localization.

MRI overlay.

In a separate session, sagittal volume acquisition fast multiplanar spoiled gradient recalled (T1) images (echo time [TE]/repetition time [TR] 3.8/248), straight coronal fast spin-echo (T2) images (TE/TR 110/4,400 to 6,000), and axial fast variable echo double echo images (TE/TR 17,102/3,700) in a slice thickness of 3 to 5 mm were obtained with a GE Signa (Milwaukee, WI) 1.5-T MRI scanner operating on a 4X Advantage Platform, or, since 1995, on a 5X Echospeed Platform, and Horizon software and hardware (Milwaukee, WI). Prior to the MRI scan, lipid-filled capsules were attached at several fiducial points, including the nasion and external ear canals. The same fiducial points were registered with the three-dimensional digitizing device immediately prior to each recording run. This allowed alignment of the MEG and MRI coordinate systems so that MEG spike source localizations could be superimposed on the appropriate MRI slices. For some patients, high-resolution MRI sequences were performed in 1.8 mm partitions and three-dimensional surface rendered images were obtained. For some patients, the MEG source localizations were superimposed on three-dimensional surface rendered images showing the placement of subdural electrodes, or superimposed on MR images obtained after placement of depth electrodes.

Invasive EEG video monitoring.

All patients reported here subsequently had implantation of subdural or depth electrodes and CCTV/EEG monitoring according to standard UCLA protocol.20 One patient (Patient 4) required two invasive studies to sufficiently localize the ictal onset zone for resection. All patients underwent surgical resection and outcomes were obtained by chart review, telephone follow-up, or clinic visits for 1 or 2 years after surgery.

Ictal recordings.

Ictal MEG recordings were obtained for 7 of the 20 patients who entered the protocol. In each case, from one to four habitual complex partial seizures were captured. The studies were relatively time consuming, despite sleep deprivation and partial tapering of anticonvulsants. For most patients, the study was truncated after 4 hours of recordings; however, for several patients, no seizures were captured in a whole day of recordings.

Patient 1.

Patient 1 was a 17-year-old right-handed man with intractable complex partial seizures since age 15 that were not preceded by an aura. The seizures consisted of unresponsiveness, wandering gaze, and gestural automatisms, followed by tonic posturing of all extremities and a brief period of postictal confusion. The average seizure duration was 1 to 2 minutes. See the table for the results of presurgical studies.

Ictal and interictal MEG recordings were obtained before (left side of figure 1, pre) and after resective epilepsy surgery (right side of figure 1, post).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. MR images of the brain of Patient 1 in axial and sagittal cuts to the right of the midline are shown. The left side, labeled “pre,” shows the first study, obtained before a right frontal resection. The right side, labeled “post,” shows the second study, obtained after a limited right frontal resection. Interictal (yellow) and ictal spikes (red) that met acceptance criteria for source localizations by magnetic source imaging are superimposed on MR images. The magnetic source localizations are represented by the localization of the triangle symbols.

The presurgical interictal MEG showed right dorsolateral frontal and orbitofrontal spikes (left side of figure 1, pre, yellow triangles). The ictal and interictal EEG during the MEG session was characterized by broadly distributed frontal spikes coincident with MEG spikes.

Two seizures with repetitive MEG spikes of 9 and 12 seconds duration characterized by unresponsiveness were captured during the presurgical MEG session. Both episodes resolved without movement, allowing recording throughout. During the 9-second episode, no spikes met acceptance criteria for source localizations. During the 12-second episode, many MEG spikes localized the ictal onset to the right dorsolateral frontal and orbitofrontal area (left side of figure 1, pre, red triangles).

Semiology and interictal and ictal scalp EEG were consistent with a right frontal ictal onset, but were not sufficiently localizing. Invasive EEG recording using subdural grid electrodes covering the right orbitofrontal, dorsolateral frontal, and supplementary motor area (SMA) regions including the interhemispheric fissure bilaterally and a left frontal subdural electrode strip was performed and revealed right orbitofrontal, dorsolateral, and superior frontal interictal spikes. Ictal invasive EEG recordings showed a broad right dorsolateral and orbitofrontal ictal onset zone, confirming but not extending as far posterior as the ictal MEG onset zone.

Owing to a concern about postsurgical frontal lobe behavioral deficits, only a limited right frontal resection was performed, including the entire invasive ictal EEG onset zone but not all of the ictal MEG onset zone, which extended more posterior (data not shown). Pathologic evaluation revealed gliosis. The patient had no worthwhile improvement in seizure control over 3 years of follow-up. Gross neurocognitive deficits were not observed, and the patient refused further testing.

A postsurgical MEG session was performed and two seizures were captured. Postsurgical interictal (yellow triangles) and ictal (red triangles) MEG spikes are shown (right side of figure 1, post), demonstrating that the posterior portion of the original MEG-defined ictal onset zone was not resected, whereas most of the invasive EEG-defined onset zone was resected. Postsurgical ictal onset spikes were also seen in areas of the right frontal lobe that were not part of the original presurgical MEG-defined ictal onset zone.

Ictal MEG confirmed that the ictal onset zone corresponded to the entire irritative zone as defined by interictal MEG. The right frontal ictal onset indicated by ictal MEG was confirmed by right frontal subdural grid electrode recordings with a broad coverage.

However, the ictal MEG was superior to invasive EEG recordings, because only ictal MEG captured the posterior portion of the ictal onset zone as defined by the postresection outcome. The poor outcome in this patient is likely due to an incomplete resection, which included most of the ictal onset zone as defined by invasive EEG but not the posterior portion of the ictal onset zone as defined by ictal MEG. Because the patient decided against repeat surgery, this could not be conclusively determined.

Patient 2.

Patient 2 was a 28-year-old right-handed woman with Sturge-Weber disease and a history of intractable complex partial seizures since age 8 that were not preceded by an aura. The seizures consisted of a blank stare, unresponsiveness, gestural automatisms, and eye flutter followed by postictal micturition, amnesia, and brief postictal confusion. The average seizure duration was 10 to 40 seconds. See the table for the results of presurgical studies.

Interictal MEG and EEG showed frequent bilateral bursts of prefrontal spikes in trains lasting from 1 to 4 seconds, which were predominant on the right side (figure 2, yellow symbols).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. An MRI of the brain of Patient 2 in two sagittal cuts to the right of the midline and in two axial cuts is shown. In addition, a three-dimensional surface rendered image seen from the anterior aspect is shown. All interictal (yellow) and ictal spikes (red) that met acceptance criteria for source localizations by magnetic source imaging are superimposed. Green triangles represent the initial spread of ictal activity after seizure onset. The magnetic source localizations are represented by the localization of the triangles or circles, and the strength and direction of the magnetic dipoles are represented by the length and direction of lines that extend from the symbols. The grid coverage is shown in blue in the three-dimensional surface rendered image.

Two complex partial seizures were recorded during the MEG session. In each case, the MEG/EEG onset preceded the clinical onset by approximately 4 seconds, and the ictal MEG onset spikes were localized to the right prefrontal region lateral and adjacent to the angioma seen on MRI (see figure 2, red symbols).

Semiology and interictal and ictal scalp EEG suggested a frontal ictal onset. The right frontal venous angioma was considered an incidental finding. Thus, invasive EEG recording using right frontal subdural grid electrodes with broad coverage of the mesial and lateral surfaces including double sided interhemispheric and left frontal subdural strip electrodes was carried out and revealed diffuse right frontal interictal spikes. Ictal invasive EEG showed nonlocalizing diffuse paroxysmal fast activity.

A right frontal resection was performed, including the MEG-defined ictal onset zone but sparing a part of the interictal MEG-defined irritative zone. Pathologic evaluation revealed mild cortical dysplasia lateral to the venous angioma. The patient had no postoperative neurocognitive deficit and has been seizure free over 2 years of follow-up.

The ictal MEG localized whereas the interictal MEG was bilateral. Ictal MEG was superior to invasive EEG recordings.

Patient 3.

Patient 3 was a 15-year-old right-handed boy with a history of intractable complex partial seizures since age 5. The seizures consisted of an aura described as a “funny feeling” followed by tonic extension of the right extremities with some flailing and kicking of the left leg, followed by brief postictal confusion. The seizures usually had a brief duration of 15 seconds.

Interictal MEG spikes localized to the left parietal region (figure 3, yellow triangles). The interictal EEG during the MEG session was characterized by repetitive broad left parasagittal spikes.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. An MRI of the brain of Patient 3 in sagittal cuts left of the midline and in coronal cuts is shown. Interictal (yellow) and ictal spikes (red) that met acceptance criteria for source localizations by magnetic source imaging are superimposed. The magnetic source localizations are represented by the localization of the triangles, and the strength and direction of the magnetic dipoles are represented by the length and direction of lines that extend from the triangles.

During the MEG session, one stereotyped seizure in which the MEG/EEG onset preceded the clinical onset by approximately 6 seconds was recorded, allowing localization of the early ictal MEG spikes to the left parietal lobe, including the lateral parietal and insular regions (see figure 3, red triangles).

The available data supported a left-sided onset but did not sufficiently localize, with ictal onsets possible from either the left frontal or parietal regions. Thus, invasive EEG recording using subdural grid electrodes covering the left lateral convexity, the bilateral interhemispheric fissure, as well as the inferior aspect of the right frontal lobe was carried out and revealed interictal spikes arising from the left inferior parietal, left superior central, and left mesial frontal regions over the cingulate gyrus. Ictal onsets were localized to the left cingulate gyrus in the mid to posterior parietal portion.

A left posterior cingulate gyrus resection was performed, including the MEG-defined ictal onset zone but sparing a part of the interictal MEG-defined irritative zone. Pathologic evaluation was normal. The patient had no postoperative neurocognitive deficit and has been seizure free over 2 years of follow-up.

Ictal MEG spikes were similar to interictal MEG spikes. Ictal MEG spikes, although close but lateral to the invasive EEG onset, were not equivalent to invasive EEG recordings, because invasive EEG did not show early insular region activity, and the insular region was not resected.

MEG-defined ictal onset zones.

Six of the seven patients who had recorded seizures had definable ictal onset zones by MEG source localization. Ictal onset zones were unambiguously defined in three patients. One patient (Patient 1) remained motionless throughout a seizure. Two patients (Patients 2 and 3) had several seconds of MEG seizure activity before clinical onset.

For three patients (Patients 4, 5, and 6), the motor onsets preceded the MEG onsets, obscuring the records with muscle artifact. For these patients, abnormal sharp waves were seen in the period from 10 to 60 seconds prior to clinical onset, whose sources were successfully modeled and localized.

In most cases, ictal MEG recordings capture only a single seizure. The possibility that other seizures have a different pattern cannot be excluded.

Correspondence of MEG-defined ictal onset and irritative zones.

In this sample of six patients with MEG-recorded seizures, most of the MEG-defined ictal onset zones were smaller than the irritative zone defined by source localizations of interictal spikes.

In one patient (Patient 4), there was no localizing interictal MEG activity. In one patient (Patient 2), interictal MEG spikes were bilateral while ictal MEG spikes were unilateral. In two patients (Patients 2 and 5), ictal MEG spikes had more specific focal localization within one hemisphere than the interictal MEG spikes, and resection included the MEG-defined ictal onset zone but not all of the interictal MEG-defined irritative zone.

In three patients (Patients 1, 3, and 6), the presurgical ictal onset zone as defined by MEG corresponded to the entire irritative zone as defined by interictal MEG.

Correspondence of MEG and invasive electrode-defined ictal onset zones.

In five of the six patients with localizing MEG ictal onset zones (Patients 1, 3, 4, 5, and 6), the MEG-defined ictal onset zones were confirmed by invasive electrode recording. Ictal MEG was superior to invasive recordings in two of the six patients (Patients 1 and 2).

For one patient (Patient 2), broad subdural grid electrode invasive recordings were nonlocalizing. However, the ictal MEG spike sources were focally localized to an area of cortical dysplasia determined by pathologic evaluation.

For the other patient (Patient 1), in which the invasive electrode-defined ictal onset zone was contained within the larger MEG onset zone, only a limited resection was performed, which included most of the invasive EEG onset zone but did not include the posterior portion of the MEG-defined ictal onset zone. The patient had no worthwhile improvement in seizure frequency.

In two patients (Patients 3 and 6), invasive EEG showed seizure onsets in the cingulate gyrus. In contrast, the ictal MEG dipoles were predominantly seen in the lateral parietal region in these patients. The recorded ictal MEG activity likely represents initial ictal spread in these cases (see discussion below).

Correspondence of MEG-defined ictal onset zones with surgical outcome.

For four patients (Patients 3 through 6), the colocalized MEG and invasive electrode-defined ictal onset zones were resected, and all were much improved or seizure free for the duration of follow-up.

For one patient (Patient 2), with a nonlocalizing invasive study but a localized MEG ictal onset zone, the region indicated by MEG was resected, and the patient was seizure free at 2 years of follow-up.

For one patient (Patient 1), only a limited resection was performed, which did not include the posterior portion of the MEG-defined ictal onset zone. The patient had no worthwhile improvement in seizure frequency.

For one of the seven patients (Patient 7), in whom ictal MEG did not yield localizing data, a left SMA resection based on a two staged depth electrode and subdural electrode study was performed, with a seizure free outcome after 18 months of follow-up.

Feasibility of ictal MEG recordings.

In our study, ictal MEG recording was possible in 7 of 20 patients who entered the protocol. Despite the low yield, our data show that successful spontaneous ictal MEG recordings following sleep deprivation and partial tapering of anticonvulsants can provide useful information. The yield is obviously greater in patients with frequent seizures than in patients with relatively infrequent seizures, and the former may be preferentially selected for ictal MEG evaluation.

Movement artifact associated with seizures is a potential obstacle to successful ictal MEG recordings, because movement both disrupts the background and moves the head out of position. The protocol used in our study includes pre-trigger acquisition of 60-second data epochs immediately preceding a seizure; this ensured that the critical period between the MEG onset and motor onset was captured. For three of the patients the immediate preictal onset data were used for localizations. These localizations were later confirmed by invasive electrode recordings.

Using the described protocol, movement artifact prevented the acquisition of usable data in only one of seven patients with seizures captured by MEG.

The single equivalent dipole model used for MEG data analysis in this study assumes a highly localized source of activity, which is unlikely to be the case. In reality, most discharges are more complex in shape than assumed by a single dipole model, and a multiple dipole model might be more accurate. A further limitation arises from the assumption of sequential activation of small brain regions, which is overly simplistic. In a study of 32 patients,21 up to 30% of interictal spikes in a given patient had magnetic fields that changed over their time course. Changes of activity over time in a complex pattern would be best addressed by spatiotemporal source modeling. However, because spatiotemporal modeling is time-consuming and results of MEG dipole modeling have been validated by simultaneous intracranial recordings22 and postoperative results,23 dipole source modeling remains the most commonly used method for MEG data analysis.24

Compared to surface EEG, which records global activity, MEG is limited by being unable to record neuronal activity from gyral surfaces because only tangential vector components can be detected.5 However, this property allows more reliable source modeling of MEG data compared to surface EEG modeling.25 In a study of 32 patients,21 interictal MEG dipoles were more localized and anatomically more reasonable than EEG dipoles, which were scattered over multiple lobes. In our study, EEG modeling was not performed, and a direct comparison of MEG dipoles with dipoles calculated from surface EEG has not been made.

Because in most patients only a few seconds of ictal data were recorded, the reported procedure has a limited capability to track seizure spread. Our sample size is too small to determine how much of a limitation this is for defining the MEG ictal onset zone. However, localization of the earliest ictal activity is the most valuable information to determine the ictal onset zone.

A major concern is whether the MEG-defined ictal onset zone represents the actual ictal onset zone as opposed to a propagated ictal pattern. This issue has been addressed by a study,26 which performed a correlation analysis between MEG and ECoG recordings and demonstrated that spikes generated in deep mesiotemporal structures may escape MEG registration. Therefore, in the case of deep mesiotemporal ictal activity, MEG may in some patients be unable to reflect the initial site of seizure onset.

This issue seems to be of less concern in neocortical ictal activity. These areas have a larger area of sulci allowing a higher rate of tangentially oriented dipoles as compared to mesiotemporal lobe epilepsy with a higher proportion of radially oriented dipoles. MEG reportedly has a preferential detection of tangentially oriented dipoles.5 MEG is capable of localizing subdural dipoles within 1, 3, and 4 mm of the actual locations.27

None of the patients in our series had mesial temporal lobe epilepsy, so areas that may escape MEG registration, such as the deep mesiotemporal structures, did not represent a problem. In our two patients with invasive EEG onsets in the cingulate gyrus (Patients 3 and 6), most ictal MEG activity was more lateral in the same lobe. Thus, these MEG spikes may represent initial ictal spread.

The results of the current study, including good clinical outcome when the MEG-defined ictal onset zone was surgically resected, suggest that these did not reflect secondary propagation from another region in most of the patients studied.

Contribution of ictal MEG over interictal MEG.

In our study, in three of the six patients with MEG-recorded seizures (Patient 1 postsurgical study, Patients 2 and 5) the MEG-defined ictal onset zones were smaller than the irritative zone defined by source localizations of interictal spikes. In one patient (Patient 4), there was no localizing interictal MEG activity.

It is well known that the irritative zone is often larger than the ictal onset zone and the epileptogenic zone.28 Noninvasive and invasive studies have demonstrated that patients with bilateral interictal abnormalities may still have a unilateral ictal onset,29 and can be rendered seizure free by a unilateral resection.30 This is in agreement with our data, which suggest that ictal MEG studies are more valuable than interictal studies in the presurgical evaluation of patients with epilepsy.

Correlation of ictal MEG with invasive EEG.

In our study, MEG-defined ictal onset zones were confirmed by invasive EEG in four of six patients with MEG-recorded seizures. In one other patient (Patient 2), invasive EEG was nonlocalizing and the MEG-defined ictal onset zone was confirmed by excellent postsurgical outcome. In Patient 5, invasive EEG did not detect all of the MEG-detected ictal onset dipoles, probably because the MEG-detected dipoles were outside the area of subdural electrodes. These findings demonstrate an advantage of ictal MEG over invasive EEG, because the area that can be covered by subdural or depth electrodes is limited by the risk of higher morbidity and mortality with an increasing number of contacts.31

In Patient 3, invasive EEG was superior to ictal MEG, because the ictal MEG onset zone, which was not resected, was lateral to the invasive EEG onset zone. This finding may represent initial ictal spread (see discussion above). The initial seizure-free outcome may indicate that the MEG results in this patient may have been misleading. However, the MEG dipoles in this patient were in the deep parietotemporal operculum deep to the angular gyrus. Subdural electrodes over the cortical and mesial surfaces of the hemisphere would not have picked up this activity or may have been recording projections. The late recurrence of seizures in this patient could possibly be explained by ictal activity generated from this region, which was not resected.

In four of the seven patients reported, ictal SPECT was obtained. As determined by invasive EEG recording and postsurgical outcome, the SPECT-defined ictal onset zone was accurate in two of these four patients, whereas the MEG-defined ictal onset zone was accurate in three of the four patients, and data were unusable in one patient (see the table). Our sample size is too small to compare the sensitivity of ictal MEG to ictal SPECT. The sensitivity of ictal SPECT has been reported to be higher in temporal lobe epilepsy (75 to 95%) than in extratemporal lobe epilepsies (∼60%).32 In our experience with neocortical epilepsies, when a seizure has been successfully captured by MEG, reliability of localization and spatial and temporal resolution seems higher than that of ictal SPECT. In addition, the success of ictal SPECT depends heavily on the timing of tracer injection,32 which is not an issue in obtaining ictal MEG studies.