Single and Multiple Clusters of Magnetoencephalographic Dipoles in Neocortical Epilepsy: Significance in Characterizing the Epileptogenic Zone


Address correspondence and reprint requests to Dr. Shigeki Kameyama at Department of Neurosurgery, Epilepsy Center, Nishi-Niigata Chuo National Hospital, 1-14-1 Masago, Niigata City, Niigata, 950-2085, Japan. E-mail:


Summary: Purpose: To characterize the epileptogenic zone in neocortical epilepsy (NE) by using magnetoencephalography (MEG).

Methods: We defined and compared locations of single and multiple clusters of equivalent current dipoles (ECDs) for interictal spikes with MRI findings, ictal-onset zones (IOZs) from subdural electroencephalography (SDEEG), resected areas, and postsurgical outcomes of 20 patients who underwent cortical resection for medically intractable NE.

Results: Fourteen patients had single clusters; six had multiple clusters. Overlap of clusters and IOZs defined group A (nine patients), in which a single cluster coincided with the IOZ; group B1 (four patients), in which a single cluster was within or partially overlapped the IOZ; group B2 (five patients), in which multiple-cluster sections overlapped IOZs; group C (two patients; one single; one multiple), in which no overlap was seen. More single clusters (nine of 14) than multiple clusters (none of six) coincided with the IOZ (p = 0.014). More patients with single clusters (10 of 14) than patients with multiple clusters (one of six) had seizure-free outcomes (p = 0.049). Eight of nine patients in group A, versus three of 11 in groups B1, B2, and C, achieved seizure-free outcomes (p = 0.0098). Correlations between MRI findings and postsurgical outcomes were not statistically significant; eight of 13 patients with single lesions, one of four with no lesions, and two of three with multifocal lesions had seizure-free outcomes.

Conclusions: In neocortical epilepsy, MEG ECD clusters correlated with SDEEG IOZs. Single clusters indicated discrete epileptogenic zones that required complete resection for seizure-free outcome. Multiple clusters necessitated that the multiple or extensive epileptogenic zones be completely identified and delineated by SDEEG.

The fundamental surgical requirement for achieving a seizure-free outcome in patients with intractable neocortical epilepsy (NE) is the identification of the complete epileptogenic zone (1,2). Invasive video subdural EEG (SDEEG) can localize the ictal-onset zone (IOZ). However, complex epileptic networks in the neocortex, with entrained functional and anatomic networks, can prevent existing diagnostic tools from clearly delineating the entire epileptogenic zone. As a result, patients with extrahippocampal neocortical epilepsy have a lower seizure-free rate than do patients with medial temporal lobe epilepsy (3–5).

The clinical use of magnetoencephalography (MEG) increased after the development of whole-head measurement systems that allowed simultaneous collection of extracranial magnetic field data from the entire cortex (6). Matching an equivalent current dipole (ECD) model to the extracranial magnetic field became a simple and convenient method for localizing sources of cerebral activity. Experimental studies proved the accuracy of ECD localization of sources of interictal discharges (7–9). Studies introducing MEG into the presurgical evaluation of epilepsy patients resulted in noninvasive detection of epileptic foci in patients with neocortical epilepsy (10,11), particularly in patients with no visible or multiple lesions on MRI (12–16).

Several studies using SDEEG or scalp EEG demonstrated the significance of interictal activity in determining the epileptogenic zone or in predicting postsurgical seizure-free outcome (17–23). However, no systematic study of neocortical epilepsy has evaluated the location, distribution, orientation, and propagation of interictal ECDs from MEG and compared the resulting single and multiple ECD clusters with SDEEG findings and postsurgical outcomes.

We hypothesized that, in patients with NE, a single cluster represented a simple resectable epileptogenic zone, and multiple clusters indicated multiple or extensive epileptogenic zones in complex epileptic networks. We retrospectively compared MEG data with SDEEG results and postsurgical outcomes to characterize the neocortical epileptogenic zone.



We retrospectively analyzed data from 20 patients who underwent surgery for medically refractory NE at the Epilepsy Center of Nishi-Niigata Chuo National Hospital between April 2000 and October 2002.

Surgical strategy

We, as a team of pediatricians, neuropsychiatrists, and neurosurgeons, evaluated patients' seizure symptoms, MRIs, interictal and ictal video scalp EEGs, interictal single-photon emission computed tomography (SPECT) images, ictal SPECT images (for patients who had frequent ictal events), neuropsychological examinations, and MEG findings to identify candidates for epilepsy surgery.

MEG recording and analysis

MEG was recorded in a magnetically shielded room by using a 204-channel helmet-shaped neuromagnetometer (Neuromag 204; Elekta-Neuromag, Oy, Finland). Details of this system were described elsewhere (6). Patients received oral pentobarbiturate to induce sleep during recording. With the patient in a supine position and head positioned in the center of the helmet, we recorded spontaneous MEG and simultaneous scalp EEG for 20–30 min in one session to three sessions over a several-day period until we captured ≥10 interictal spikes for which ECDs met preset criteria. Sampling rate was 300 Hz with a band-pass filter of 0.03–160 Hz.

During off-line data analysis, we used a bandpass filter of 3 or 5–45 Hz. We visually identified MEG epileptiform discharges, spikes and sharp waves; we then cross-referenced them with the simultaneous EEG recordings. At times, we identified MEG epileptiform discharges without corresponding EEG spikes (24). After we manually selected interictal epileptiform discharges, we applied the single ECD calculation for the peak of each discharge. When one discharge had several peaks, we performed ECD calculations for several consecutive peaks to confirm the temporal course of the activity, as described elsewhere (25). The location, orientation, and strength of each ECD were calculated by using a single-sphere model of the skull. We defined acceptable ECDs as those with a goodness-of-fit to the model of >80% and with ECD strength between 100 and 400 nAm. All acceptable ECDs were superimposed onto the patients' MRIs.

Definitions after evaluation of MEG findings

We defined a cluster as 10 or more ECDs located contiguously within neighboring gyri. We identified a cluster in one area as a single cluster. A single cluster could be either focal, that is, occurring within two neighboring gyri, or extensive, distributed continuously over three or more gyri. We described several clusters separated by a distance of more than two gyri as multiple clusters.

We visually evaluated the orientations of the ECD moments in each type of cluster. We classified the orientations as uniform when the ECD moments had similar directions or as random when the ECD moments occurred in various directions.

Within multiple clusters, we analyzed the temporal relation among consecutive interictal-discharge ECDs. We defined multiple clusters as having a consistent relation when the one-way propagation from an initial spike in one cluster to the following spikes in another cluster was consistent.

SDEEG recordings

For SDEEG recording, we used 40–120 electrodes in a grid-and-strip combination (Nihon-Koden, and Unique Medical, Tokyo, Japan). The electrodes covered all areas indicated by presurgical evaluations, including MEG. Recordings were usually started on the day after electrode implantation and continued from 4 days to 4 weeks until we obtained at least three habitual seizures.

We performed SDEEG to identify the IOZ and the eloquent cortex. We analyzed seizure-onset patterns and classified them as either low-amplitude fast discharges (LAFDs), consisting of electrodecremental recruiting rhythms, or as non-LAFDs, including repetitive spike-and-waves or sharp waves. The IOZ was defined as the area, as determined by distribution of the electrodes, showing those seizure patterns before and at the onset of clinical seizures. Multiple IOZs consisted of more than one IOZ separated by more than two gyri activating either a single type of seizure or multiple types of seizures.

Correlation of clusters with IOZs

To determine the correlation or overlap between clusters and IOZs, we anatomically compared the distribution of clusters projected onto 3D-surface renderings of MR images with the IOZs overlaid onto intraoperative photographs of the brain surface with electrodes. Because an ECD from a single MEG interictal spike is a point source that represents the center of activity of an area of cortex (26), we defined the margin between cluster and IOZ as coinciding or overlapping when the marginal ECDs located within the same gyrus or within one electrode adjacent to the margin of the IOZ or both on SDEEG.

Surgical procedures

We used an intraoperative navigation system (Insight; Medtronic Sofamor Danek, Minneapolis, MN, U.S.A.) to excise the IOZ. When the initial IOZ consistently propagated after ≥5 s to other areas, the initial IOZ was resected, but the propagated IOZ was left. We removed en bloc the IOZ for pathological diagnosis. We performed subpial cortical resection by cavitron ultrasonic aspiration of gray and convolutional white matter, sparing the deep white matter. For some patients with neocortical temporal lobe epilepsy, we performed a cortical resection in addition to the anterior temporal lobectomy, sparing the amygdala and hippocampus.

We divided the surgical procedures into focal resections (those within two gyri), and extensive resections (those involving three gyri or more). One month after surgery, we performed postoperative MRIs to evaluate the residual areas of clusters and IOZs.

Neuropathologists from the Brain Research Institute, Niigata University, Japan, examined the surgical specimens.

Postsurgical seizure outcome

Postsurgical seizure control was rated according to Engel's classification (27). We regarded a seizure-free outcome as Engel's class I and improved seizure control as classes II or III.

Statistical analysis

We used multiple t test and Fisher Exact test for statistical analyses.


Table 1 summarizes the patients' profiles, MRI and MEG findings, and IOZs. This table also shows the correlations between MEG clusters and IOZs, surgical resections, pathologies, and postsurgical outcomes.

Table 1. Summary of patient profiles and examinations
Patient No.Age (years)/SexSeizure SymptomsIctal onset of video scalp EEG electrodesMRI findingsClustered ECDsIctal onset zone on SDEEGCorrelation between clusters and IOZSurgical resectionResidualPathologyEngel's class / f-u period
  1. A, ECD zone coincides and overlaps IOZ; ant, anterior; ATL, anterior temporal lobectomy sparing amygdara hippocampus; AVM, arteriovenous malformation; B, ECD partially overlaps IOZ; bas, basal; bil, bilateral; C, ECD does not overlap IOZ; CD, cortical dysplasia; CPS, complex partial seizure; DNT, dysembryoplastic neuroepithelial tumor; e, extensive; f, focal; F, frontal; f-u, follow-up in months; IOZ, ictal onset zone; inf, inferior; lat, lateral; LAFD, low amplitude fast discharges; Lt, left; med, medial; mid, middle; M, multiple; N.A., not available; O, occipital; P, parietal; post, posterior; R, random; Rt, right; S, single; sGTC, secondary generalized tonic-clonic seizures; SPS, simple partial seizure; sup, superior; TS, tuberous sclerosis; T, temporal; U, uniform; →, consistent propagation; →→– propagation with long latency.

  2. *1) prior epilepsy surgery at another hospital, *2) surgery for cavernous angioma over 10 years ago, *3) surgery for brain abscess in childhood, *4) surgery for cyst in childhood, *5) tumor removal (mixed-neuroglial tumor) 5 years ago.

 134/MSPSUndeterminedSFhigh signal on FLAIRS/fF (lat, mid-sup)US/fF (lat, mid-sup)+AfF (lat)CDIA / 31
 221/FSPS, CPST3,T5STAVM & scarS/fT (lat, mid-sup)US/fT (lat, mid-sup)+AfT(lat) + AVMAVM, scar tissueIA / 38
 322/MSPS to sGTCT4 (clinical onset preceded)normalS/fT (lat, mid-sup)US/fT (lat, mid-sup)+AeT(lat)CDIB / 49
 430/MsGTCUndeterminedSFsurgical scar*1)S/fF (bas)US/fF (bas)+B1fF(bas)F (bas)CDIIIA / 49
 524/FCPSUndeterminednormalS/fT (med-bas)US/fF (med, ant)+CfF (med,ant)T (med-bas)F (med)normal tissueIIIB / 45
 624/MSPSF4, T4M(bil)F>10 tubers (F dominant)S/fF (operculum-precentral)US/fF (operculum-precentral)AfF(lat) + tuberTSIA / 26
 735/MSPS, CPSF4, Fz, C4SFsurgical scar*2)S/fF (lat, mid-sup)US/fF (lat, mid-sup)+AfF(lat)scar tissueIIA / 43
 824/MCPSUndeterminedM(bil)Ftubers (2 Rt & 1 Lt)S/fF (lat, inf-mid)RS/fF (lat, inf-mid)AeF(lat)TSIA / 29
 918/MSPS to sGTCT4STatrophy with high signal on FLAIRS/fT (lat, mid-sup)RS/fT (lat, mid-sup)+AfT(lat)CDIA / 50
1017/FSPS, sGTCC3, P3SPporencephaly & abnormal gyriS/fP (lat, inf)RS/fP (lat)+B1eP(lat); Porencephalic margin + cortical excisionporencephalic cystIC / 49
117/MCPS to sGTCT3, T5STdiffuse atrophyS/eT (lat, mid-sup)RS/fT (med & lat)+AeATL + cortical excisionCDIA / 37
1229/FCPS to sGTCUndeterminedSPsurgical scar*3)S/eP-T(lat)RS/fP (lat, sup)+AeP(lat)scar tissueIA / 49
1352/FCPS to sGTCN.A.ST, insulatumorS/eF (operculum-insula)RS/eF,T,insula+B1eperisylvian + lesion (Insula)DNTIC / 25
1413/MCPSUndeterminednormalS/eP (lat; inf-sup)RS/fP (lat)B1eP(lat)CDIIIB / 31
1530/MCPS to sGTCF7, T3 (clinical onset preceded)SThigh signal on T2M/2T(bas)→T(lat)R&UMT (bas, ant) →→ T (lat)+B2eATLT (lat)CDIIB / 51
1630/MCPST6,O2 (clinical onset preceded)STcortical defect*4)M/2O(lat)→T(bas)RMO (lat) → T (bas)B2eT(bas) & O(lat)CDIIIA / 24
1749/FCPS, sGTCUndeterminednormalM/2T (med-bas) & T (sylvian)UMsimultaneously T (bas, ant) & T (sylv)+B2eATL & T(lat)CDIIIA / 24
1833/FCPST4, rapidly spreading to T3STsurgical scar after tumor removal *5)M/2T (bas) & P (lat)R&UMT (bas) →→ P (lat)+B2eT(bas)P (lat)CDIC / 39
1922/MSPS to CPS or sGTCUndeterminedSTdiffuse atrophyM(bil)/2bil T (med- bas)RMT (bas, post)CfT(bas,post)ipsilateral T (med), contralateral T (med-bas)CDIIC / 46
207/MCPSgeneralized but P3 dominantM(bil)P, sylvianulegyriaM(bil)/3P(Lt)→F(Lt) or P(Rt)RS/eP (lat; Lt) → (diffuse)B2eP(lat)F (lat) & contralateral P (lat)CD + destructive tissueIIIA / 23

Patients' profiles

Ages at surgery of the 13 male and seven female patients ranged from 7 to 52 years (mean, 26.5 years). Video scalp EEG showed that simple partial seizures occurred in eight patients, complex partial seizures in 14 patients, and secondarily generalized tonic–clonic seizures with or without partial seizures in 10 patients. Video scalp EEG localized seizure onset in 11 patients. Muscle or movement artifacts precluded localization of seizure onset in eight patients. In one patient, no seizures were captured during 14 days of monitoring.

MRI findings

Preoperative MRIs showed that 13 (65%) patients had single lesions: six had a single focal lesion, five had postsurgical scars or defects due to previous brain surgery, and two had diffuse temporal lobe atrophy. Three (15%) patients had multifocal lesions, such as multiple tubers or bilateral ulegyria; four (20%) had no abnormalities on MRI.

MEG findings

Based on the distributions of their ECDs, we divided patients into those with a single cluster and those with multiple clusters. Fourteen (70%) of 20 patients had a single cluster. Single clusters were focal in 10 patients and extensive in four.

Multiple clusters occurred in six (30%) of the 20 patients: five patients had two clusters; one had three clusters. The multiple clusters distributed bilaterally in two patients, but of the total number of ECDs, more were present in one hemisphere.

Seven of 10 single focal clusters had uniform orientations (Fig. 1A); three were random (Fig. 1B). The four single extensive clusters had random orientations (Fig. 1C). Of six multiple clusters, three had random orientations in all clusters; two contained one random and one uniform cluster; and one consisted of two uniform clusters.

Figure 1.

Equivalent current dipole (ECD) locations superimposed on each patient's MRI. A: Patient 1 shows a single focal cluster with uniform ECD orientations in the right middle to superior frontal region. B: Patient 8 shows a single focal cluster with random ECD orientations in the right inferior to middle frontal region. C: Patient 14 shows a single extensive cluster with random ECD orientations in the right lateral parietal region.

Three patients (patients 15, 16, and 20) with multiple clusters showed a consistent interictal propagation pattern within consecutive interictal discharge ECDs among multiple clusters. No consistent propagation pattern was seen for other patients with multiple clusters.

Ictal-onset zone on subdural EEG

SDEEG revealed a single IOZ in 15 (75%) of the 20 patients: 13 patients had focal IOZs; two had extensive IOZs. Five (25%) patients had multiple IOZs.

Patients with IOZs in the frontal lobe were older (seven patients; mean, 31.9 years) than were those with IOZs in the temporal lobe (eight patients; mean, 25.3 years; p = 0.3) or in the parietal/occipital lobe (five patients; mean, 19.2 years; p = 0.06).

Of six patients with multiple IOZs, four (patients 15, 16, 18, and 20) showed a consistent ictal propagation among multiple IOZs. Patients 15 and 18 showed the consistent propagation with long latency.

Fourteen patients had LAFDs at the seizure onset; the remaining six had non-LAFDs. LAFDs occurred in 11 (78%) of the 14 SDEEG patients with a single cluster; non-LAFDs were found in three (22%). Three (50%) of the six patients with multiple clusters and SDEEG results had LAFDs, and three (50%) had non-LAFDs.

Comparison of clusters and IOZs

Overlap of clusters and IOZs defined the following groups. In group A (nine patients), a single cluster coincided with the entire IOZ; in group B1 (four patients), either a single cluster was smaller than or partially overlapped the IOZ; in group B2 (five patients), parts of multiple clusters overlapped the IOZs, and in group C, the single or multiple cluster did not overlap the IOZ (two patients).

Surgical resections

Of 14 patients with single clusters, seven with single focal clusters in orientations of six uniform and one random received focal resections; three with single focal clusters in orientations of one uniform and two random underwent extensive resections; and four with single extensive clusters all in random orientation also had extensive resections.

Five of six patients with multiple clusters in orientations of one with all clusters uniform, two with all clusters random, and two with both random and uniform clusters underwent extensive resections. One patient (patient 19) in group C received a focal resection of the IOZ alone; tissue in the area of the clusters (random orientation) was not removed.

Postoperative MRI findings

One (patient 5) of 14 patients with single clusters and four of six patients with multiple clusters had residual tissue in cluster areas on postoperative MRI. In three (patients 15, 18, and 20), we performed extensive resection for only the initial IOZ and intentionally left the propagated IOZs. SDEEG had demonstrated long-latency ictal propagations between IOZs in patients 15 and 18; patient 20 had a rapid and diffuse spread of seizures from the parietal IOZ, with consistently related ipsilateral frontal and contralateral parietal ECD zones, which were not removed.

Two patients (patients 4 and 5) had residual tissue in the medial or basal frontal IOZ.

Pathology findings

Twelve patients had cortical dysplasia; two had tuberous sclerosis; three had scar tissue or gliosis, including one with additional arteriovenous malformation; one had a dysembryoplastic neuroepithelial tumor; one had a porencephalic cyst; and one patient had normal tissue.

Seizure outcomes

After a mean postsurgical follow-up period of 38 months (range, 23–51 months), 11 (55%) patients were Engel class I, three (15%) were class II, and six (30%) were class III, so that 11 (55%) patients had seizure-free outcomes, and nine (45%) had improved seizure control.

Single clusters

Table 2 correlates MEG clusters and MRI findings with IOZs and surgical outcomes.

Table 2. Correlations among MEG ECD clusters, MRI findings, ictal onset zones (IOZs), and surgical outcomes
#/Type of MEG clusters:14 Single6 Multiple
#/Type of MRI lesions:Single 9Normal 3Multifocal 2TotalSingle 4Normal 1Multifocal 1Total
  1. A, ECD zone coincides IOZ; B, ECD partially overlaps IOZ; C, ECD does not overlap IOZ; *, p < 0.05.

Correlation between MEG clusters and IOZs:A612 9 9 (65%)
B314(B1) 5 (35%)3115(B2)6 (100%)*
C1 1 11 
Surgical outcome (Engel's classification):I52121010 (71%)111 (17%) 
II1 14 (29%)1125 (83%)*
III111 3 1113 

In the 14 patients with single clusters, MRI showed a single lesion in nine patients, normal findings in three, and multiple lesions in two. Nine patients were in group A, four were in group B1, and one (patient 5) was in group C.

Seizure-free outcome was obtained in 10 (group A, eight; group B1, two) of 14 (71%) patients. Improved seizure control was seen in one group A patient (patient 7), two group B1 patients (patients 4 and 14), and one group C patient (patient 5). Postoperative MRI revealed residual tissue in the deep basal frontal IOZ of patient 4, although tissue in the cluster area had been resected. Patient 14 underwent complete resection of the IOZ overlapped by the cluster, yet seizures continued. For patient 5, SDEEG recorded consistent interictal discharges at only one electrode in the cingulate gyrus, but MEG showed ECDs in the anterior mediobasal temporal area. Postoperative MRI revealed residual tissue in the anterior cingulate gyrus of the IOZ; tissue in the mediobasal temporal cluster area had been intentionally left because it did not correlate to the IOZ.

Multiple clusters

In the six patients with multiple clusters, MRI showed a single lesion in four patients, normal findings in one, and multiple lesions in one.

No multiple-cluster patients were in group A, five multiple-cluster patients were in group B2, and one was in group C. Patient 19 (group C) had bilateral, vertically oriented anterior mediobasal temporal clusters. The IOZ was in the posterior basal–temporal area; however, SDEEG showed no interictal discharges there. Interictal discharges were recorded only in the anterior mediobasal temporal area, which was concordant with the cluster. We reported this case elsewhere (28).

In three patients (patients 15, 16, and 20) the MEG spike trains of the multiple clusters had a consistent temporal relation. In these patients, the initial IOZ correlated with the cluster of initial MEG spikes in the consecutive spikes (Fig. 2).

Figure 2.

The relation between the multiple clusters in patient 16. A 3-D view of the magnetoencephalography (MEG) coil placements shows selected left temporal and occipital coil positions (shaded squares, upper left). Consecutive MEG wave forms show the left occipital lobe peak (a) preceding the left temporal lobe peak (b) by 50 ms (upper middle). Equivalent current dipole locations superimposed on MRI show two clusters in left occipital lobe with random orientations (a) and in the left basal temporal region with random orientations (b) (upper right). Scalp radiograph shows subdural grid placement over the left temporooccipital regions (middle left). Ictal subdural EEG shows high-amplitude repetitive spikes over the left occipital regions at the seizure onset (circles and triangles) followed by low-amplitude fast discharges over the left basal temporal region (squares), within 2 s (middle right). Extensive cortical resections of both left occipital and temporal regions corresponding to two clusters are shown (lower). The patient had recurrent seizures 6 months after the surgery.

One (17%) of six patients with multiple clusters experienced a seizure-free outcome (patient 18, group B2). The remaining five patients, including three with unilateral dual clusters and two with bilateral clusters, had residual seizures; four were in group B2, and one was in group C.

Correlation between MEG clusters, IOZs, MRI findings, and seizure outcomes

Correlations between MEG single clusters and IOZs were statistically significant. More single clusters (nine of 14) than multiple clusters (none of six) coincided with the IOZ (p = 0.014).

Correlations between MEG clusters and postsurgical outcomes showed statistical significance: more patients with single clusters (10 of 14) than patients with multiple clusters (one of six) had seizure-free outcomes (p = 0.049). Correlations between MEG cluster groups and seizure-free outcomes also had statistical significance: eight of nine patients in group A achieved seizure-free outcome versus three of 11 in groups B and C (p = 0.0098). Correlations between MRI findings and postsurgical outcomes were not statistically significant; eight of 13 patients with single lesions, one of four with no lesions, and two of three with multifocal lesions had seizure-free outcomes.


This study is the first to analyze and characterize single and multiple clusters of MEG ECDs in patients with neocortical epilepsy and to compare the results with SDEEG and long-term postsurgical seizure control. A single cluster correlates to a single focal epileptogenic zone. Seizure-free outcome occurs when cortical resection eliminates a single cluster area that equates to an IOZ, because both involve the epileptogenic zone. We conclude that further careful study is needed to determine whether single clusters from noninvasive interictal MEG alone can predict the resection area for good seizure control in patients with intractable NE. Multiple clusters correlate to extensive and complicated IOZs and to limited seizure control. Surgical resection of only the IOZ is not sufficient to remove the entire epileptogenic zone, as suggested by residual ECDs. We conclude that when interictal MEG indicates multiple clusters, SDEEG will be necessary to define completely the extensive or multiple epileptogenic zones for seizure control.

Implications of a single cluster

The single cluster reflected the existence of a discrete local epileptic neuronal group capable of generating seizures. When, as indicated by postoperative MRIs, the resection area included the area of MEG localizations, good seizure control occurred (29). Similarly, in pediatric patients with lesional extrahippocampal epilepsy, the epileptogenic zone correlated with focal and asymmetrical ECD distributions (30). Our current results for patients with a single cluster were similar to those of previous studies that showed good seizure control in patients with focal spikes but without residual spikes on SDEEG (17,18,20,22,23). Pediatric MEG studies using an intraoperative navigation system reported that clustered ECDs indicated a primary epileptic zone requiring complete surgical excision (31). When clusters of ECDs overlapped unresectable eloquent areas, the authors found limited seizure control because of the residual clusters. Because most seizure-free outcomes occurred in our patients who had complete resection of a single cluster, the extent of the single cluster likely included the entire epileptogenic zone. Although 13 patients had single focal lesions and, in these cases, it would be unusual for the IOZ from SDEEG not to arise in and around these lesions, the unique value of MEG in defining an onset is that MEG is noninvasive.

Previous studies have correlated good postsurgical results with the LAFD ictal-onset pattern, because in this pattern, discrete small neuronal populations produced seizures (32–34). The LAFD ictal-onset pattern occurred more frequently in patients with a single cluster and seizure-free outcome than in those with the repetitive spike-and-waves of multiple clusters. Our finding was consistent with the concept that interictal activity producing a single cluster is related to a discrete epileptogenic zone.

In the subset of NE patients with lesions on MRI, complete resection of a single cluster adjacent to the lesion may provide seizure remission without invasive studies. In our series, however, two of three patients with a nonlesional single cluster and ictal onset undetermined by scalp video EEG had limited seizure control. In the absence of a lesion, patients with a single cluster require careful review of their seizure semiology and scalp EEG ictal findings. If invasive studies cover the entire cluster area to define the epileptogenic zone for complete resection, the nonlesional epilepsy patient might achieve a seizure-free outcome. In the subset of patients without lesions, MEG ECDs may be as informative as they are in patients with lesions; however, patients without epileptogenic lesions need careful reviews of their entire epileptic networks.

Orientation of ECDs

Randomly oriented ECDs in either single or multiple clusters require extensive cortical resection. When ECDs show random orientations, different neuronal populations may underlie each interictal discharge, especially in cases with intractable epilepsy (35). To detect a magnetic field outside the skull, MEG discharges require synchronized neuronal activity involving a certain extent of epileptic cortex (36,37), and the ECD method defines the center of the synchronized neuronal activity during the discharge. When cortical activity arises from various sites within a set neuronal population, the activity may be too extensive to be explained by a single cluster and result in randomly oriented ECDs. These ECDs may represent the dissimilar, irregularly, and partially synchronized neuronal activities of a rather extensive epileptogenic zone.

Implications of multiple clusters

Multiple clusters correlating with extensive and complicated IOZs indicate extensive or multiple, primary or potential epileptogenic zones. Residual ECDs may relate to the residual primary epileptogenic zone even after the IOZ is removed. When interictal MEG indicates multiple clusters, SDEEG will be necessary to define completely the complete epileptic network.

A current concept in epilepsy is that the epileptogenic zone represents a network often involving different interconnected regions rather than a discrete focus (3,38,39). Multiple clusters of interictal MEG discharges derived from analyses of interictal activities might be considered as part of the dynamics of epileptic activity rather than as only source localizations. The multiple clusters in remote cortical areas can represent distant activities from a large epileptic network that generates seizures. Ictal-onset behaviors without LAFD, mostly repetitive spikes in patients with multiple clusters, imply epileptic discharges through a large epileptic network of structurally and functionally abnormal connections (33).

The consistent interictal propagation pattern within consecutive spikes among multiple clusters seen in patients 15, 16, and 20 reflected the ictal propagation on SDEEG. Similar propagations within the epileptogenic zone were reported from interictal MEG studies (25,40–42). In our intractable neocortical epilepsy series, 14 (70%) of 20 patients had cortical dysplasia and tuberous sclerosis. Studies reported that focal cortical dysplasias were highly and intrinsically epileptogenic, exhibiting specific ictal fast activity or interictal continuous repetitive and rhythmic spikes and polyspikes on EEG and ECoG (43,44). Clustered MEG spike sources represented focal cortical dysplasia as occurring within and extending from MRI lesions (44). Spatial and temporal information from interictal ECDs could not determine interrelations within the extensive focal cortical dysplasia and among multiple cortical tubers (45).

Only when seizures recurred after surgically sparing ECD regions did we recognize that residual ECDs must represent part of extensive or multiple epileptogenic zones. Among the patients with multiple IOZs on SDEEG, only patient 18, with a long latency of ictal propagation, had a seizure-free outcome. This finding implies that the initial IOZ governed the propagated IOZ and was the single primary epileptogenic zone. In this case, a second cluster correlating to the propagated IOZs did not require removal because it represented a potential epileptogenic zone. Interictal MEG studies alone, however, could not differentiate a primary epileptogenic zone or a potential epileptogenic zone among multiple clusters. In our other multiple-cluster patients, ictal SDEEG could show the temporal course and define the interrelation among IOZs but did not identify the complete extent of the epileptogenic zones. Even when multiple clusters correlated to multiple IOZs, the extent of multiple clusters only partially represented the entire epileptic network.

The exact role of multiple clusters in an epileptic network is still unclear. The present method of single-source modeling of interictal MEG discharges does not allow the derivation of these dynamic relations. In patients with multiple clusters, we still have to use SDEEG to carefully demarcate the margins of large or multiple epileptogenic zones. Extensive resection of multiple epileptogenic zones in multistaged operations achieves seizure control and improves the quality of life in selected patients (46). Similarly, our surgical decisions gave some seizure control to patients with multiple clusters and identical partial seizures. Patients with these characteristics can be carefully considered for epilepsy surgery to improve their seizure control.

Migration spikes in multiple clusters

The regional differences in the IOZs in patients with a single cluster and in those with multiple clusters may correlate with the migration of epileptic foci. Multiple clusters were often seen in younger patients with IOZs in temporal/parietal/occipital lobes, whereas single clusters were frequently seen in older patients with an IOZ in the frontal lobe. In children, more occipital spikes migrate anteriorly than frontal spikes migrate posteriorly (47,48). This may suggest that posteriorly dominant epilepsy can extend anteriorly to expand the epileptic network through anatomic and functional connections, whereas frontal lobe epilepsy less frequently migrates to other lobes. Therefore multiple clusters associated with the posterior epileptic network may require extensive resection, especially in young patients. Conversely, the single cluster that correlated with a discrete anterior epileptic region in relatively old patients may predict a successful focal resection.


Acknowledgment:  This study was partially supported by ELEKTA KK, Tokyo, Japan, and the Uehara Memorial Foundation. We thank Kuniko Tsuchiya, Clinical Technology, Nishi-Niigata Chuo National Hospital, Toru Kondo, SUM Ltd, Niigata, Katsumi Takahashi, ELEKTA KK, Tokyo, Japan, for technical assistances, and Mrs. Carol L. Squires and Dr. Ayako Ochi for their editorial assistance. The histologic diagnoses were made by Drs. Hitoshi Takahashi and Akiyoshi Kakita, Brain Research Institute, Niigata University, Niigata, Japan. Elekta K.K. Tokyo provided support for software analysis. The Uehara Memorial Foundation provided financial support.