Magnetoencephalography Is More Successful for Screening and Localizing Frontal Lobe Epilepsy than Electroencephalography


Address correspondence and reprint requests to Pauly Ossenblok, PhD, P. O. Box 61, 5590 AB Heeze, The Netherlands. E-mail:


Purpose: The diagnosis of frontal lobe epilepsy may be compounded by poor electroclinical localization, due to distributed or rapidly propagating epileptiform activity. This study aimed at developing optimal procedures for localizing interictal epileptiform discharges (IEDs) of patients with localization related epilepsy in the frontal lobe. To this end the localization results obtained for magnetoencephalography (MEG) and electroencephalography (EEG) were compared systematically using automated analysis procedures.

Methods: Simultaneous recording of interictal EEG and MEG was successful for 18 out of the 24 patients studied. Visual inspection of these recordings revealed IEDs with varying morphology and topography. Cluster analysis was used to classify these discharges on the basis of their spatial distribution followed by equivalent dipole analysis of the cluster averages. The locations of the equivalent dipoles were compared with the location of the epileptogenic lesions of the patient or, if these were not visible at MRI with the location of the interictal onset zones identified by subdural electroencephalography.

Results: Generally IEDs were more abundantly in MEG than in the EEG recordings. Furthermore, the duration of the MEG spikes, measured from the onset till the spike maximum, was in most patients shorter than the EEG spikes. In most patients, distinct spike subpopulations were found with clearly different topographical field maps. Cluster analysis of MEG spikes followed by dipole localization was successful (n = 14) for twice as many patients as for EEG source analysis (n = 7), indicating that the localizability of interictal MEG is much better than of interictal EEG.

Conclusions: The automated procedures developed in this study provide a fast screening method for identifying the distinct categories of spikes and the brain areas responsible for these spikes. The results show that MEG spike yield and localization is superior compared with EEG. This finding is of importance for the diagnosis and preoperative evaluation of patients with frontal lobe epilepsy.

Patients with frontal lobe epilepsy (FLE) comprise about 20% of the group of patients with localization related epilepsy. For mesiotemporal lobe epilepsy, it has been reported (e.g., by Boon and D'Havé, 1995; Gilliam et al., 1997; Schultz et al., 2000; Sakamoto, 2004; Baldauf et al., 2006) that electroencephalography (EEG) studies including dipole localization and MRI hippocampal studies provide satisfactory localizing information in almost all cases. The causes of FLE are, however, more diverse, while the epileptiform activity in the EEG of these patients may show a more diffuse, less dipolar voltage field pattern because of the involvement of extended brain regions (Salanova et al., 1993), or the rapid propagation within the neocortex to deep structures of the brain (Laskowitz et al., 1995). Therefore, the electroclinical localization of FLE is thought to be poor, although this finding may depend on the specific areas of the frontal lobe from which the epilepsy originates (Kotagal and Arunkumar, 1998; Kellinghaus and Luders, 2004; Vadlamudi et al., 2004).

It has been shown that source analysis of interictal EEG spikes may be helpful for localizing the brain area that is responsible for FLE (Ossenblok et al., 1999; Stefan et al., 2000; Ochi et al., 2001). However, distributed or rapidly propagating epileptiform activity may lead to numerous morphologically distinct epileptiform EEG discharges with maximal amplitude at variable surface electrode locations (Engel, 1993; Lantz et al., 1998; Shiraishi et al., 2005). Therefore, the population of interictal spikes should be grouped into distinct categories before averaging and source reconstruction. Because visual inspection of spikes in prolonged high resolution recordings can become highly complex, in this study, we developed automated procedures in order to group the consecutive interictal epileptiform discharges (IEDs) occurring in these recordings.

A number of studies (e.g. Knowlton et al., 1997; Barkley, 2004; Baumgartner, 2004; Park et al., 2004) reported that magnetoencephalography (MEG) is superior to EEG with respect to sensitivity of spike detection and localization in case of extratemporal lobe epilepsy. However, until now only a few studies directly compared source localizations obtained from MEG and EEG in patients with extratemporal lobe epilepsy. These studies included either a small number of patients (Nakasato et al., 1994; Merlet et al., 1997; Shiraishi et al., 2001; Yoshinaga et al., 2002) or were dealing with presurgical assessment (Pataraia et al., 2002; Stefan et al., 2003). We compared the relative strengths of MEG and EEG of patients with a localization related epilepsy with a frontal onset, while dealing with clinically relevant sources of interictal epileptiform events.

We used MEG in addition to EEG because it is generally assumed that solutions to the inverse problem are more accurate with MEG. This is primarily because magnetic fields, compared with electric potentials, are less attenuated or distorted by intervening tissue layers between the brain and recording sensors (Hämäläinen and Sarvas, 1989). Furthermore, it is typical for patients with FLE that the epileptiform activity propagates rapidly from a localized area, related e.g. to an epileptogenic lesion, to distant and more deep lying structures of the brain (Ossenblok et al., 1999). MEG is less sensitive to activity of deep lying sources then EEG (De Jongh et al., 2005). This typical feature of MEG may result in MEG spikes clearly distinguishable from ongoing background activities, whereas the corresponding EEG spike onset may be masked by the activity due to propagation. Thus, suppressing activity from deeper structures may provide more discernable MEG spikes and more discrete localization of cortical phenomena. In this article, we describe the results of the comparison of spike sensitivity and localizability of the underlying sources of the interictal MEG and EEG of patients with FLE, using automated methods for focus localization.



A group of 24 patients with localization related partial epilepsy who had a mean age of 29 years (range, 7–59) participated in the study. All patients had experienced standard EEG/CCTV monitoring before their participation in this study and underwent for diagnostic purposes clinical MRI. The selected patients did meet the following criteria: the patient suffered of medically intractable frontal lobe epilepsy and it was assumed—on basis of EEG/CCTV monitoring—that the patient had an epileptogenic lesion. Patients who did not fulfill the last criterion, but were candidates for preoperative subdural grid recordings also were included in this study. Furthermore, EEG/CCTV monitoring indicated sufficient IEDs and the patients had the mental and physical ability to participate in the simultaneous MEG and EEG (M/EEG) recordings. Exclusion criteria were: EEG/CCTV monitoring indicating multifocal epilepsy, history of epilepsy unrelated to the frontal lobe lesion, history of severe mental disease on (probably) prior brain damage predisposing the patient to epilepsy and the presence of a pacemaker or intracranial metals. Despite the precautions regarding intracranial metals 3 out of the 24 patients studied had large artifacts in their MEG, probably due to intracranial metals left behind after an earlier operation.

The electro clinical studies indicated for 22 of the patients studied that their epilepsy was related to a clearly identifiable lesion at MRI, as described in Table 1. In patient B, there was suspicion of a cortical dysplasia in the right frontal lobe. Patient H had MR-abnormalities suggestive of a right hemispheric mesial temporal sclerosis and a lesion in the parietal cortex due to an earlier stroke. However, the semeiology of the seizures was suggestive of a right frontal lobe seizure onset. Patients were on maintenance doses of their habitual antiepileptic drugs as prescribed.

Table 1.  Patient characteristics and automated analysis results
PatientEpileptogenic lesionEEG focus (cm)MEG focus (cm)
  1. Patient characteristics: location of the epileptogenic lesion of 22 of the 24 patients studied, EEG and/or MEG cluster located closest to the lesion and the distance to the lesion (in cm) or located in the same lobe as the irritative zone identified by ECoG (RF is right frontal, LF is left frontal, RT is right temporal, LFP is left fronto parietal, LP is left parietal). The symbol (–) indicates whether the number of spikes occurring in the EEG or MEG was less than 6 spikes per hour. The symbol X (column 4) indicates that there were no MEG data available. Plain fields: for 2 of the patients (B and H) the epilepsy was not related to a visible lesion at MRI (column 2). Spike clustering and localization of the EEG failed for 10 of the18 patients for whom interictal EEG and MEG was obtained successfully (column three) and for MEG for 3 of these patients (column 4).

1/ALF: dysplasia, left frontal parasagittal(–) 
2/B RF
3/CLF: postoperative defect, left fronto-polarLFX
4/DRF: postoperative defect, right fronto-basal(–)(–)
5/ELF: dysplasia, left frontal (subcortical)(–)LF(1.72)
6/FLF: trauma capitis, left frontalLF(1.64)LF(2.00)
7/GLF: postoperative defect, left frontal 
8/H RT
9/IRF: DNET, right fronto-central area RF(4.10)
10/JLFP: parenchyma defect, left fronto-parietalLFP (1.28)LFP (1.03)
11/KLF: hyperintensity, left fronto-basal(–)(–)
12/LLF: postoperative defect, left frontal X
13/MLF: postoperative defect, right frontal X
14/NRF: hamartoma, right fronto-basal RF(0)
15/OLF: intraparenchymae haematoma, left frontalLF (1.78)LF (1.84)
16/PLF: subcortical hyperdensity, left frontalLF (0.36)LF (1.90)
17/QLF: hyperdensity, left frontal parasagittal area(–)(–)
18/RLF: DNET, left frontal parasagittalLF(0.03)LP
19/SLF: local atrophy, left frontal LF (1.76)
20/TLF: dysplasia or tumor, left frontal LF (0.35)
21/ULF: postoperative defect and gliosis, left frontalLF (0.92)LF (0.45)
22/VRF: low graded angioma, gyrus rectusRF (0.80) 
23/WLF: low graded angioma, left frontal LF (0.69)
24/XRF: dysplasia, right frontalRF (0.26)RF (0.80)

This study had the approval of the Medical Ethics Committee of Epilepsy Center Kempenhaeghe and all patients gave informed consent. The patients included in this study were patients of Epilepsy Centre Kempenhaeghe or were referred via the Dutch Collaborative Epilepsy Surgery Program.

M/EEG recordings

The M/EEG measurements were carried out at the Free University Medical Center of Amsterdam (VUmc) using a 151-channel whole head MEG system with a base length of 5 cm (VSM Inc., Vancouver, BC, Canada). Squid noise levels were below 10 fT/Hz. Recordings were made with third order gradiometer noise cancellation. The antialiasing filter of the CTF hardware system was set to 200 Hz and the data were sampled at 625 Hz. Up to 71 EEG electrodes were placed on the scalp according to the international 10% system. Recordings were made with a common reference at CPz and the electrode impedances were kept below 5 kohm.

Patients were asked to keep themselves relatively deprived of sleep in the night before the measurement. During the measurements the patients were lying in a supine position with eyes closed and were allowed to fall asleep. The data were collected in consecutive files of 5, 10, or 15 min depending whether the patient was awake, drowsy or asleep. The mean value of the duration of M/EEG recorded effectively was 81 min (range, 30–115 min), with three outliers in the range of 30–60 min for three of the children younger than 12 years. Before and after each file the position and orientation of the patients head relative to the MEG sensors was measured using head coils located at three anatomical landmarks. The landmarks are the impressions before the left and right ear (preauricular points) and the impression above the nose (nasion). The electrode positions were derived relative to these landmarks by deploying a MEG device as a 3-dimensional (3D) digitizer (De Munck et al., 2001).

MR acquisition

A whole head scan, using a 3D T1-weighted spoiled gradient echo (FLASH) sequence (TR 11.8 ms, TE 5 ms, flip angle 30 degrees, 2 signals averaged, 1 mm in-plane resolution, FOV 256mm) producing 2 mm sagittal partitions, was performed at the VUmc. Vitamin E capsules, that are visible on MR scans, were placed at the same landmarks as the head coils. The positions of the sets of head coils and vitamin E capsules defined the matching between the M/EEG and MRI coordinate systems.

Semiautomatic spike detection

The initial selection of IEDs was made by a computer based algorithm (Persyst Spike Detector; Persyst Development Corporation, Prescott, AZ, U.S.A.), published by Wilson et al. (1996). The IEDs in both EEG and MEG were independently reviewed visually by an EEG technician and confirmed by a clinical neurophysiologist (A.C.) who made the final selection. Two types of IEDs occurring in EEG or MEG were identified: spikes and spike-and-wave discharges. Sharp signals (duration < 200 ms) clearly distinguishable from ongoing background activities, seen on at least three to five nearby channels of the individual region, were selected and then regarded as EEG or MEG spikes. Sharp signals suspected of clear contribution from heart beats, eye movements, physiological rhythmic discharges, or vertex sharp activities during sleep were rejected. The spike-and-wave discharges are spikes followed by a slow wave of variable duration. Magnetic spikes may differ in morphology from electrical ones, but because no standardized definition of magnetic spikes currently exists (Zijlmans et al., 2002), the EEG criteria were used for both MEG and EEG spikes.

After trend-removal the signals were band-pass filtered between 1 and 70 Hz and the data in 1 s time windows around the selected maximum of the spike (spike marker) were extracted and merged into a new file. The MEG data in this new file were interpolated to an MEG grid corresponding to the average head position across the entire recording session (De Munck et al., 2001). The EEG recordings were converted to average reference. The IEDs in either EEG or MEG were reevaluated visually and if clear dipole patterns were identified at the maximum of the spike, markers were placed at the onset of the spike and at the spike maximum.

Source analysis of spike clusters

Clustering of large numbers of spikes recorded from a large number of channels on the basis of visual inspection is highly complex, because the distinction between spike categories is not always well-defined. Therefore, an automated clustering procedure was developed in order to group the consecutive spikes in prolonged magnetoelectric recordings before averaging and source reconstruction (Van't Ent et al., 2003). A computer based cluster algorithm was used to align and cluster the spikes and to visualize the data focused around the marker. First, the spike markers in the merged new file were outlined on the spike maximum. After the realignment step, M/EEG data in epochs of 17.6 ms (11 samples given a 625 sample rate) centered on the spike markers were stored and were put in a hierarchical clustering algorithm. The agglomerative grouping can be visualized as a hierarchical dendrogram based on a table of Eucledian distances representing the similarity of the field distributions of all pairs of spikes (Guess and Wilson, 2002).

Spike averages were computed for each spike group and underlying sources of the averaged magnetic and electric fields were reconstructed using a single equivalent current dipole model. A dipole was estimated at each individual time sample during the same time window as used in the clustering procedure. That is, across a 17.6 ms epoch centered on the peak of each epileptic spike. The goodness of dipole fit (%err) is an estimate of the difference between the recorded and dipole field. A realistic Boundary Element Model was used to describe the volume conductor. This model, consisting of three compartments (brain, skull and skin), was derived from the individual's MRI, using a segmentation method described in Van't Ent et al. (2001). The computed dipole locations are presented as dipole density plots, with the maximal density (in red) coinciding with the slices shown.

Evaluation of the source analysis results

The results of the automated localization procedures were evaluated relative to the epileptogenic lesion. In cases that such lesion was not clearly visible at MRI (for patient B and H), the location of the dipole cluster was compared to the interictal onset zone (irritative zone) identified preoperatively by subdural electroencephalography. When multiple dipole clusters were found, comparison was based on that dipole cluster, which was located in the same lobe and closest to this lesion or in the same lobe as the irritative zone identified by electrocorticography (ECoG). In case of a lesion, the localization accuracy of the MEG versus EEG spike clusters was expressed in terms of the distance of the selected dipole cluster to the lesion. This distance (in cm) was defined as the median distance of the dipoles within a cluster to the nearest border of the lesion. The insert in Fig. 3 schematically shows the method used for obtaining the distance of the centre of gravity of the dipole cluster to the lesion.

Figure 3.

Topographic maps of the average magnetic fields for the three spike clusters of patient W and the channel overlay plots. At the top of each panel the cluster number and, within brackets, the amount of spikes is given. The channel overlays are shown during a time window from −100 to 100 ms around the spike marker, indicated by the dashed vertical red line. The maps show the magnetic field distributions at the time of the marker (t = 0 ms). The arrows indicate the epoch centered around the maximum of the spike subaverages used for equivalent dipole fitting. The anatomical locations of the equivalent current dipole estimates of spike clusters 1–3 are shown as density plots on sagittal and axial MR scans. If present in the MR slices shown, the structural lesion in the left frontal lobe is delineated by a white dotted line. The insert (middle right) shows a schematic drawing of a surface reconstruction of a lesion and the shortest distances of the dipoles (red dots) to the lesion as black arrows. The white arrow gives the median distance of the lesion to the cluster of dipoles.


Patient material

In 22 of the 24 patients studied a clearly identifiable lesion at MRI was found (see Table 1, column 2). In 3 patients, no MEG data were available, because of large artifacts due to intracranial metal fragments left after an earlier operation. Because we were not able to compare in these cases the interictal MEG and EEG, the M/EEG data of these patients were excluded from further analysis. The interictal M/EEG data were recorded successfully with more than six spikes per hour in either MEG or EEG in the remaining 18 of the 24 patients studied.

Visual review of the M/EEG datasets

Large and unexpected differences were observed in both the morphology and topography of the IEDs occurring in MEG and EEG. Therefore, a systematic comparison between the appearance of interictal EEG and MEG spikes (number and duration) was made in the successfully recorded M/EEG data sets. It is apparent from Fig. 1 that the spike yield is larger in MEG than in EEG for the majority of the patients studied. The number of EEG spikes exceeds the number of MEG spikes (X > Y) in only 5 of the 18 patients. Furthermore, when sharpness is defined by the slope of the spike, measured for each of the patients studied from the onset until the spike maximum (Fig. 2, upper left), a tendency of the MEG signal to look sharper was observed in many of the patients (Fig. 2, bottom). The histograms (Fig. 2, upper right) show that the distributions of the duration of the total number of EEG and MEG spikes differ significantly (Wilcoxon rank sum test; p < 0.05).

Figure 1.

Results of the MEG versus EEG spikes for 18 patients (A–X) with successful recording of MEG and EEG and at least 6 spikes per hour occurring either in MEG or EEG. The dashed line represents the line were the number of MEG spikes equals the number of EEG spikes.

Figure 2.

The duration of the spikes was measured from the onset till the spike maximum (insert upper left). The normalized distributions of the duration of the total number of EEG (402) and of the total number of MEG (567) spikes are shown at the insert in the upper right corner of the panel. The mean and standard deviation of the duration estimated for the EEG (open circles) and MEG (filled triangles) spikes are given for the 18 patients studied (bottom).

Automated source analysis results

Cluster analysis was used to classify the spikes of the same 18 patients on the basis of their spatial distribution, followed by equivalent dipole analysis of the cluster averages. We found for most of the patients studied distinct spike subpopulations with clearly different topographical field maps. For example, cluster analysis of the 57 MEG spikes of patient W yielded four subclasses of events. The MEG signals of the smallest subdivision with three spikes were very noisy and were discarded as outliers during the cluster analysis procedure. The average magnetic field distributions of the three spike clusters left are given in Fig. 3, together with the subaverages of the spikes. The equivalent dipoles, fitted at an epoch across the maximum of these subaverages, are plotted at the sagittal and axial MR scans of the patient (Fig. 3, bottom). The equivalent dipoles fitted for spike cluster 1 and 3 (% err < 10) are bordering the left frontal lesion of the patient. The averaged magnetic field distribution of spike cluster 2 is more bilateral, while the location of the equivalent dipoles fitted at the maximum of this spike cluster (% err < 15) may indicate propagation towards the area of the cingulum. The average potential distributions (Fig. 4, left) of all 3 EEG spike clusters obtained for 34 EEG spikes were widely distributed and corresponded to spurious locations of the equivalent dipoles in the central regions of the brain (Fig. 4, right).

Figure 4.

The topographic field maps of the potential fields (left) of the three EEG spike clusters of patient W. The cluster number, number of spikes per cluster and the mean Euclidian distance are shown in the insert at the lower right corner of the panel. The equivalent dipoles obtained for the three spike clusters are plotted together in a sagittal and axial MR scan of the patient (right).

The location of the epileptic focus

In Table 1 the locations of the selected dipole clusters of EEG (EEG focus) and MEG (MEG focus) are described. For MEG spike clustering and localization failed for 3 of the patients studied (A, G and V), probably because of the high variability of the single spikes within one cluster. The variability of the MEG spikes of patient V was large, because during the recordings a technical problem (the helium level became too low) resulted in a substantial increase in noise. In patients A and G, strong background activity could be responsible for the low signal-to-noise ratio (SNR). For each of the other 15 patients studied, the main dipole cluster was located in the same lobe as the epileptogenic lesion or in the same lobe as the irritative zone identified by ECoG, except for patient R. Preoperative ECoG identified the brain area responsible for the epilepsy of two of the patients (B and H). Although there was a suspicion for patient B of a right frontal cortical dysplasia, MR studies failed to identify a clear-cut structural lesion that possibly could be responsible for the epilepsy. Preoperative subdural grid recordings yielded evidence that the ictal onset zone was located in the SMA of the right hemisphere of this patient. This location was considered to be too close to the primary motor area to allow a complete resection without causing neurological deficits (Ossenblok et al., 2003). The MEG localizations of patient H have previously been discussed in the context of the electro clinical examinations of this patient by Van't Ent et al. (2003). These authors showed that the 2 plausible MEG spike clusters corresponded to equivalent dipole locations in the right temporal lobe and the right frontobasal area. Preoperative subdural grid recording indicated an irritative zone located in the right temporal lobe of this patient. The EEG spike clusters, on the other hand, obtained for 62 interictal epileptiform events all pointed at an implausible localization at an area bordering temporally the fissura silvii. For patient R, the MEG spike cluster was located in the left parietal lobe, far from the lesion, and the EEG spike cluster was located at the borders of the left frontal parasagittal lesion. Although the subdural grid recordings of this patient supported the MEG localization, the patient was rejected for resective surgery because of these and other in congruencies.

In conclusion, 2 of the 18 patients had an insufficient number of EEG spikes (A and E), for 1 patient with a left frontal lesion the spike cluster was located in the right frontal lobe (G), whereas for another 7 patients, spike clustering failed only for the EEG and not for the MEG. Thus, automatic source analysis of the MEG spikes was successful in 14 of the 18 patients studied (77.8%) and of the EEG in 7 of these patients (38, 9%), whereas the results were inconclusive in patient R.

The accuracy of source reconstruction

The epilepsy in most of the patients was likely to be related to a clearly identifiable lesion at MRI. Therefore, this study enabled us to compare systematically the localization accuracy of the M/EEG spike clusters located within the epileptogenic region in terms of the distance of the irritative zone to the border of the structural lesion. However, we did not find a systematic difference between the distances of the selected MEG and EEG dipole clusters (i.e., the MEG vs. EEG focus) relative to the lesion. For some of the patients, the centre of the selected dipole cluster was located close to the lesion or even within the lesion (patient N) and for others distant to the lesion (Table 1). For example, patient I, who had a dysembryoplastic neuroepithelial tumor (DNET), the median distance of the dipole cluster to the nearest border of the lesion amounted to 4.1 cm. In this study, 8 of the patients had post traumatic (patients F, J) or postsurgical holes in their skulls (patients C, D, G, L, M, U). We found for one of these patients (G) clearly distinguishable EEG spikes with discrete localization, however, in the right instead of the left frontal lobe were the lesion was located. For 3 other patients with a hole in the skull (F, J, and U), the differences in distance to the lesion of MEG dipoles versus EEG dipoles were less than 0.5 cm.


The primary aim of this study was to evaluate whether MEG yields additional information compared with EEG for identifying clinically relevant sources of epileptiform events of patients with FLE. To this end, we compared systematically the MEG and EEG datasets consisting of expert-selected and confirmed IEDs of 18 well-documented patients with FLE. It was found that MEG spikes are far more often present than EEG spikes and generally have a sharper appearance. The higher spike yield of MEG compared with EEG is in accordance with a model that predicts higher SNRs for MEG than EEG in the frontal lobes, thus explaining that MEG spikes appear to be more distinguishable from the background activity than EEG spikes (De Jongh et al., 2005). The same model predicted similar spike yields for MEG and EEG in patients with temporal lobe epilepsy, which is consistent with the findings of Leijten et al. (2003) and Iwisaki et al. (2005) for patients with temporal lobe epilepsy. These authors regarded the relative insensitivity of MEG to deep lying sources as a disadvantage, especially because epileptiform activity generated in mesial temporal lobe regions might not be recorded by MEG. That MEG shows a significantly higher sensitivity than EEG for a patient with lateral frontal lobe epilepsy, but not for a patient with basal temporal epilepsy was also previously shown by Oishi et al. (2002), by comparing the interictal EEG and MEG spikes with spikes in the electrocorticogram. Furthermore, in a study of Fernandes et al. (2005) a computer-based algorithm was applied to extract parameters that could be used to quantitatively describe the morphology of the epileptiform transients. It appeared that EEG and MEG coincident spike events were statistically different with respect to several morphologic characteristics, such as duration, sharpness, and shape. These topographic and morphologic differences are the consequence of volume propagation through the tissues with different conductivities that surround the brain and affect EEG but not MEG, and of the difference in sensitivity of MEG and EEG to the orientation of the underlying dipolar sources.

The localizability of MEG versus EEG

The comparison of the MEG and EEG cluster analysis results followed by dipole localization raises the question why EEG based analysis failed for twice as many patients as MEG based analysis. For some of the patients the interspike variability was too high for clustering and reliable localization. However, diffuse field distributions due to overlap of primary (onset) and secondary source activity may be another reason for failure of source localization of EEG spikes, e.g., for patient H. Equivalent dipole fitting at an epoch centered around the maximum of each single interictal EEG event of this patient followed by clustering of these dipoles resulted in a single widely distributed cluster of dipoles covering the right temporal and frontobasal area (Fig. 5, bottom right). Cluster analysis of the equivalent dipoles fitted at an epoch centered around the maximum of each single MEG spike of this patient indicated, in accordance to the spatiotemporal cluster analysis results published by Van't Ent et al. (2003), 2 plausible clusters of dipoles, located at the right temporal lobe (red dots), and a second cluster located at the right frontobasal area (dark brown dots) (Fig. 5, upper right). Thus, the interictal MEG of patient H enabled the discrimination between clearly delineated primary and secondary epileptic active areas, probably as a direct result of the MEG feature being less sensitive to distributed and deep lying sources. The EEG sharp waves of this patient, on the other hand, corresponded to a distributed area of activity, probably resulting from the rapid propagation of the epileptiform activity along the right temporal lobe to the right frontal basal area (see Fig. 5, bottom right).

Figure 5.

A single epileptiform event occurring simultaneously in the MEG (black) and EEG (red) of patient H (left). The arrows indicate the epoch centered around the maximum of the single spikes that were used for dipole fitting. The equivalent dipoles obtained for 71 single MEG and 62 single EEG epileptiform events are projected for MEG at the sagittal and axial MR scan of the patient (upper right) and for EEG at the sagittal MR scan (bottom right). The strength of the equivalent dipoles is color coded and is maximal in red.

The results of equivalent dipole fitting of single spikes, as shown in Fig. 5, indicate that one has to be careful when localizing average spikes. This finding was also reported by Chitoku et al. (2003), who showed discrepancies of single interictal spike localization and localization of average spike clusters in case of extratemporal lobe epilepsy. Equivalent dipole analysis of single spikes instead of spike subaverages is, however, in most of the cases not feasible, because of insufficient SNR of single spikes. Therefore, in this study spike clusters are localized assuming that averaging of spikes with similar spatiotemporal patterns yields a more accurate localization, because of the increased SNR of the average spike.

Accuracy of MEG versus EEG localizations

An indication for accuracy might be inferred from the site of the main dipole cluster relative to the lesion, because neuropathology is most of the time directly related to the lesion (Genow et al., 2004). Assuming that the area at the margins of the structural lesion is related to the epileptogenic zone the accuracy of EEG versus MEG localizations was estimated as the distance of the irritative zone to the border of the lesion (Table 1). We did, however, not find a systematic relationship between the distances of the EEG versus MEG localizations. We sometimes found the centre of the dipole cluster located within the lesional area, which is probably due to the widely distributed area involved in the epilepsy of the patient, while it still contained intact neuronal cells. It is well known that in case of a tumor, like for patient I, the epileptogenic lesion may be located at a remote distance of the lesion (Zaatreh et al., 2003). Furthermore, for the EEG, holes in the skull can lead to very large errors in the inverse solution (Oostenveld and Oostendorp, 2002), but not for the MEG. Although, the location errors due to underestimating skull conductivity—which is unknown for the individual patient—are typically higher than those found due to neglecting a hole in the skull (Vanrumste et al., 2000). We found for three out of four of the patients with a hole in the skull, that the differences of MEG and EEG in location relative to the lesion were falling within the localization error inherent to inverse solutions for EEG and MEG (Leahy et al., 1998).

It has been reported that the combined analysis of MEG and EEG may yield a better accuracy than a separate analysis of these modalities (Diekman et al., 1998). In practice, however, a combined analysis is hampered by the large differences that can be observed in both the morphology and topography of the MEG versus EEG spikes of our group of patients.

The diagnostic value of the localization procedures

Evaluation of epilepsy improves with yield of interictal epileptiform activity, with neurophysiologic SNRs of such activity and with the localization of the full range of interictal IEDs. In this study, we successfully applied algorithms for the characterization of the single spike events with varying morphology and topography occurring in prolonged M/EEG recordings of patients with FLE. The automated clustering and localization of these spike events indicated distinct spike clusters within a widely distributed epileptogenic region surrounding a lesion (e.g. patient W in Fig. 4) or corresponding to distinct brain areas involved during the IEDs (e.g patient H). For the latter patient, additional research—including spatiotemporal analysis of the MEG spikes, ECoG, and successful surgical intervention that rendered the patient seizure free—pointed to a primary onset zone in the right temporal lobe with secondary involvement of the right frontal lobe. Thus, the results presented here support the hypothesis that the spike clusters located within the epileptogenic region, as defined by an epileptogenic lesion or ECoG, correspond to the irritative zone, whereas others only represent the propagation of the epileptiform activity (see also Ossenblok et al., 1999). This finding is in accordance with the results of Ossadtchi et al. (2004), who reported multiple spike clusters with some of them located in the vicinity of the area that subsequently was resectioned (for three of the four patients studied), while the other remaining “spurious clusters” might correspond to secondary epileptic areas because of propagation.

The generally higher spike yield of MEG and the generally sharper profiles of the spikes is of importance for deciding whether the patient has epilepsy or not. Furthermore, because localization of MEG is possible in more patients on basis of MEG than of EEG, MEG might result in a better treatment and prognoses based on a more accurate diagnosis and localization of the epilepsy of the patient. These findings are in support of the conclusion of Baumgartner and Pataraia (2006), who stated: “In extratemporal lobe epilepsy MEG provides unique information in nonlesional cases and helps to define the relationship of epileptic activity with respect to lesions and eloquent cortex.”


Automated single equivalent dipole analysis—next to clustering—is a quick and easy way to determine the brain areas generating the IEDs. It enables the identification of the irritative zone relative to an epileptogenic lesion, if present at MRI or relative to ECoG. The higher spike yield of MEG and the sharper morphology together with a better localizability of the MEG spikes indicate that interictal MEG is superior to interictal EEG for presurgical diagnosis in groups of patients who otherwise are not considered as candidates for presurgical evaluation or let alone resective surgery.


A grant (no. 20–10) of the Dutch Epilepsy Foundation was highly appreciated to support this work. Furthermore, we would like to thank Irma Van Velzen and Petra Van Mierlo for all their work for this research project and Roy Krijn for his technical assistance.