Surgical Implications of Neuromagnetic Spike Localization in Temporal Lobe Epilepsy

Authors


Address correspondence and reprint requests to Dr. M. Iwasaki at Department of Neurosurgery, Tohoku University Graduate School of Medicine, 1-1 Seiryo-machi, Aoba-ku, Sendai 980-8574, Japan. E-mail: iwa@nsg.med.tohoku.ac.jp

Abstract

Summary:  Purpose: To investigate the clinical usefulness of magnetoencephalography (MEG) as a guide to the surgical treatment of temporal lobe epilepsy (TLE).

Methods: Preoperative spike localization by MEG was compared with seizure outcome and postoperative spike localization at 12 months after resective surgery in 16 patients with TLE. Spike localization was classified into anterior temporal (AT) and non-AT localization in 11 patients without neocortical lesion treated with anterior temporal lobectomy (ATL); and lesion and lobar localization in five patients with neocortical lesion treated with lesionectomy (n = 3) or lesionectomy with medial temporal resection (n = 2).

Results: All five patients with AT localization became seizure free and spike free after surgery. Among the six patients with non-AT localization, two became seizure free and spike free, two became seizure free with residual spikes, one had residual seizures but no spikes, and one had both residual seizures and spikes. All three patients with lesion localization and two with lobar localization had favorable seizure outcome and became spike free after surgery.

Conclusions: MEG spike localization can identify neocortical sources remote from the presumed epileptogenic area. Favorable seizure outcome can be expected in patients with AT localization after ATL and patients with lesion localization after lesionectomy. In contrast, non-AT localization indicates either nonmedial TLE or spike propagation to the posterior and extratemporal neocortex. Similarly, lobar localization indicates spike propagation from an epileptogenic lesion or extensive epileptogenicity. Patients with non-AT localization or lobar localization should undergo intensive evaluations, such as intracranial EEG, for improved seizure outcome.

Interictal spiking is important for the clinical diagnosis of partial epilepsy. The site of the interictal spikes is highly suggestive of the seizure origin, but not identical to the epileptogenic area (1). Scalp-recorded EEG has been used to investigate interictal spike propagation (2,3). Intracranial EEG has revealed that leading spikes near the epileptogenic area may propagate over various neocortical areas (4,5). Therefore, the interictal spike area is usually more extensive than the epileptogenic area (1). Nevertheless, identification of the interictal spike area is helpful for clinical decision making in the surgical treatment of epilepsy. Interictal spike localization by scalp-recorded EEG is correlated with postoperative seizure outcome in patients with temporal lobe epilepsy (TLE) (6,7). Complete resection of the interictal spike area identified by intracranial EEG is associated with a favorable seizure outcome (8).

Spike localization by magnetoencephalography (MEG) has been compared with localization by intracranial EEG (2,9–14), positron emission tomography (PET) (15) and single-photon emission computed tomography (SPECT) (9). Previous studies suggested that the localization accuracy of MEG was higher than that of conventional scalp EEG (10,14) because of the absence of disturbance by the inhomogeneous conductivity of cranial tissues. However, only a few studies have correlated the MEG spike localization with the postoperative seizure outcome (13,14,16,17). Excellent surgical outcome was achieved when spike source localization was concordant between MEG and EEG (16), and the reliability of interictal MEG localization was verified in patients with extratemporal neocortical epilepsy followed by postoperative good outcome (13). However, little is known about the correlation between MEG findings and surgical outcome in patients with TLE, although MEG may provide the basis for the diagnosis, such as medial or lateral TLE, based on the spike localization (18–20).

In the present study, interictal MEG was performed in patients with TLE both before and after surgery. Preoperative MEG spike localization was correlated with postoperative spike localization and seizure outcome.

METHODS

Patient population

Twenty-six patients underwent temporal lobe surgery for intractable TLE at the Kohnan Hospital between September 1998 and August 2000. Of the 26 patients, 24 (13 male and 11 female patients) who underwent simultaneous EEG/MEG recording during preoperative examination and were followed up for ≥12 months were selected for this study. Seizure diagnoses were based on the International League Against Epilepsy classification (21). Informed consent for this study was obtained from all participants.

The patients were aged from 7 to 40 years (mean, 23.5 years) at surgery, and the duration of epilepsy ranged from 1 to 34 years (mean, 12.2 years). Seizure symptoms included complex partial seizures in 22 patients with aura (n = 18) and/or automatism (n = 19), simple partial seizures in two, and secondarily generalized tonic–clonic seizures in six. Twelve patients had a history of febrile convulsion in childhood. Thirteen patients underwent left temporal surgery, and 11 underwent right temporal surgery.

Comprehensive preoperative evaluations were performed according to our protocol, including brain magnetic resonance imaging (MRI), continuous video-EEG monitoring, interictal PET or SPECT, and neuropsychological testing with intracarotid amobarbital injection. The MRI protocol included T1- and T2-weighted, and fluid attenuation inversion recovery (FLAIR) imaging in axial and coronal sections. The MRI diagnosis was based on routine inspection for hippocampal atrophy and other structural lesions. Hippocampal atrophy was defined on coronal sections as apparent volume reduction or high-intensity change in the FLAIR image. Continuous video-EEG monitoring was performed for 7 days to capture at least three habitual seizures, by using scalp electrodes applied according to the international 10-20 system with bilateral sphenoidal electrodes. Interictal PET or SPECT scans were reviewed by visual inspection for hypometabolic or hypoperfusion areas. Surgical treatment without long-term intracranial EEG was offered to patients showing concordant findings in all of these noninvasive evaluations. The surgical strategy could not be determined by these noninvasive evaluations in five patients, so long-term intracranial EEG was performed to identify the seizure-onset and functional areas. The indications for intracranial EEG implantation were normal MRI findings, atypical electroclinical features, or suspected epileptogenic areas near the eloquent cortex.

Patients were divided into two groups based on the structural lesion identified by preoperative MRI: patients without neocortical lesion, defined as the absence of neocortical structural lesion with/without hippocampal atrophy, and patients with neocortical lesion, defined as the presence of neocortical structural lesion without hippocampal atrophy.

Anterior temporal lobectomy with amygdalohippocampectomy (ATL) was performed in the patients without a neocortical lesion. The anterior temporal neocortex was initially resected for 3.5–4.5 cm along the middle temporal gyrus on the dominant/nondominant language side. Then the medial temporal structures were resected with the intact hippocampus, parahippocampal gyrus, and amygdala. Hippocampal resection was extended to the anterior 3.0 cm of the head. Lesionectomy was performed in the patients with a neocortical lesion. Additional resection of the anterior temporal neocortex and medial temporal structures was performed in two patients, because participation of the medial structures was anticipated in seizure generation. Histologic examination was made of all resected tissue by the same neuropathologist.

Seizure outcome was evaluated in August 2001, based on the classification described by Engel et al. (22): seizure free (class I), rare seizures (II), worthwhile improvement (III), and no worthwhile improvement (IV).

EEG/MEG recording

Interictal MEG was recorded before surgery in all patients and 12 months after surgery in 22 patients. All recordings were performed during hospitalization. Pre- and postoperative MEG was recorded under decreased and continuing antiepileptic medication (AED), respectively. Oral barbiturate was used to induce a light sleep state in 17 and eight patients before pre- and postoperative MEG, respectively.

EEG and MEG were simultaneously recorded in a magnetically shielded room. EEG electrodes were placed according to the international 10-20 system with bilateral sphenoidal electrodes or surface anterior temporal electrodes (23). Whole-head neuromagnetometer systems with 122 and 204 channels (Neuromag Ltd., Helsinki, Finland) were used in 18 and six patients before surgery, and in 21 and one patients after surgery, respectively. The MEG sensors consisted of rectangular pairs of planar-type gradiometers aligned in a helmet-shaped dewar vessel, into which the patient's head was inserted during the measurement. The details of this system are described elsewhere (24).

EEG and MEG were recorded for a continuous 30 min in drowsiness to light-sleep states. The data was bandpass filtered between 0.03 and 130 Hz and sampled at 400 Hz. The total recording was divided into five or six sets and stored in a hard disk for later offline analysis. Head coordination was acquired before each set to minimize the error in coordinate integration between MEG and MR imaging.

Spike dipole estimation

EEG and MEG data were inspected in parallel for interictal spikes. Ipsilateral ear reference montage, longitudinal, and transverse bipolar montages were used for reviewing the EEG data. MEG spikes were defined as specific when the spike amplitude was double or larger than the background activity and corresponded with the EEG spike in timing and approximate potential distribution. The earliest peak was used for source analysis when multiple peaks were prominent. Dipole source was estimated for ≥10 interictal spikes. Cases with <10 interictal spikes were considered to have inadequate sampling and were excluded from the present study.

The acquired data were low-pass filtered at 40 Hz. High-pass filtering was used at appropriate settings between 2 and 8 Hz to extract the spike component from the slower background activity. When a clear dipole pattern (i.e., a pair of in-flux and out-flux peaks of magnetic fields) was seen in the magnetic field distribution around the spike peak, the single equivalent current dipole (ECD) model was fitted in the patient's spherical head model to the recorded signals from a subset of 60 to 140 channels around the maximum peak. The least-squares minimization method was used to find the optimal solution. The spike dipole was represented by a dipole with the highest goodness-of-fit and lowest confidential volume. Dipoles with goodness-of-fit >80% were accepted as spike samples. The estimated dipoles were co-registered on the patient's MRIs by using a MEG–MRI coordinate integration system. The three-dimensional digitizer system integrated the two coordinates based on three fiduciary points, the nasion and bilateral preauricular points (25).

The results were principally assessed based on the dipole localization on the patient's MRIs. Spike localization was classified into two groups in the patients without neocortical lesion: anterior temporal (AT) localization and non-AT localization. AT localization was defined as >70% of total dipoles localized in the unilateral anterior temporal lobe, and other cases were defined as non-AT localization. In this study, the anatomic boundary between the anterior and posterior temporal lobes was defined as the point where the central sulcus reaches the sylvian fissure (26). The boundary line was defined as perpendicular to the chiasmatic–commissural line, which is known to be parallel to the temporal lobe axis (27).

Spike localization was classified according to the relation with the MR lesion in patients with neocortical lesions: lesion localization and lobar localization. Lesion localization was defined as >70% of total dipoles localized within 10 mm of the margin of MR lesion, and other cases were defined as lobar localization.

Brain MRI was performed before and after surgery in all patients for dipole co-registration. Spike dipole localization before and after surgery was compared with the seizure outcome. The ratio of preoperative spike dipoles localized inside the surgical resection was assessed on postoperative MRI.

The dipole localizations were compared between patients by using Talairach's coordinate system (28), and the mean dipole coordinates and standard deviations were calculated from the total dipole samples of each patient. The standardization was based on the method described elsewhere (29).

RESULTS

Preoperative MEG observed an adequate number of spikes in 16 patients, 11 without and five with neocortical lesions (Table 1). Preoperatively, 13 to 51 (mean, 28.4) interictal spikes were obtained for analysis. All spike dipoles were lateralized to the surgically treated side. No postoperative spike was found in both MEG and EEG in the eight patients who did not show enough MEG spikes before surgery. Of these eight patients, six patients without a neocortical lesion had hippocampal sclerosis followed by excellent seizure outcome after ATL. The other two patients had a neocortical lesion, which was cavernous angioma, followed by excellent outcome after lesionectomy.

Table 1.  The list of 16 cases with temporal lobe epilepsy
CasesSexAge
[yr.]
Epileptogenic
side
MRIMEG spike dipole
localization
SurgeryPathologyMEG
12 mo. after
surgery
Seizure
outcome
(follow-up
period
[mos.])
  • HA: hippocampal atrophy; AT: anterior temporal lobe; HS: hippocampal sclerosis; CD: cortical dysplasia.

  • CA: cavernous angioma; AVM: arterio-venous malformation; DNT: dysembryoblastic neuroepithelial tumor.

  • ATL: anterior temporal lobectomy with amygdalo-hippocampectomy.

  • *

    evaluated by invasive monitoring with grid electrodes implantation.

TLE without neocortical lesionAT dipoles
[% (n/N)]
Dipoles within
the resection
[%]
    
 1M28LtHAAT75.8 (25/33)54.5 (18/33)ATLHSno spikeI (21)
 2F10LtHAAT100.0 (20/20)100.0 (20/20)ATLHSno spikeI (28)
 3M17LtHAAT93.1 (27/29)51.7 (15/29)ATLHSno spikeI (35)
 4M22RtHAAT76.7 (23/30)70.0 (21/30)ATLHSno spikeI (30)
 5M16RtHAAT100.0 (21/21)38.1 (8/21)ATLHSno spikeI (16)
 6F25Lt*HAnon-AT4.4 (2/45)4.4 (2/45)ATLHSno spikeI (24)
 7F36LtHAnon-AT47.4 (9/19)31.6 (6/19)ATLHSno spikeI (19)
 8M17RtHAnon-AT41.2 (21/51)49.0 (25/51)ATLHSresidual spikesI (32)
 9F22Lt*normalnon-AT23.1 (6/26)19.2 (5/26)ATLgliosisresidual spikesI (27)
10M29LtHAnon-AT31.3 (10/32)21.9 (7/32)ATLHSno spikeII (27)
11M12Lt*HAnon-AT8.0 (2/25)8.0 (2/25)ATLHSresidual spikesIII (22)
            
TLE with necortical lesionLesion dipoles     
12M23Ltanterior temporal AVMLesion81.0 (17/21)14.3 (3/18)lesionectomyAVMno spikeI (23)
13M7Rtanterior temporal CDLesion100.0 (13/13)46.2 (6/13)lesionectomyCDno spikeI (24)
14M23Rtlateral temporal and multiple CALesion100.0 (18/18)100.0 (18/18)lesionectomyCAno spikeII (20)
15F37Lt*basal temporal CDTemporal lobe48.6 (18/37)29.7 (11/37)ATL + lesionectomyCDno spikeI (31)
16F28Lt*lateral temporal tumorTemporal lobe17.1 (6/35)91.4 (32/35)ATL + lesionectomyDNT + HSno spikeI (15)

Patients without neocortical lesion

AT localization was observed in five patients, and non-AT localization in six patients before surgery. The postoperative follow-up period was 16 to 35 months (mean, 25.5 months).

Figure 1 shows patients with AT localization: 89.1% (75.8–100%) of all estimated dipoles were localized within the AT area, and 62.9% (38.1–100%) were within the anterior temporal resection. All patients had hippocampal sclerosis and became seizure free (class I) after surgery. Postoperative MEG showed no interictal spikes.

Figure 1.

Anterior temporal (AT) localization in patients without neocortical lesion. Sagittal images and axial images along the temporal lobe axis (the chiasmatic–commissural plane) are shown before and after surgery. Spike dipoles within 7.5 mm thickness are overlaid on each slice. Dashed line indicates the boundary between the anterior and posterior temporal lobes, which was defined as the line perpendicular to the temporal lobe axis and passing through the crossing point of the sylvian fissure and central sulcus. In case 2, all spike dipoles were localized in the anterior temporal tip and included in the anterior temporal resection. In case 3, spike dipoles were localized in a more posterolateral area compared with case 2. Anterior temporal resection included 51.7% of the preoperative spike sources. Both patients had hippocampal sclerosis and became seizure free after AT lobectomy. Residual spike activity was not found at 12 months after surgery.

Figure 2 shows patients with non-AT localization: 25.9% (4.4–47.4%) of all dipoles were localized within the AT area, and 22.4% (4.4–49.0%) were within the anterior temporal resection. Various patterns were seen preoperatively in MEG spike localization in these patients. Four patients became seizure free after ATL, whereas the other two patients with frontal lobe dipoles had residual seizures. Hippocampal sclerosis was revealed in five patients. Residual spikes were detected at 12 months after surgery in three patients, including one patient with residual seizures.

Figure 2.

Figure 2.

Non–anterior temporal (AT) localization in patients without neocortical lesion. Spike dipoles were localized in the posterior temporal region in case 6, and diffusely localized over the midtemporal region in case 8. In case 9, a small number of the spike dipoles were localized in the AT area, but the majority occurred in the posterosuperior temporal lobe. These patients became seizure free after anterior temporal lobectomy (ATL), but postoperative magnetoencephalography (MEG) identified residual spiking in the corresponding region to preoperative MEG in cases 8 and 9. Spike dipoles were diffusely localized in the anterior frontal and temporal lobes in case 10, and in the posterosuperior temporal lobe and frontal lobe in case 11. These patients had residual seizures after ATL. Residual spikes were identified in the same region as by preoperative MEG in case 11.

Figure 2.

Figure 2.

Non–anterior temporal (AT) localization in patients without neocortical lesion. Spike dipoles were localized in the posterior temporal region in case 6, and diffusely localized over the midtemporal region in case 8. In case 9, a small number of the spike dipoles were localized in the AT area, but the majority occurred in the posterosuperior temporal lobe. These patients became seizure free after anterior temporal lobectomy (ATL), but postoperative magnetoencephalography (MEG) identified residual spiking in the corresponding region to preoperative MEG in cases 8 and 9. Spike dipoles were diffusely localized in the anterior frontal and temporal lobes in case 10, and in the posterosuperior temporal lobe and frontal lobe in case 11. These patients had residual seizures after ATL. Residual spikes were identified in the same region as by preoperative MEG in case 11.

Figure 2.

Figure 2.

Non–anterior temporal (AT) localization in patients without neocortical lesion. Spike dipoles were localized in the posterior temporal region in case 6, and diffusely localized over the midtemporal region in case 8. In case 9, a small number of the spike dipoles were localized in the AT area, but the majority occurred in the posterosuperior temporal lobe. These patients became seizure free after anterior temporal lobectomy (ATL), but postoperative magnetoencephalography (MEG) identified residual spiking in the corresponding region to preoperative MEG in cases 8 and 9. Spike dipoles were diffusely localized in the anterior frontal and temporal lobes in case 10, and in the posterosuperior temporal lobe and frontal lobe in case 11. These patients had residual seizures after ATL. Residual spikes were identified in the same region as by preoperative MEG in case 11.

Figure 2.

Figure 2.

Non–anterior temporal (AT) localization in patients without neocortical lesion. Spike dipoles were localized in the posterior temporal region in case 6, and diffusely localized over the midtemporal region in case 8. In case 9, a small number of the spike dipoles were localized in the AT area, but the majority occurred in the posterosuperior temporal lobe. These patients became seizure free after anterior temporal lobectomy (ATL), but postoperative magnetoencephalography (MEG) identified residual spiking in the corresponding region to preoperative MEG in cases 8 and 9. Spike dipoles were diffusely localized in the anterior frontal and temporal lobes in case 10, and in the posterosuperior temporal lobe and frontal lobe in case 11. These patients had residual seizures after ATL. Residual spikes were identified in the same region as by preoperative MEG in case 11.

Spike dipoles were localized in the posterior temporal lobe in case 6, and over the middle temporal lobe without clustering in cases 7 and 8 (Fig. 2A). These three patients had hippocampal sclerosis and became seizure free after ATL. Postoperative MEG showed spike disappearance in cases 6 and 7, but residual spikes were localized in the posterior temporal lobe adjacent to the anterior temporal resection in case 8. Most of the spike dipoles was localized in the posterior temporal lobe in case 9 (Fig. 2B). No lesion was found on the MRI, and intracranial EEG revealed the seizure origin in the mediobasal temporal cortex. Excellent seizure outcome was achieved after ATL. Postoperative MEG showed that the posterior temporal dipoles remained in the same region indicated by preoperative MEG. Frontal lobe dipoles were associated with temporal lobe dipoles in cases 10 and 11 (Figs. 2C and D). Spike dipoles were diffusely localized in the anterior frontal and temporal lobes in case 10, and in the posteroinferior frontal and posterior temporal lobes in case 11. Case 10 had atypical features including sensory seizure of the right upper limb and possible etiology of encephalitis, and the MRI revealed no extratemporal lesion other than hippocampal atrophy. Intraoperative electrocorticography (ECoG) confirmed frontal lobe spikes at a lower frequency than temporal lobe spikes. Case 11 showed atypical high-intensity change in the medial temporal structures extending to the insular cortex. Prolonged intracranial EEG study revealed seizure onset from the medial temporal lobe sometimes accompanied by simultaneous orbitofrontal onset. Both patients had hippocampal sclerosis, followed by residual seizures (classes II and III, respectively) after ATL. Postoperative MEG detected no interictal spikes in case 10, but spikes in the same location on preoperative MEG in case 11.

Figure 3 summarizes the mean and standard deviation of the dipole coordinates on preoperative MEG in patients without a neocortical lesion. The data were converted and plotted by using Talairach's coordinate system. All five patients with AT localization became seizure free and spike free, whereas four of the six patients with non-AT localization had residual seizures and/or spikes.

Figure 3.

Summary of spike dipole localization before surgery in 11 patients without neocortical lesion. Mean and standard deviation of dipole coordinates were converted to Talairach's standard brain coordinates (28) and viewed from the left lateral side. The crossing point of the central sulcus and sylvian fissure (white circle) was used as a landmark to separate the anterior temporal lobe from the posterior temporal lobe. The boundary line (solid line) was drawn perpendicular to the temporal lobe axis, which was defined at an angle of 24 degrees with the y-axis (29). Mean dipole position was localized in the anterior temporal (AT) lobe in five patients with AT localization. None of these patients had residual seizure or spike activity after anterior temporal lobectomy (ATL). However, four of the six patients with non-AT localization had residual seizures (n = 2) or residual spikes (n = 3) after ATL.

Patients with neocortical lesion

Lesion localization (Fig. 4) was found in three patients (cases 12–14) and lobar localization (Fig. 5) in two (cases 15 and 16). Lesionectomy without medial temporal resection was performed in the former three patients, followed by favorable outcome. In case 15, posterosuperior temporal spikes were found in addition to lesional spikes. In case 16, spike dipoles were localized diffusely in the basal temporal cortex. These two patients underwent prolonged intracranial EEG to identify the seizure origin. ECoG seizure onset was identified in the mesiobasal temporal cortex in case 15. Medial temporal onset was observed independently from the onset near the lateral temporal lesion in case 16. Both patients underwent lesionectomy with medial temporal resection, followed by excellent outcome. Histologic examination revealed hippocampal sclerosis in case 16. The follow-up period was 15 to 31 months (mean, 22.6 months). No interictal spike was detected in postoperative MEG in all patients.

Figure 4.

Lesional spike localization in patients with neocortical lesion before and after surgery. Spike dipoles were localized around the magnetic resonance imaging lesion (white arrowhead) in case 12 and on the lesion in case 14. Case 12 became seizure free after lesionectomy. Case 14 had favorable outcome (class II) after lateral temporal lesionectomy. No residual spike activity was found in these patients by magnetoencephalography at 12 months after surgery.

Figure 5.

Lobar spike localization in patients with neocortical lesion before and after surgery. Of spike dipoles, 48.6% were localized within 10 mm from the lesion, but the others were localized in the posterosuperior temporal lobe in case 15. Spike dipoles were localized diffusely in the basal temporal lobe in case 16. Both cases underwent lesionectomy with medial temporal resection after prolonged intracranial EEG study, followed by excellent seizure outcome. No interictal spike was found by postoperative magnetoencephalography.

DISCUSSION

This study compared neuromagnetic spike localization in patients with TLE before and after surgery, and found a positive correlation between MEG spike localization and epileptogenic area. Patients without neocortical lesion showing AT localization became seizure free and spike free after ATL, whereas some patients with non-AT localization had residual seizures and/or residual spikes, although the other patients became seizure free and spike free after ATL. Patients with neocortical lesion showing lobar localization suggested that the epileptogenic area was more extensive than the MR lesion.

Correlation between postoperative seizure outcome and magnetoencephalography spikes

The present study found that AT localization of the interictal MEG spikes indicated excellent seizure outcome in patients without neocortical lesion. Previously, anterior temporal spikes in scalp EEG and MEG were found to be highly specific to medial TLE (7,18–20). Classification of temporal lobe MEG spikes into anterior temporal horizontal (ATH), anterior temporal vertical (ATV), and posterior temporal vertical (PTV) dipoles (18,19) and comparison with ECoG findings showed that dominant ATH and ATV dipoles were correlated with seizure onset from the medial temporal structures, whereas PTV dipoles were correlated with onset from the lateral temporal neocortex or nonlocalized onset. Investigation of the relation between the dipole pattern and clinical diagnosis found that patients with medial TLE showed ATH or ATV dipoles, whereas patients with nonlesional TLE (no MRI abnormalities) showed dipoles localized on the anterior medial or posterior lateral temporal lobe (20). In the present study, all five patients with AT localization, corresponding to ATH or ATV dipoles, had hippocampal sclerosis and achieved excellent seizure outcome after ATL.

The spike dipole location was not totally included in the surgical resection in our patients with AT localization. However, the patients became seizure free and spike free after ATL. The spike dipole area is not identical to the epileptogenic area, but is usually more extensive, suggesting that the MEG spikes may be a remote propagation phenomenon distant from the epileptogenic area (2,30). The epileptogenic area in patients with medial TLE may have a strong tendency to propagate medial temporal spikes over the anterior temporal lobe. Thus AT localization indicates a possible epileptogenic area in the medial temporal structures, and complete resection of the MEG spike area in the lateral neocortex may be unnecessary. No MEG spikes were detected in the medial temporal structures, possibly because of the limitations of the method, as discussed later.

Non-AT localization implies that the epileptogenic area is located in the posterior or extratemporal lobe. Posterior temporal dipoles may indicate a lateral TLE or nonlocalizable focus (18). We agree that these dipoles indicate posterior temporal activity. However, four of our six patients with posterior temporal dipoles became seizure free after ATL. Therefore, lateral temporal epileptogenicity is difficult to determine based only on the MEG spike localization. This study had a relatively small patient population with homogeneous surgical outcome [i.e., all except one patient belonged to the good-outcome group (class I or II)], so no statistical difference in seizure outcome could be identified. A longer follow-up period is needed to evaluate the actual outcome of patients with non-AT localization, because late seizure recurrence is likely in patients with hippocampal sclerosis compared with other etiologies (31,32). Patients with non-AT localization should undergo further presurgical evaluations, such as intracranial EEG, to improve the seizure outcome.

The non-AT dipoles disappeared in three patients, but persisted in the other three patients on follow-up MEG at 12 months after ATL. This result suggests two possible mechanisms generating the non-AT dipoles: remote propagation from the medial temporal structures; and extended epileptogenic area. Disappearance of the interictal spikes might be due to insufficient sampling time. However, the non-AT dipoles were obviously reduced in the two patients with excellent seizure outcome; thus we believe that the non-AT localization was caused by remote activity propagated from the medial temporal epileptogenic area, and removal of the medial structures resulted in spike disappearance. Interictal cortical spikes can be propagated from the primary leading region near the epileptogenic area to remote secondary regions (4). Interictal spike propagations from deep to superficial temporal, anterior to posterior temporal, and temporal to frontal lobes were observed in spike-averaging studies (2,3). In these cases, the leading spike preceding the most prominent non-AT dipole may be detected by intensive analysis of MEG spikes (18,19).

The persistent residual spiking in the other three patients suggests that the anterior temporal resection did not include part of the epileptogenic area (i.e., the remnant of the posterior hippocampus or other neocortical area). How this activity is associated with in situ seizure generation is unknown. The residual epileptogenicity causing the interictal spikes might not have been sufficient to cause clinical seizures in the two patients with excellent outcome (33). However, postoperative interictal spikes detected at 6 to 12 months' follow-up examination were correlated with residual seizures in previous studies (34,35). Further investigation with a longer follow-up period is needed to assess the relation of residual MEG spike with seizure outcome.

Non-AT localization was related to residual seizures after ATL in our cases 10 and 11, who both had frontal lobe dipoles. Unfavorable seizure outcome after ATL was predicted by incomplete resection of the extratemporal spike area, indicated by implanted grid electrodes (8) and scalp EEG (6). Persistent spike activity was observed postoperatively in one patient, but not in the other patient. Sampling time may have been insufficient in the latter case; thus non-AT localization, especially the coexistence of frontal spikes, may suggest unfavorable seizure outcome in patients with TLE.

The perilesional irritable area was successfully localized by MEG in our patients with neocortical lesion, as found previously (20). Additionally, we found extensive spike localization in two patients. The intracranial EEG study revealed participation of the medial temporal structures in seizure generation in both cases, and histologic examination showed coexistence of hippocampal sclerosis in one patient. Thus an extensive epileptogenic area may be suggested when a significant number of dipoles are localized remote from the lesion. Lesion localization may indicate favorable seizure outcome after lesionectomy, whereas intensive presurgical evaluations are desirable in patients with lobar localization to improve seizure outcome.

Sensitivity of neuromagnetic spike localization

MEG does not always detect all interictal spike sources of patients, because of the limited sensitivity. Radial currents to the scalp (i.e., gyral cortical activity in the lateral cortex) may not be detected by MEG (18,19). In contrast, tangential currents in the superior or basal temporal plane and temporal tip are often localized by MEG. Deep sources (i.e., medial temporal spikes) may not be detected by MEG (20,30,36). In the present study, no medial temporal dipole was localized, even in the patient with hippocampal sclerosis and excellent seizure outcome. We believe that the present MEG study detected predominantly neocortical spikes tangential to the scalp.

Short sampling time MEG may overlook interictal spikes. The recording time is limited in current MEG technology compared with conventional EEG (20,37). In this study, sufficient interictal spikes were obtained in 16 (67%) of 24 patients. This sensitivity agrees with that (73%) of the previous study of TLE (17). Oral barbiturate was used to induce the sleep state to increase interictal spike density in this study (38). However, insufficient spike sampling does not exclude the presence of epileptic activity. No interictal spike was detected by postoperative MEG in two of our three patients with residual seizures (cases 10 and 14). Induced light sleep and decreased antiepileptic medication may be useful to increase the density of interictal spikes (20).

Limitation of source-estimation accuracy in magnetoencephalography

We applied single ECD modeling, assuming that all neuronal activity can be represented as a point generator. However, this model has limited capacity to accommodate extended or multiple generators (18,19). The size of the cortical area required to generate detectable MEG spikes is unknown (18). Nevertheless, in this study, spike localization characterized the dominant irritative site for each patient, which was correlated with the postoperative findings. Thus we believe that single ECD modeling is clinically useful to localize epileptic activity.

Spike dipoles were estimated on the single-spike basis in the present study, so background brain noise might affect the source estimation accuracy. Sleep induction and the high-pass filtering technique were used to improve the signal-to-noise ratio of the spike component. We believe that these techniques provide source localization accuracy acceptable for clinical use.

Ancillary