MEG Predicts Outcome Following Surgery for Intractable Epilepsy in Children with Normal or Nonfocal MRI Findings

Authors


Address correspondence and reprint requests to Rajesh RamachandranNair at Division of Pediatric Neurology, McMaster University, McMaster Children's Hospital, 1200 Main St W, Hamilton, Ontario, Canada L8N 3Z5. E-mail: rajeshrnair@hotmail.com

Abstract

Summary: Purpose: To identify the predictors of postsurgical seizure freedom in children with refractory epilepsy and normal or nonfocal MRI findings.

Methods: We analyzed 22 children with normal or subtle and nonfocal MRI findings, who underwent surgery for intractable epilepsy following extraoperative intracranial EEG. We compared clinical profiles, neurophysiological data (scalp EEG, magnetoencephalography (MEG) and intracranial EEG), completeness of surgical resection and pathology to postoperative seizure outcomes.

Results: Seventeen children (77%) had a good postsurgical outcome (defined as Engel class IIIA or better), which included eight (36%) seizure-free children. All children with postsurgical seizure freedom had an MEG cluster in the final resection area. Postsurgical seizure freedom was obtained in none of the children who had bilateral MEG dipole clusters (3) or only scattered dipoles (1). All five children in whom ictal onset zones were confined to ≤5 adjacent intracranial electrodes achieved seizure freedom compared to three of 17 children with ictal onset zones that extended over >5 electrodes (p = 0.002). None of six children with more than one type of seizure became seizure-free, compared to eight of 16 children with a single seizure type (p = 0.04). Complete resection of the preoperatively localized epileptogenic zone resulted in seizure remission in 63% (5/8) and incomplete resections, in 21% (3/14) (p = 0.06). Age of onset, duration of epilepsy, number of lobes involved in resection, and pathology failed to correlate with seizure freedom.

Conclusions: Surgery for intractable epilepsy in children with normal MRI findings provided good postsurgical outcomes in the majority of our patients. As well, restricted ictal onset zone predicted postoperative seizure freedom. Postoperative seizure freedom was less likely to occur in children with bilateral MEG dipole clusters or only scattered dipoles, multiple seizure types and incomplete resection of the proposed epileptogenic zone. Seizure freedom was most likely to occur when there was concordance between EEG and MEG localization and least likely to occur when these results were divergent.

Focal MRI abnormalities significantly increase the chance of successful delineation of the epileptogenic zone preoperatively in patients who are candidates for the surgical treatment of their epilepsy (Duncan and Sagar, 1987; Fish et al., 1993; Spencer, 1995; Armon et al., 1996; Cascino, 2004; Tonini et al., 2004). Therefore, epilepsy surgery provides a good postsurgical outcome in patients who have focal or abnormal MRI findings (Duncan and Sagar, 1987; Fish et al., 1993; Spencer, 1995; Armon et al., 1996; Tonini et al., 2004). In particular, MRI evidence of hippocampal sclerosis or tumor is associated with good postsurgical outcome following surgery (Tonini et al., 2004). However, MRI does not aid in the presurgical evaluation in nearly 29% of patients in whom it is normal or shows nonspecific findings (Semah et al., 1998).

The outcome following epilepsy surgery in patients with normal brain MRI depends on the case selection criteria and expertise of the epilepsy center. There is no accurate estimate of the prevalence of normal MRI findings in children with intractable epilepsy who are potential surgical candidates. The majority of children with normal MRIs would seem not to be surgical candidates during the presurgical evaluation. However, the weight given to a normal or nonspecific MRI depends on the case selection criteria adopted by the epilepsy center, since each center decides surgical candidacy in patients with normal MRI findings based on their own presurgical evaluation modalities. For example, in a cohort of 75 children below the age of 12 yr, who underwent resective surgery for intractable epilepsy at a pediatric epilepsy center, 35 had no identifiable focal lesion in MRI (Paolicchi et al., 2000).

We retrospectively analyzed clinical profiles, interictal magnetoencephalography (MEG), intracranial EEG findings from extraoperative intracranial invasive monitoring, surgical procedures, postoperative electrocorticography (ECoG), and pathology as to their relation to postsurgical seizure outcomes in a cohort of children with normal, subtle or nonfocal MRI findings, who underwent epilepsy surgery for intractable epilepsy. We hypothized in children with normal MRI and intractable localization-related epilepsy a single MEG cluster represented the epileptogenic zone.

METHODS

We analyzed patients with normal, subtle or nonfocal MRI findings, who underwent surgery for intractable epilepsy during the period of July 1998–June 2005 following extraoperative video EEG (VEEG) monitoring from a subdural electrode array. A data abstraction form was used to gather information on the clinical profiles, neurophysiological findings of scalp and extraoperative intracranial VEEG, MEG, postoperative ECoG, and neuroradiological features.

All the children underwent prolonged scalp VEEG (BMSI System 5000, Nicolet, Madison, WI, U.S.A.; Harmonie 5.4, Stellate, Montreal, PQ, Canada) to evaluate interictal epileptiform discharges, and seizures using the International 10–20 scalp-electrode placement system with a single reference electrode. Three or more typical seizures were captured.

All patients underwent MEG studies with a whole-head gradiometer Omega system (151 channels, VSM MedTech Ltd., Port Coquitlam, BC, Canada). During MEG, they also underwent simultaneous EEG that was recorded from 19 electrodes (International 10–20 system). We recorded 2 min periods of spontaneous MEG data 15 times. The sampling rate for data acquisition was 625 Hz. We visually identified MEG epileptic discharges, spikes and sharp waves (referred to as spikes), by examining the MEG recordings and cross-referencing them with the simultaneous EEG recording using a band pass filter of 3–70 Hz and a notch filter of 60 Hz. We applied a single moving dipole analysis with a single-shell, whole-head spherical model. We defined the MEG spike dipole for each spike as a single dipole fit from the earliest phase of each spike with the criteria of a residual error of less than 30%. The MEG spike dipole sources were mapped onto the MRI (T1-WI; 2 mm thickness, no skip) pixels by using the MARK VOXEL program (VSM MedTech Ltd., Port Coquitlam, BC, Canada).

We defined the MEG spike source distributions by the number and density (Iida et al., 2005a). Clusters consisted of six or more spike sources with ≤1 cm between adjacent sources; scatters consisted of fewer than six spike sources regardless of the distance between sources, or spike sources with >1 cm between sources regardless of the number of sources in a group (Iida et al., 2005b).

Preoperative MRI of brain was performed using a GE 1.5 T Signa MRI (GE Medical Systems, Milwaukee, WI, U.S.A.). The epilepsy protocol included the following sequences: sagittal T1-WI; axial and coronal dual-echo T2-WI; coronal FLAIR sequence; and coronal volumetric three-dimensional Fourier transform (3DFT) gradient echo sequence. A pediatric neuroradiologist (MMS) who was blinded to the clinical information analyzed all the MRI pictures. We included children with subtle and nonfocal abnormalities such as mild diffuse brain atrophy, periventricular leukomalacia, and nonspecific white matter signal changes in the study.

Extraoperative intracranial VEEG monitoring from subdural grid electrodes was performed before epilepsy surgery. We placed intracranial electrodes (subdural grids with or without depth electrodes) based on the seizure semiology, scalp VEEG and MEG findings, and neuropsychological profiles as described (Minassian et al., 1999; Snead, 2001).

Based on analysis of the extraoperative intracranial VEEG monitoring which captured at least three habitual seizures, we delineated the resection areas involving ictal onset zones, part of ictal symptomatogenic zones, and active interictal zones adjacent to the ictal onset zone (Minassian et al., 1999; Snead, 2001; Park et al., 2002). Localization of the eloquent cortices was determined as described (Snead, 2001) and defined the completeness of the resection. Surgical procedures consisted of lobectomy, corticectomy, multiple subpial transaction (MST) or a combination. We performed MST prior to 2004 when the epileptogenic zone involved eloquent cortex (Blount et al., 2004). Since 2004, we have used resective surgery over facial motor cortex. Postoperative ECoG was performed at the margin of the resection and was graded A–D (Nolan et al., 2004). Pathology of the resected brain tissue was reviewed and was classified as either specific, and well defined; nonspecific, or normal.

Using the neuronavigation system (Carl Zeiss Canada, Ltd., Toronto, ON, Canada), we compared the MEG spike source distribution with the epileptic zone demarcated by intracranial VEEG on the exposed brain surface (Iida et al., 2005a). A single spike source from an MEG spike can not determine the spatial extent of the epileptic zone because the model represents the center of activation by a point source rather than the area of activated cortex (Pataraia et al., 2002). Therefore, we defined colocalization as MEG spike sources that occurred within the same gyrus and/or within one electrode adjacent to the epileptic zone. We retrospectively analyzed the part of the MEG spike source distribution that had been surgically treated, using the 3D surface renderings of MRI-MEG images, the intraoperative digital pictures of brain surface imaging with electrocorticography, and the digital pictures of intracranial video EEG monitoring (Rutka et al., 1999; Otsubo et al., 2001).

We classified the seizure outcome at the last follow-up according to Engel's classification (class I–IV) (Engel, 1993). More than 90% reduction in seizure frequency was classified as Engel class IIIA (worthwhile improvement). Good outcome was defined as Engel class I–IIIA. This study was approved by the research ethics board of the Hospital for Sick Children.

STATISTICS

Fisher's exact test was used to evaluate possible associations of clinical, neurophysiological and pathological findings with postsurgical seizure freedom.

RESULTS

During the study period, 61 children underwent extraoperative intracranial VEEG monitoring from subdural electrodes for intractable epilepsy. Twenty-four children satisfied the MRI criteria of normal or subtle and nonfocal MRI findings. Two children did not undergo surgery because the intracranial VEEG data suggested bilateral ictal onset zones.

Table 1 describes clinical profiles, MRI findings, EEG and MEG findings, surgical procedures, pathology, and the postsurgical seizure outcomes. Twenty-two children (10 boys and 12 girls) were included in the study. Mean age of seizure onset was 4.3 yr (0.5–9 yr) and the mean duration of epilepsy was 7.4 yr (2.5–13 yr). Mean age at surgery was 11.7 yr (4–18 yr). Six children (patients 1, 6, 7, 14, 19, and 20) had more than one type of seizure. Fourteen children had a normal MRI. Eight children showed subtle abnormalities including cortical thickening (patients 18 and 20), small white matter hyperintensity (patients 11 and 12) (Fig. 1A), and abnormal signal intensity periventricularly (patients 1, 5, 19, and 22).

Table 1. Clinical profiles, MRI, EEG, MEG findings, surgical procedures, pathology, and postsurgical seizure outcomes
 Scalp EEG Sugical procedures and locations of resection 
No.Age (yr) /sexSeizure duration (yr)Types of seizureMRIInterictal dischargesIctal onset zoneMEG Interictal dipolesIntracranial EEG Ictal onset zoneLobectomyCorticectomyMSTComplete/IncompletePathologyFollow-up (Months)Postsurgical seizure outcomes
  1. C, central; CD, cortical dysplasia; CPS, complex partial seizure; DNET, dysembryoplastic neuroepithelial tumor; F, frontal; GSW, generalized spike and wave; GTCS, generalized tonic–clonic seizure; GTS, generalized tonic seizure; IOZ, ictal onset zone; L, left; MST, multiple subpial transaction; MTS, mesial temporal sclerosis; NA, not available; O, occipital; P, parietal; PVL, periventricular leukomalacia; R, right; SPS, simple partial seizure; T, temporal; 2GS, secondary generalized seizure.

19/Female6CPS with visual aura, R-focal motor with 2GSPVLL-T,P,O, rare bilateral-F,GSWL-T,OCluster L-OExtensiveL-OL-PPosterior portion of L-sup F gyrusIncompleteCD184B
216/Female12CPSNormalBilateral-PNonlateralizedCluster R-PExtensiveR-perirolandic, superior T gyrusIncompleteNA364B
314/Female7SPS, 2GSNormalR-FR-FScatter R-F,CExtensiveR-anterior FR-middle F gyrusR-angular, supramarginal gyriIncompleteNormal243A
45/Male4.5CPS,2GSNormalR-FC,GSWNonlateralizedCluster R-F,CRestrictedR-precentralR-precentral (middle portion)IncompleteGliosis331A
517/Male13CPS, 2GSPeriventricular hyperintensity, RR-F,C,PNonlateralizedCluster R-CExtensiveR-PR-perirolandicIncompleteCD161D
611/Female6CPS, Myoclonic, 2GSNormalBilateral-C,TNonlateralizedCluster R-P,TExtensiveR-TR-postcentral, posterior TR-perirolandic (facial motor area)IncompleteMTS183A
76/Male2.5SPS,GTC, MyoclonicNormalL-F,T,GSWNonlateralizedClusters bilateral TExtensiveL-TL-superior T gyrus, inferior, middle F gyriIncompleteCD363A
815/Male12.5CPSNormalR-F,GSWR-FCluster R-FExtensiveR-F-CompleteNormal12IA
913/Female8GTSNormalR-F,T, L-FNonlateralizedCluster R-CRestrictedR-middle and inferior F gyri (posterior portion), supramarginal and angular giri-CompleteCD321A
1015/Female6CPS, 2GSNormalR-F,TR-F,TCluster R-FRestrictedR-anterior FR-middle F, superior T, superior PR-P, perirolandic (facial motor area)IncompleteGliosis651A
1118/Female13Focal clonicSubtle hyperintensity L-P white matterL-C, midlineNonlateralizedCluster L-PExtensiveL-postcentral-Incomplete, hand motor area (IOZ) leftGliosis182D
1212/Female9CPS,2GSSmall hyperintensity R- F white matterL-F,T,P, R-TL-hemisphericCluster L-TExtensiveL-anterior TL-posterior T-CompleteCD181A
137/Female4.5CPSNormalL-T,PL-F,TCluster L-F,TExtensiveL-anterior TL-postcentral (inferior portion), superior T (posterior portion)L-supramarginal, angular gyri, inferior F gyrus (posterior portion)IncompleteNormal444B
144.5/Male4CPS,2GS, MyoclonicNormalL-F,C,T, R-C,TNonlateralizedCluster L-inf C,P, R-C,TExtensiveL-anterior TL-perirolandic, T (posterior portion), angular gyrusIncompleteCD253A
1516/Female9CPS,2GSNormalBilateral-T,CNonlateralizedClusters, bilateral post C, L- post TExtensiveR-T (lateral cortex)R-perirolandicIncompleteGliosis583A
166/Male3.5CPS,2GSNormalL-F,T,OL-T,P,OCluster L-T,PRestrictedL-T (posterior portion),P,OL-superior T gyrus, angular gyrusIncompleteDNET401A
1717/Male14CPSNormalBilateral-F,TNonlateralizedCluster R-F,TRestrictedR-periSylvian-CompleteGliosis241A
1814/Male7SPS, 2GSIsolated cortical thickening, L-ant FL-CL-CCluster L-FExtensiveL-perirolandicL-perirolandic (hand motor area)IncompleteCD183A
1912/Male9Atonic, CPS, Tonic, GTCSPVLBilateral-P,T,O, GSWNonlateralizedCluster R-T,P,O, scatter L-hemisphericExtensiveR-anterior TR-F,P,O-CompleteCD203A
2012/Female5CPS, TonicSubtle cortical thickening, L-FBilateral-F (L>R),GSWNonlateralizedCluster L-FExtensiveL-anterior FL- inferior, middle F gyri-CompleteNormal154B
2114/Male5SpasmsNormalL-F,CL-F,CCluster L-FExtensiveL-ant FL-suprior, inferior F gyri, middle T gyrus-CompleteCD161A
224/Male2.5SpasmsBilateral periventricular heterotopiaL-hemispheric, rare R-hemisphericNonlateralizedCluster L-C,P,OExtensiveL-OL-F (posterior portion), postcentral-CompleteGliosis94B
Figure 1.

MRI, MEG and intracranial ictal onset zone in patient 12. (A) MRI axial T2 image shows an area of high signal intensity in the right anterior frontal subcortical area. (B) Sagittal T1 MR image shows a cluster of MEG spikes sources in the left temporal region. A few scatters are also seen in the parietal and posterior temporal regions. (C) Extensive ictal onset zone corresponding to 12 electrodes is shown within yellow lines on the exposed left frontal, parietal and temporal brain surface. Numbers in small white squares denotes the location of intracranial subdural electrodes.

Interictal epileptiform discharges on scalp EEG were located unihemispherically in only eight children. Six children had generalized spike and wave discharges. Scalp VEEG correctly localized the ictal onset in eight patients and hemispheric lateralization was possible in another child.

MEG dipole clusters were located in a single hemisphere in 18 children (Figs. 1B, 2A). Three children had bilateral MEG dipole clusters and one child had only scatters.

Figure 2.

Figure 2.

MEG and intracranial ictal onset zone in patient #17. (A) Sagittal T1 MR image shows a cluster of MEG spike sources in the right frontotemporal region. (B) Ictal onset zone restricted to two electrodes is shown within yellow lines on the exposed right frontal, parietal and temporal brain surface. Numbers in small white squares denotes the location of intracranial subdural electrodes.

Figure 2.

Figure 2.

MEG and intracranial ictal onset zone in patient #17. (A) Sagittal T1 MR image shows a cluster of MEG spike sources in the right frontotemporal region. (B) Ictal onset zone restricted to two electrodes is shown within yellow lines on the exposed right frontal, parietal and temporal brain surface. Numbers in small white squares denotes the location of intracranial subdural electrodes.

All the children had intracranial extraoperative video EEG monitoring from intracranial subdural grids (number of electrodes, mean = 100). Eighteen children had additional strip/depth electrodes. Contralateral intracranial electrodes (strips and depths) were placed in 13 children. Two children underwent repeat invasive monitoring after ictal lateralization with bilateral strip electrodes. Surgical procedures consisted of lobectomy (2), corticectomy (9), combined lobectomy and corticectomy (10), and MST alone (1). Twelve children who had lobectomy and or corticectomy had MST in the eloquent cortex. In 14 children, including 13 who received MST, resection was deemed incomplete. Surgery involved one lobe in five, two lobes in nine and more than two lobes in eight. Postoperative ECoG was as follows: grade A in seven, grade B in 12, grade C in one and grade D in two.

Histology of the resected brain tissue in 21 children showed a specific pathology in eleven (cortical dysplasia in nine, dysembryoplastic neuroepithelial tumor in one and mesial temporal sclerosis in one), and nonspecific gliosis in six. The pathology was normal in four children.

POSTSURGICAL OUTCOMES

The mean follow up was 27 months (9–67 months). Eight children (36%) were seizure free. Good postsurgical seizure outcome was attained in 17 (77%) children (including the 8 who were seizure free). All the children except one (patient 22) completed at least 12 months follow-up after surgery. This child with 10-month follow up had a class IVB outcome.

MEG dipole clusters were present in the region of final resection in 18 children. Thirteen of these children (72.2%) had good postsurgical seizure outcomes; including eight who were seizure-free (8/18). All the children with postsurgical seizure freedom (8) had the MEG dipole cluster in the final resection area (Figs. 1B, 2A). Postsurgical seizure freedom was obtained in none of the four children who had bilateral MEG dipole clusters (patients 7, 14, and 15) or only scatter (patient 3). In the 17 cases with good postsurgical outcome, the MEG dipole cluster correctly localized the resection area in 13, interictal EEG in seven, and ictal scalp VEEG in six.

Postsurgical seizure freedom was not achieved in any of the six children with more than one type of seizure (p = 0.04). Similarly, none of the four children with generalized interictal spike and waves and focal epileptic spikes outside the resection area were seizure free (p = 0.13).

Intracranial ictal onset zone was confined to 10.5 (mean) electrodes (median 5) in seizure free children and 15.5 (mean) electrodes (median 10) in children with residual seizures. When we took the median as cutoff (five electrodes), there were five children who had intracranial ictal onset restricted to five or less adjacent electrodes (Figs. 1C, 2B). All of them were seizure free (p = 0.002). Only three of the 17 children with ictal onset zone extended over more than five electrodes (extensive ictal onset zone) became seizure free.

Complete resection of preoperatively localized epileptogenic zone resulted in seizure freedom in 62.5% (5/8) and incomplete resections in 21.4% (3/14) (p = 0.06).

There was no statistically significant association of outcome with age of onset or duration of intractable epilepsy, interictal epileptic discharges on scalp EEG outside the resection area, presence of interictal generalized spike and wave discharges, undetermined ictal onset localization by scalp VEEG, number of lobes involved in resection, or pathology of resected tissue. Similarly, postoperative ECoG did not predict seizure-free outcome since only three out of seven children with postoperative ECoG grade A were seizure free, and four out of 15 with postoperative ECoG grades B–D were seizure-free.

DISCUSSION

Seventeen of 22 (77%) children with normal or nonfocal MRI findings achieved good postsurgical outcomes. This group included eight (36%) seizure free children. In a metaanalysis of 47 studies of epilepsy surgery, an abnormal MRI was reported as a good prognostic factor (Tonini et al., 2004). There is much variation in the reported success rate of epilepsy surgery in patients with normal or nonlocalizing features depending on the center (Paolicchi et al., 2000; Blume et al., 2004; Chapman et al., 2005; Lee et al., 2005). This could be related to the difference in the selection criteria adopted by individual epilepsy programs. In a large series of 89 patients (adult and pediatric) with cryptogenic neocortical epilepsy, good postsurgical seizure outcome was obtained in 80% and seizure freedom in 47% (Lee et al., 2005). Similar rates were reported by Blume et al. (37% seizure free and 61% good outcome) (Blume et al., 2004). In a group of 24 adult and pediatric patients, Chapman et al. reported 37% seizure freedom and more than 90% seizure reduction in 75% (Chapman et al., 2005). An exclusive pediatric series by Paolocchi et al reported a higher overall percentage of postsurgical seizure freedom (51%) in 35 children with nonlesional epilepsy. However, 21 children were classified as ‘nonlesional’ based on 0.5 T MRI. Only seven children had multilobar resection (Paolicchi et al., 2000). In this group, one child (14%) became seizure free and a total of three children (43%) had good outcome. Our cases differed from these series because the majority of our children had two or more lobes involved in the surgical resection. Despite the complexity of the cases, we had similar results. The high success rate in the series by Lee et al. (2005) was attributed to case selection based on PET data. We believe that case selection based on MEG data and extensive coverage using intracranial electrodes helped us to obtain good postsurgical outcomes in children with extensive epileptogenic zone and normal, subtle or nonfocal MRI findings.

The absence of a MEG dipole cluster located exclusively within the region of final resection had 100% negative predictive value with respect to postsurgical seizure freedom since postsurgical seizure freedom was obtained in none of the four children who had a bilateral MEG dipole clusters or only scattered dipoles. Conversely, all children who had postsurgical seizure freedom also had a MEG dipole cluster confined to the resection area. The ictal onset zone was localized in the three children who had bilateral MEG dipole clusters after two-stage intracranial extraoperative monitoring (two children) or simultaneous bilateral recording (one child). However, surgery failed in all four of these children. These findings lead to two interesting conclusions: (1) The presence of a MEG dipole cluster confined to the resection area is a prerequisite for postsurgical seizure freedom, and (2) The absence of a MEG dipole cluster or the presence of bilateral MEG dipole clusters or scatter may predict seizure recurrence following epilepsy surgery.

These MEG data are in agreement with our previous report that the localization of the MEG dipole cluster correlates highly with that of the ictal onset zone and further, that failure to resect the brain region containing the MEG dipole cluster (e.g., eloquent cortex) leads to postoperative seizure recurrence (Minassian et al., 1999). A recent study by Oishi et al. showed that a single MEG cluster, the localization of which comported with that of the ictal onset zone, was associated with very high postoperative seizure-free outcomes (eight of nine patients), whereas only three of the 11 patients who had a single MEG cluster smaller than or partially overlapping the ictal onset zone, multiple clusters overlapping the ictal onset zone, or MEG clusters not overlapping the ictal onset zone, had a postoperative seizure-free outcome (Oishi et al., 2006). This study and our findings support the hypothesis that potential ictal onset zone could arise postoperatively in the unresected MEG dipole cluster area, although the apparent ictal onset zone was mapped to a different cortical area. It is difficult to draw any conclusion from the single case where MEG showed only scattered dipoles. The absence of a single MEG dipole cluster may suggest the extensive or multiple epileptic networks. Fischer et al. (2005) applied a novel technique designed to generate an ellipsoidal volume from the scattering of single MEG source localizations to represent MEG results in 33 adult patients who underwent surgery for epilepsy. This volume was compared voxel wise with the resection volume generated from pre and postoperative MR images. A high coverage of the MEG results ellipsoid by the resection volume and a low distance between the mass centers of both volumes correlated to a favorable outcome (Fischer et al., 2005). We have been using the neuronavigation system to compare the MEG spike source distribution with the epileptic zone demarcated by IVEEG on the exposed brain surface. We included the entire MEG cluster in the resection area, the limiting factor being the eloquent cortex.

A restricted intracranial ictal onset zone predicted postsurgical seizure freedom. Restricted ictal onset zone has been reported to be more common in extratemporal epilepsy than temporal lobe epilepsy in adult patients (Lee et al., 2003). Two of the earlier adult series had reported no association with the surgical outcome and restricted intracranial ictal onset in cryptogenic epilepsy (Schiller et al., 1998; Lee et al., 2003). We postulate that restricted intracranial ictal onset zone is associated with a smaller epileptic network in children. Our policy was to include the area of active interictal activity adjacent to the ictal onset zone in the resection to include the future potential ictal zone, that Rosenow et al. called high-threshold ictal onset zone during intracranial VEEG (Rosenow and Luders, 2001). When the ictal onset zone was small (≤5 electrodes), the resected neighboring active interictal zone may be the potential ictal onset zone. When the ictal onset zone was large (>5 electrodes), the resection area may not be proportionally large enough to cover the entire epileptogenic zone including potential ictal onset zone.

Children with multiple seizure types failed to achieve seizure freedom. During intracranial VEEG monitoring all types of seizures were captured and analyzed to decide on area of resection. Extensive and or multiple epileptic networks might have existed in these children. None of the interictal features on scalp EEG (focal spike outside resection area, bilateral spikes and generalized spikes) were associated with outcome. However, there was some indication that a combination of generalized spikes and focal spikes outside the resection might be a poor prognostic indicator. Presence of a spike outside the resection area, multiple spike foci and generalized spike-waves were previously reported as independent poor prognostic factors (Blume et al., 2004). Since we did not do dipole analysis using the scalp EEG data, we were unable to compare the concordance between the MEG dipole cluster and EEG dipole cluster. Most authors agree that combined EEG and MEG analysis adds to the sensitivity and specificity of presurgical localization of the irritative zone (Barkley, 2004; Bast et al., 2004; Baumgartner, 2004; Iwasaki et al., 2005). Ochi et al used EEG and MEG dipole lateralizations to identify the primary epileptogenic hemisphere in 41 children with intractable localization-related epilepsy. Concordant lateralizations predicted good seizure control after surgery by identifying the primary epileptogenic hemisphere. Discordant lateralizations signified an undetermined epileptogenic hemisphere (Ochi et al., 2005).

We included children with subtle and nonfocal findings in MRI because these features are unlikely to provide useful clues regarding the localization of the potential epileptogenic zone during the presurgical workup. Similar MRI criteria have been used by others as well (Cukiert et al., 2001; Chapman et al., 2005). We found that subtle MRI abnormalities could be misleading, as in patients 12 (Fig. 1A) and 18, where the epileptogenic zone proved to be distant from the site of MRI abnormality.

We found no correlation between outcome and the pathology of resected tissue. Forty-three percent (9/21) of the resected brain tissue in our series showed evidence of cortical dysplasia. These data comport with reports demonstrating cortical dysplasia histologically in 20%–70% of patients who had undergone surgery for cryptogenic epilepsy (Paolicchi et al., 2000; Siegel et al., 2001; Chapman et al., 2005; Lee et al., 2005). Similarly, our one case of DNET (patient 16), not seen on MRI was analogous to the single case of ganglioglioma missed by MRI in a cohort of 43 patients (Siegel et al., 2001). Imaging with 3T MRI, newer MR techniques like multiplanar and curvilinear reconstruction might increase the yield of neuroimaging in these children (Montenegro et al., 2002).

We used MST over eloquent cortex until 2004 (Blount et al., 2004); hence many children up until that time had only partial resection. However, from 2004 we increasingly used resection of the eloquent cortex depending on the case. A higher proportion of incomplete resection in the earlier patients in this series could have adversely affected the surgical results since there was a trend toward seizure freedom when the resection was complete. In the pediatric series by Paolicchi et al. (2000), complete resection of the epileptogenic zone resulted in seizure freedom in 76%. Only 27% of the children with incomplete resections had seizure freedom. Those authors combined lesional and nonlesional cases for their analysis.

In summary, this study demonstrates that a good postsurgical outcome can be obtained in the majority of children with intractable epilepsy who have a normal or nonfocal MRI. A restricted ictal onset zone within an MEG spike cluster was associated with postsurgical seizure freedom. Children with multiple seizure types failed to achieve postsurgical seizure freedom, as did those in whom the MEG spike sources were either bilateral or diffuse.

Ancillary