Epilepsy surgery in patients with normal or nonfocal MRI scans: Integrative strategies offer long-term seizure relief


Address correspondence to Dr. Prasanna Jayakar at the Department of Neurology, Brain Institute, 3100 SW 62nd Avenue, Miami, FL 33155, U.S.A. E-mail: prasanna.jayakar@mch.com


Purpose: Excisional surgery achieves seizure freedom in a large proportion of children with intractable lesional epilepsy, but the outcome for children without a focal lesion on MRI is less clear. We report the outcome of a cohort predominantly of children with nonlesional intractable partial epilepsy undergoing resective surgery.

Methods: We studied 102 patients with nonlesional intractable partial epilepsy who underwent excisional surgery. The epileptogenic region was identified by integrating clinical exam and video-EEG data complemented by ictal SPECT (n = 40), PET (n = 10), extraoperative subdural monitoring (n = 80), and electrocorticography (n = 22). All patients had follow-up greater than 2 years, 76 patients had 5-year follow-up, and 43 patients had 10-year follow-up.

Results: A total of 66 resections were unilobar; 36 were multilobar. One patient died of causes unrelated to seizures or surgery. At 2-year follow-up, 44 of 101 patients were seizure-free, 15 experienced >90% reduction, 17 had >50% reduction, and 25 were unchanged. At 5-year follow-up, 35 of 76 patients were seizure-free, 12 experienced >90% reduction, 12 had >50% reduction, and 17 were unchanged. At 10-year follow-up, 16 of 43 patients were seizure-free, 13 experienced >90% reduction, 7 had >50% reduction, and 7 were unchanged.

Outcomes correlated with the presence of convergent focal interictal spikes (p < 0.005) on the scalp EEG and completeness of resection (p < 0.0005).

Conclusions: Our findings demonstrate that excisional surgery is successful in the majority of children with nonlesional partial epilepsy. A multimodal integrative approach can minimize the size of resection and alleviate the need for invasive EEG monitoring. Focal interictal spikes and completeness of resection predict good outcome. The benefits of surgery are long-lasting.

A discrete lesion on the MRI scan is arguably one of the most reliable features for defining the epileptogenic region (Engel et al., 1993; Fish et al., 1993; Spencer, 1995; Armon et al., 1996; Wyllie et al., 1998; Cascino, 2004; Tonini et al., 2004). Its absence renders accurate seizure focus localization more difficult. Correspondingly, several studies conducted primarily in adults claim lower rates of seizure freedom following resective surgery in patients with normal or nonspecific MRI scans (nonlesional) compared to patients with discrete lesions (Siegel et al., 2001; Blume et al., 2004; Sylaja et al., 2004; Chapman et al., 2005; Lee et al., 2005; Alarcon et al., 2006).

The ability to localize seizure origin is even more challenging in children with nonlesional epilepsy in whom widespread and extratemporal epileptogenesis related to developmentally malformed cortex is common. The absence of an MRI lesion thus weighs heavily against surgical candidacy, and pediatric epilepsy centers typically defer surgical consideration in this population.

There is a paucity of information to guide surgical strategy in children with nonlesional epilepsy, especially information regarding long-term outcomes. Our experience with resective surgery in this population has suggested that success is often possible despite the challenges presented (Duchowny et al., 2000). We report the outcome of resective surgery in patients with normal/nonfocal MRI scans and attempt to identify predictive variables that facilitate candidate selection and guide surgical strategies.


Patient population

We did a retrospective chart review and identified 128 patients from a population of 622 patients in our epilepsy surgery database, who were (1) 21 years of age or younger at time of surgery; (2) had MRI scans reported as normal or nonfocal at time of surgery [nonfocal MRI findings were defined as diffuse atrophy, ventricular dilatation, or bilateral nonspecific white matter signal changes] and (3) had follow-up of at least 2 years. From this cohort, 22 patients who underwent callosotomy alone and four children who had no resection following implantation were excluded. In the remaining 102 patients who underwent resective surgery, we performed a retrospective chart review and statistical analysis of pre- and perioperative variables.

Presurgical evaluation

All patients had a detailed history and clinical exam. The history and medical records from the referral source were reviewed for consistency in the laterality/focality of seizure semiology and EEG findings over the duration of the child's epilepsy with particular attention to data at the time of onset.

Video EEG monitoring was performed to capture multiple seizures and was repeated in two or more monitoring sessions at least several months apart to document consistency. Recordings were obtained on 32 channel analog or digital systems with standard 10–20 electrode placements; closely spaced electrodes including Silverman (T1/T2) positions were utilized when indicated and montage reformatting was utilized routinely.

Field plots of focal interictal spikes and ictal onsets were analyzed to further define the region of abnormality. Special attention was paid to defining focality within the context of apparently widespread or bisynchronous interictal discharges. Besides identifying spike voltage maxima, the analyses included identifying complex horizontal dipolar or multipolar fields and defining subtle time leads using reference-subtraction montage (Jayakar et al., 1991a, 1991b).

MRI studies were performed with a 0.3-Tesla magnet in 28 patients operated prior to 1992; the remaining 74 patients had scans utilizing a 1.5-Tesla magnet (GE Medical Systems, Milwaukee, WI, U.S.A.). Our imaging protocol included T1-WI, T2-WI, rapid gradient echo, and 3D volumetric acquisition SPGR sequences. Closely spaced 0.3-cm cuts were utilized in all cases; special sequences including flow attenuation inversion recovery (FLAIR) have been routinely employed since 1996 and surface coils are used to further define suspect areas. Scans were viewed in all three planes and results confirmed independently by two pediatric neuroradiologists; the first interpretation was blind to other findings and the second occurred at a multidisciplinary case conference after reviewing data from video EEG, ictal SPECT, and PET/interictal SPECT when available. None of the patients had evidence of focal or lateralized cerebral structural abnormality, including subtle blurring of gray–white matter interface or an unequivocal thickened gyrus.

Ictal SPECT scans were obtained in patients operated after 1993, utilizing a three headed Multispect Siemens' Medical System (Hoffman Estate, IL, U.S.A.). To maximize the yield, ictal SPECT capture was attempted in patients with multiple daily seizures that lasted at least 30 s and had overt clinical manifestations soon after EEG onset. Parents or caretakers were routinely at the bedside to assist in early seizure detection and a simultaneous video EEG recording allowed timing of HMPAO injection (300 microcuries/kg) in relation to the electrographic seizure onset. Interictal SPECT scans were additionally obtained in 12 patients. Ten patients were referred elsewhere for a PET study. We have been performing functional MRI (fMRI) tests since 2000 and have recently incorporated routine PET scanning and 3D spike source localization techniques into our presurgical evaluation protocols to help localize the epileptogenic foci. These data are reported elsewhere (Mirkovic et al., 2003; Medina et al., 2005).

All data were reviewed at a multidisciplinary case conference. Patients in whom noninvasive data revealed a discrete and convergent zone of abnormality remote from critical regions (e.g., interictal spikes/seizure onset over the frontal polar region and convergent, discrete hyperperfusion on ictal SPECT) underwent a one-stage resection tailored by intraoperative electrocorticography. Extraoperative intracranial monitoring was deemed necessary when the noninvasive data were inconclusive or divergent, or close proximity of the epileptogenic region to eloquent cortex prompted the need for functional mapping. Extraoperative recording was performed using subdural strip or grid electrodes; depth electrodes were employed to target deep regions, including mesial temporal lobe structures. Functional mapping was performed using paradigms described previously (Jayakar et al., 1992). Motor cortex was defined by direct electrical stimulation mapping; defining the central sulcus through evoked potentials alone is in our experience not always reliable.

Epileptogenic zone and surgical resections

The region to be resected was defined on the basis of multiple convergent findings, including seizure semiology, focal EEG abnormalities, and ictal SPECT scans. Interictal SPECT or PET findings were used mainly for corroboration or guiding grid placement and were not used per se to guide resection planes. Resection plans based on noninvasive data localization were further tailored by intraoperative electrocorticography or invasive EEG monitoring and functional mapping when required.

Scalp interictal EEG abnormalities considered significant included spikes that were consistently focal especially during wakefulness or occurred in repetitive or rhythmic runs, and focal background slowing or bursts of fast activity. Scalp ictal onsets were used mainly to lateralize or regionalize the epileptogenic zones. Ictal SPECT studies revealing a localized region of hyperperfusion unrelated to secondary propagation were considered significant (Koh et al., 1999). Cross-cerebellar hyperperfusion was regarded as corroborative data for lateralization. In particular, a single discrete region of increased perfusion with surround hypoperfusion that was convergent with other data strongly influenced our decision to plan a one-stage resection tailored by intraoperative electrocorticography.

Intracranial EEG abnormalities considered significant for defining the epileptogenic zone in patients at our center have been described previously (Jayakar et al., 1994; Jayakar, 1999; Paolicchi et al., 2000). These include the ictal onset zone defined as the region showing focal transformations into rhythmic activity, bursts of high-frequency discharges, repetitive spiking, or electrodecremental patterns. Secondary foci that consistently activate during a seizure were included in the resection if they occurred in tissue adjacent to or in regional proximity to the primary ictal focus. Rhythmical theta or delta frequency activity that occurred over widespread regions after ictal onset did not influence the resection margins. Interictal discharges were considered significant if they were consistently focal or had rhythmic features, occurred as bursts or in trains of focal fast activity, or were associated with focal background attenuation; infrequent spikes or random spikes without consistent focality were ignored as were nonspecific changes in background activity.

Resections were tailored to the epileptogenic abnormalities as defined earlier and the results of functional mapping. Special attempts were made to preserve eloquent motor, language, or visual cortex when feasible and additional direct cortical stimulation was performed intraoperatively to reconfirm functional integrity when resections encroached on critical regions. Even in preverbal patients or patients with delayed language acquisition, dominant language sites were not resected unless they resided within the “core” epileptogenic zone.

All temporal lobectomies included anterior neocortical and mesial limbic structures. Corticectomy of the temporal convexity and/or basal neocortex was performed when seizure origin was more posterior; the vein of Labbe was preserved. Extratemporal resections consisted of corticectomy or lobectomy tailored to the epileptogenic region. Anterior frontal epileptogenic regions typically consisted of medial or lateral wedge resections; posterior frontal, parietal, and occipital foci were more likely to undergo corticectomy.

The surgical resection was defined as complete or incomplete at time of surgery by the epilepsy team. Completeness implied removal of all regions of significant EEG and ictal SPECT abnormalities as defined previously. Resections were deemed incomplete if the epileptogenic region could not be completely excised and was usually attributable to proximity of the epileptogenic zone to either eloquent tissue or critical vascular structures.


Seizure status was assessed after surgery by telephone contact or visits in the outpatient clinic and the seizure outcome was classified according to Engel's classification: class 1 = Seizure-free; class 2 > 90% reduction; class 3 > 50% reduction; and class 4 < 50% reduction or unchanged.


Fisher's exact test was used to evaluate correlation between preoperative variables and postsurgical seizure outcome. For the purposes of the statistical analyses, outcome was dichotomized as favorable [class 1 + 2] or unfavorable [class 3 + 4]. Significant correlation was defined as p < 0.05.


Clinical variables (Table 1)

Table 1.  Clinical variables in the 102 patients
SexM = 60, F = 42
Age at seizure onset1 day–15.2 years [mean = 3.5 years]
Age at surgery1.5 months–21 years [mean = 10.7 years]
Seizure frequency
 Monthly 8
 <Monthly 2
Seizure type
 2° generalized32
 Status epilepticus31
 Infantile spasms 8
Language function
 Preverbal 3
Cognitive evaluation

Sex and age distribution

There were 60 males and 42 females. Age of seizure onset ranged from birth to 15.2 years (mean 3.5 years). Age at time of surgery was between 1.5 months and 21 years (mean 10.7 years). Ninety-three patients were under age 18 years, nine were between ages 18 and 21 years.


Seventy-eight patients manifested daily seizures, 14 had weekly seizures, 8 seized monthly, and 2 seized less frequently. The most frequent seizure type was complex partial in 86 patients, simple partial seizures occurred in 16, and secondary generalization was documented in 32 patients. Thirty-one patients also had episodes of status epilepticus; eight had a prior history of infantile spasms.


Neurological exam revealed postictal hemiparesis in five patients. Three patients were preverbal, language delay was observed in 41 of the remaining 99 patients. Neuropsychological testing was available in 87 patients; 54 had normal cognitive function, and 33 had significant delays.

Presurgical investigations (Table 2)

Table 2.  Presurgical investigations
TestNNormalFocalNonfocal (divergent)
Interictal spikes10267719
Scalp ictal onset102 9012
Ictal SPECT 40 31    8 (1)
Interictal SPECT 12  6 6
PET 10  6 4

Scalp EEG findings

Interictal EEGs epileptiform discharges were focal in 77 patients and nonlocalizing in 19; 28 patients also had persistent or significant background abnormality. Interictal EEGs were normal in six patients. Ictal EEGs revealed focal or regional onsets in 90 patients, 12 were nonlocalizing but none revealed more than one independent foci.

Nuclear imaging findings

Ictal SPECT scans were performed in 40 patients; 31 revealed focal hyperperfusion, 8 were nonlocalizing, and 1 was focal but divergent. Interictal SPECT scans were performed in 12 patients, 6 revealed focal hypoperfusion, and 6 were nonlocalizing. PET scans were obtained in 10 subjects, 6 revealed focal hypometabolism, 4 were nonlocalizing, and none was divergent.

Intracranial EEG recordings

The findings and utility of intracranial recordings is the subject of a separate study. To summarize, the interictal and ictal onset zones varied from subject to subject and sometimes from one seizure to another in the same subject. In patients where the noninvasive data were divergent, intracranial recordings helped to identify the ictal onset zone and allowed focal resections.

Surgical variables (Table 3)

Table 3.  Surgical variables
 One stage with EcoG22
 Two stage with subdural implantation80
Hemispheric side
 Focal corticectomy17
 Hemispherectomy 3
 Additional callosotomy 3
 Additional multiple subpial transection 6
 Nonspecific abnormal/mild dysplasia71
 Taylor-type dysplasia 6
 Hippocampal sclerosis 7

Twenty-two patients had one-stage resection guided by intraoperative electrocorticography; a two-stage procedure with extraoperative subdural/depth electrode recording was performed in 80 patients.

Resection of the epileptogenic region was judged to be complete in 64 patients; in 38 patients resections were incomplete because of involvement of eloquent cortex. Resections were deemed complete in 18 [82%] of the 22 patients who underwent one-stage procedure as compared to 46 [57%] of 80 subjects who underwent two-stage procedure with subdural monitoring.

Resections were performed in 102 patients, 54 involved the left, and 48 the right hemisphere. Focal corticectomies were performed in 17 patients, lobectomies in 49, multilobar resections in 33, and hemispherectomy in 3. Six patients also had multiple subpial transection over critical cortex, 3 patients also had anterior 2 of 3 corpus callosotomy. Resections involved the frontal lobe in 54 patients, temporal lobe in 47, and parietooccipital region in 29.

Histopathology was reported as normal in 18, minimal nonspecific abnormalities/low-grade cortical dysplasia in 71, Taylor-type dysplasia in 6; hippocampal sclerosis was observed in 7 temporal lobe specimens.

Outcome (Table 4)

Table 4.  Outcomes at 2, 5, and 10 years' follow-up
Class2 years n (%)5 years n (%)10 years N (%)
  1. p: ns; note: one patient died.

1 Seizure-free44 (44)35 (44)16 (38)
2 >90% reduced15 (14)12 (15)13 (30)
3 >50% reduced17 (17)12 (15)7 (16)
4 Unchanged25 (24)17 (22)7 (16)
Total Number1017643

At 2-year follow-up, 44 of 101 (44%) patients were seizure-free, 15 had >90% reduction, 17 had >50% reduction, and 25 were unchanged; 1 patient had died of cause unrelated to seizures or surgery. Five-year follow-up was available for 76 patients; 35 (44%) were seizure-free, 12 had >90% reduction, 12 had >50% reduction, and 17 were unchanged. A 10-year follow-up was available in 43 patients; 16 (38%) were seizure-free, 13 had >90% reduction, 7 had >50% reduction, and 7 were unchanged.

Predictive variables (Table 5)

Table 5.  Predictors of outcome at 2 years' follow-up in 101 patients
VariableTotal nSeizure-free>90%>50%<50%
MRI strength
 0 3 Tesla2812 5 5 6
 1.5 Tesla7332101219
p = ns
 2° generalized3213 4 6 9
p = ns
Interictal spikes
 Focal7545814 8
 Nonfocal26 7 3 9 7
p < 0.005
Background slowing
 Focal2716 3 4 4
p = ns
Ictal onset
 Nonfocal12 3 3 1 5
p = ns
 Focal3214 4 5 9
 Nonfocal 8 3 3 1 1
p = ns
 One stage2212 2 5 3
 Two stage7932131222
p = ns
 Complete63351111 7
 Incomplete38 9 4 618
p < 0.0005
 Temporal4722 7 810
 Extratemporal5422 8 915
p = ns
 Unilobar6631 91413
 Multilobar3515 6 311
p = ns

Outcomes were significantly better (p < 0.0005) following complete resections with 46 of 63 (73%) patients having favorable outcome as compared to 13 of 38 (31%) who had incomplete resections. Outcome also correlated significantly (p < 0.005) with the presence of focal interictal spike discharges on the scalp EEG convergent with the area resected. None of the other variables examined including the lobe of surgery (temporal vs. extratemporal), one-stage versus two-stage procedure, or the extent of resections (unilobar vs. multilobar) correlated with outcomes. There was no significant difference in outcome based on MRI field strength; seizure freedom was achieved in 12 of 28 (43%) patients scanned on a 0.3-Tesla MRI compared to 32 of 73 (44%) of patients scanned on a 1.5-Tesla MRI.


Our findings demonstrate that excisional surgery benefits a large proportion of patients with intractable epilepsy who lack a demonstrable structural lesion. Many patients remain seizure-free or >90% improved even 10 years later, enjoying relief during a critical phase of brain development. The challenges of localization can be overcome through the integrated use of clinical, neurophysiologic, and functional imaging data. Collectively, the integrated approach often permits adequate excision and firmly justifies surgical candidacy for these patients.

Most previous studies of nonlesional epilepsy surgeries have included adults (Siegel et al., 2001; Blume et al., 2004; Sylaja et al., 2004; Chapman et al., 2005; Lee et al., 2005; Alarcon et al., 2006). Overall, these citations report outcomes comparable to ours. For example, 9 of 24 (37%) patients in the Chapman series were seizure-free and 75% had at least a 90% reduction at 1-year follow-up; in the series of 89 patients reported by Lee et al. (2005), 47% were seizure-free and 80% had a favorable outcome. In a series restricted to children, Ramachandran Nair et al. (2007) reported seizure freedom in 36% of the 22 subjects, while 77% had good outcome. Our study is the first to demonstrate long-term benefit in a large cohort composed predominantly of children.

Compared to surgical strategies in adults, developmentally specific issues generally warrant added conservatism in children. Clinical and EEG patterns may evolve with maturation even if the seizure "source" remains constant thus making differentiation from truly multifocal epilepsy more difficult. In some children, partial seizures may masquerade as generalized epilepsy with only isolated and often subtle indications to the truly partial nature of the epileptogenic process. Further compounding factors are the difficulty in assessing the effect of surgery on the establishment of language dominance/plasticity and the often early catastrophic presentation of the epileptic disorder that challenge the surgical skills and critical care expertise needed to deal with the younger age groups.

Given the complexities of the nonlesional cohort in childhood, it is not surprising that surgical strategies vary from center to center. The pursuit of candidacy often hinges on referral patterns and is strongly influenced by the availability of collective expertise and experience in clinical, neurophysiologic (including invasive), and functional imaging interpretation. Some centers defer surgery altogether citing poor outcomes as the reason; others advocate multilobar or hemispheric resections in the belief that larger resections will help to ensure a favorable outcome. Consequently, there are no “standardized” guidelines dealing with children with nonlesional epilepsy.

Our presurgical evaluation strategy is driven by the singular goal to maximize outcome with respect to seizure control while preserving functionality at predestined locations by performing the smallest resection possible; larger resections may perhaps have resulted in better seizure control. For example, hemispherectomies are reported to achieve seizure freedom in excess of 66% of operated children (González-Martínez et al., 2005; Basheer et al., 2007). As part of the presurgical evaluation, the past history and medical records from the referral source are thoroughly examined for consistency of both seizure semiology and EEG findings over the entire duration of the child's epilepsy with particular attention to data collected at time of seizure disorder onset. When past data are inconclusive, multiple sessions of video EEG monitoring several months apart may be necessary to document consistency.

The choice of additional functional tests in our program is not investigative modality-driven but rather seeks to utilize an integrative approach to help minimize the need for invasive EEG monitoring/functional mapping. We have used ictal SPECT scans effectively for over a decade. Till recently, we sent select patients elsewhere for PET studies; interictal SPECT was done when referral for PET scans was not feasible. Three-dimensional spike source localization technique is being increasingly incorporated into our evaluation protocol (Mirkovic et al., 2003) and functional MRI has mitigated the need for invasive brain mapping in certain cases (Medina et al., 2005). Collectively, the noninvasive data often adequately lateralize the epileptogenic region and bilateral implantation is rarely required. We have not used magnetoencephalography or EEG triggered fMRI.

The only investigative variable in our series that predicted favorable outcome was the presence of focal spikes on the scalp EEG, a finding discrepant from that reported by Chapman et al. (2005). Part of the discrepancy may be explained by our efforts in defining “focality” using rather unconventional complex field and time lead analyses, techniques akin to those used in computed 3D source localization algorithms. In that sense, our finding is in agreement with the study by Ramachandran Nair et al. (2007), where localized clusters of MEG dipoles correlated with good outcomes, whereas bilateral or scattered dipoles predicted failure.

The variable that had the most predictive power for seizure freedom or >90% improvement was completeness of resection of the epileptogenic region. Completeness of resection is conceptually a composite of multiple individual test variables, and implies that success in a nonlesional case can only be expected if data from several tests converge to reveal a discrete epileptogenic region remote from eloquent cortex. Over a third of our cases who underwent a two-stage procedure had incomplete resections and lower rates of seizure freedom. Thus, if the noninvasive tests indicate a diffuse epileptogenic region encroaching on perirolandic and sylvian eloquent tissues or insular and opercular cortex, the family needs to be counselled prior to implantation that resection may not be an option at all or may be incomplete leading to a more guarded prognosis in terms of seizure freedom. We no longer offer MST as an option in our program.

The concept of nonlesional epilepsy keeps evolving with advancing imaging capabilities. Most epilepsy centers, including ours, have utilized progressively sensitive MRI imaging paradigms over the past decade and in a longitudinal study, it becomes difficult to analyze a homogeneous nonlesional patient cohort. Our study failed to show significant difference in outcome in patients scanned on a 0.3-Tesla or 1.5-Tesla MRI. This suggests that MRIs with higher field strength correlate with better outcomes only when they reveal a focal lesion—when no lesion is detectable, the MRI field strength has no significant bearing on surgical planning or outcome. Hippocampal sclerosis and Taylor-type cortical dysplasia are reliably identified on modern MR imaging; low-grade cortical dysplasia or borderline (“normal/nonspecific”) histopathology now remain the predominant substrate of nonlesional epilepsy. The widespread use of 3-Tesla MR imaging and improved sequencing will decrease the proportion of nonlesional surgical candidates still further. This trend, however, will likely be offset by the pursuit of more aggressive case selection.

As pediatric centers gain more experience with newer functional techniques, specific utility/cost–benefit studies of various investigative tools will need to develop evidence-based guidelines for optimal application of limited resources. Several studies including ours already suggest that two-stage invasive recordings can be avoided when several noninvasive functional studies are convergent. Amidst the growing optimism prompted by the success of excisional surgery, multicentric studies will be needed to analyze the long-term neurocognitive and functional benefits of various surgical strategies in children who do not have a demonstrable MR lesion.

Conflict of interest: We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. All the authors affirm full disclosure and state that they have no conflicts of interest.