Patients with medically refractory epilepsy can be considered for surgery if there is a reasonable possibility that they have a single resectable region of brain that is responsible for seizure origin (1–2). Surgical success in these patients is directly related to the ability to localize precisely the region of seizure onset (3,4). Locating and documenting this area involves a progressive series of diagnostic evaluations and tests, starting with a careful history and, in selected cases, intracranial monitoring (5). In all such evaluations, a pivotal role is played by high-resolution magnetic resonance imaging (MRI) (6,7). MRI findings are critical in defining the course of evaluation and can serve to divide patients into five categories: (I) MRI evidence of unilateral hippocampal sclerosis (8–10); (II) MRI detection of small circumscribed lesions (7,11,12); (III) MRI evidence of large gliotic or destructive lesions often involving more than one lobe (13); (IV) MRI evidence of neuronal migration disorder (14–16); and (V) normal MRI findings (7,17). In the present study, we have explored the hypothesis that in patients with normal MRI, invasive monitoring with electrode placements determined by the results of the noninvasive monitoring can localize the seizure focus and lead to successful epilepsy surgery.
Summary: Purpose: High-resolution magnetic resonance imaging (MRI) plays a crucial role in the presurgical evaluation of patients with medically refractory partial epilepsy. Although MRI detects a morphologic abnormality as the cause of the epilepsy in the majority of patients, some patients have a normal MRI. This study was undertaken to explore the hypothesis that in patients with normal MRI, invasive monitoring can lead to localization of the seizure-onset zone and successful epilepsy surgery.
Methods: A series of 115 patients with partial epilepsy who had undergone intracranial electrode evaluation (subdural strip, subdural grid, and/or depth electrodes) between February 1992 and February 1999 was analyzed retrospectively. Of these, 43 patients (37%) had a normal MRI.
Results: Invasive monitoring detected a focal seizure onset in 25 (58%) patients, multifocal seizure origin in 12 (28%) patients, and in six patients, no focal seizure origin was found. Of the 25 patients with a focal seizure origin, cortical resection was performed in 24, of whom 20 (83%) had a good surgical outcome with respect to seizure control. Six of the 12 patients with multifocal seizure origin underwent other forms of epilepsy surgery (palliative cortical resection in two, anterior callosotomy in two, and vagal nerve stimulator placement in two).
Conclusions: Successful epilepsy surgery is possible in patients with normal MRIs, but appropriate presurgical evaluations are necessary. In patients with evidence of multifocal seizure origin during noninvasive evaluation, invasive monitoring should generally be avoided.
From February 1992 to February 1999, 115 patients in the Dartmouth Epilepsy Program underwent intracranial EEG monitoring. At our program, intracranial EEG monitoring was performed in surgical candidates when prior noninvasive investigations produced discordant data or when MRI was normal or nonspecific. Of the 115 patients, 43 (37%) had normal MRI findings. These 43 patients all had comprehensive noninvasive presurgical evaluations before intracranial electrode placement including inpatient long-term video-EEG monitoring with scalp electrodes. Clinical seizure characteristics were evaluated throughout the noninvasive evaluation to help predict region or regions of seizure origin (18–26). Ictal single-photon emission computed tomography (SPECT) scanning was attempted in most patients. Interictal and ictal scalp EEG data were often not informative, particularly in patients with extratemporal seizure origin. When present and localizing, scalp EEG data were used in conjunction with clinical seizure characteristics and ictal SPECT results to plan intracranial electrode placements. Electrode placements included various combinations of depth electrodes, and subdural strip and subdural grid electrodes (Ad-Tech Medical Instrument Corp., Racine, WI, U.S.A.).
The standard MRI protocol consisted of sagittal T1, axial fast spin-echo (FSE) T2, and proton density as well as coronal FSE, and MPGR with a coronal SPGR volume (1.5-mm slices) orientated along the short axis of the temporal lobes. All preoperative MRI scans were reviewed prospectively before or during noninvasive study. The MR images were analyzed for the presence of structural abnormalities. In addition, MRIs were evaluated qualitatively for the presence of hippocampal atrophy and signal alterations indicating mesial temporal sclerosis [hippocampal anatomy on T1-weighted inversion recovery images, and signal hyperintensity on T2 fluid-attenuated inversion recovery (FLAIR) sequences](9,27,28). Quantitative analyses of hippocampal volumes were performed in two of the six patients with mesial temporal lobe epilepsy, but this did not alter the visual interpretation. The remaining four patients were studied before quantitative volumetric studies were available at our institution.
Brain SPECT scanning was performed using a Picker Prism 3000 three-headed camera; 4 × 7-min sequential acquisitions were obtained, and frames excluded in cases of patient movement. Scans were reconstructed in sagittal, coronal, and axial planes. Coronal images were registered to the patient's coronal SPGR MR scans using an automated image registration program (MMVREG) (29). After image registration, the counts in each study were normalized using the cerebellum counts as reference. Ictal minus interictal image subtraction was performed to enhance ictal focus conspicuity. The subtracted images were displayed alongside the orthogonal MRI, ictal, and interictal studies for interpretation. For ictal SPECT scans, seizures were detected by direct observation of the patient or the EEG, or both. As soon as a clinical or electrographic seizure was detected, the patient was injected with either 20 mCi (740 MBq) technetium-99m hexamethylpropylamine oxime [99mTc-HMPAO (Ceretec)], or 20 mCi technetium-99m ethylene cysteinate dimer [99mTc-ECD (Neurolite)]. Times between seizure onset and injection ranged from 3 to 48 s (mean, 17.6 s). Interictal scans were performed before or after the ictal scans.
Pathology was available from all patients who underwent resections, and the one patient who had multiple subpial transections had a biopsy taken at the time of surgery. Initial interpretation of the surgical specimens was by one author (C.H.R.) and later reviewed with a second author (P.D.W.).
Forty-three patients (24 males, 19 females) had normal MRIs. The mean age at seizure onset was 12.2 years (range, 0.5–40.3 years). The mean duration of epilepsy was 19.2 years (range, 2.9–40.6 years). The mean age at evaluation was 31.3 years (range, 16.8–51.8 years).
Seventeen (40%) patients had possible risk factors for epilepsy. Of these patients, 14 had one, and three had two risk factors. Significant birth anoxia occurred in one (2%) patient. Minor birth problems were not considered risk factors. Complicated febrile seizures occurred in one (2%) patient who had frontal lobe epilepsy. Four (9%) patients had histories of encephalitis during childhood as the almost certain cause of their epilepsy. Three (7%) had sustained anoxia. Four (9%) patients had head trauma with brief loss of consciousness. None of these patients required hospitalization for head trauma. Seven (16%) patients had a family history of epilepsy. Of these, none had more than one affected relative. Discounting the family history of epilepsy and the histories of probably insignificant head trauma, only 13 (30%) patients had reasonable risk factors.
Ictal SPECT was performed in 31 (72%) patients. Ictal SPECT could not be obtained in 12 patients despite several attempts. Of the 17 patients in whom a focal seizure origin was later identified, ictal SPECT correctly localized seizure origin in 13 (76%). In three (18%) patients, ictal SPECT was correctly lateralized but not localized, and in another, it was nonfocal. Of the 10 patients with multifocal epilepsy who had an ictal SPECT, eight (80%) had focal hyperperfusion that was concordant with one of the electroencephalographically demonstrated seizure origins, whereas two patients had nonfocal SPECT scan. Of the four patients with ictal SPECT in whom invasive study failed to localize seizure origin, three had focal hyperperfusion, and one patient showed bilateral hyperperfusion.
Results of intracranial electrode studies
The results of intracranial EEG studies are shown in Fig. 1. Focal seizure origin was ultimately defined in 25 (58%) patients, including three patients with frontal lobe epilepsy who had two intracranial EEG studies. Of these 25 patients, 14 (56%) had frontal lobe epilepsy, six (24%) had mesial temporal lobe epilepsy, and five (20%) had temporal neocortical seizure origin. Although strongly suspected in two, there were no patients with documented seizure origin in the posterior neocortex. In the 25 patients with focal epilepsy, retrospective review showed that correct lobar localization was predicted by history alone in 16 (64%) patients, after noninvasive evaluation in five (20%), and determined only with invasive monitoring in four (16%) (Table 1).
|No.||By history||After noninvasive|
|FLE||14||10 (71%)||2 (14%)||2 (14%)|
|MTLE||6||3 (50%)||2 (33%)||1 (17%)|
|NTLE||5||3 (60%)||1 (20%)||1 (20%)|
|Total||25||16 (64%)||5 (20%)||4 (16%)|
In 12 (28%) patients, invasive monitoring demonstrated multifocal seizure origin. Of these, multifocal epilepsy could have been predicted after noninvasive study in seven patients, whereas multifocal seizure origin was detected only with invasive monitoring in five patients. Localization of seizure origin failed in six of the 43 patients, presumably due to the intracranial electrodes not being close enough to the region of seizure origin (sampling error).
Surgical procedures were performed in 31 (72%) of the 43 patients. The procedures performed are outlined in Fig. 2. One patient with mesial temporal lobe seizure origin refused surgery. Surgical treatment was not recommended in 11 patients after intracranial EEG. In two patients with multifocal seizure origin, palliative surgery was performed with resection of the focus from which the majority of seizures originated. Of these, one had a palliative amygdalohippocampectomy, and the other, a resection of the supplementary motor area.
The result of surgery in the 24 patients with intracranially localized focal seizure origin is shown in Fig. 3. All patients had ≥2 years of postsurgical follow-up. Of these, 20 (83%) had a satisfactory surgical outcome, including two patients who required a more extensive reoperation after a failed focal frontal resection. Of these 20 patients, 15 were seizure free (Engel's class I), and five had rare seizures (Engel's class II) (30). Of the 14 patients with frontal lobe epilepsy, 11 (79%) had a satisfactory surgical outcome (eight patients were seizure free). All five patients with mesial temporal lobe epilepsy had a satisfactory outcome (four patients were seizure free). Four (80%) of the five patients with neocortical temporal lobe epilepsy had a satisfactory surgical outcome (three patients seizure free). In four (17%) patients, resective surgery failed to control seizures (i.e., they were classified into Engel's classes III and IV). Both patients with multifocal epilepsy who had palliative resections did poorly but for different reasons. The patient with resection of the supplementary motor area (SMA) did not have further SMA seizures but continued to have dysphasic seizures from an independent left lateral temporal focus in Wernicke's area. The second patient had no improvement after an amygdalohippocampectomy. Three patients underwent an anterior corpus callosotomy because of multifocal seizure origin (two patients) or nonlateralized frontal lobe epilepsy (one patient). All three of these had a >80% reduction in frequency of their most disabling seizure types. Four patients had vagal nerve stimulators implanted, but although early results have been promising, there has been insufficient follow-up.
Pathology was available from all 24 patients who had surgical resections. Table 2 lists the various pathologic diagnoses. More than half of the surgical specimens were considered normal (nine neocortex only, two neocortex and hippocampus, two hippocampus only). When definite pathology was found, such as the single ganglioma, MRIs were retrospectively examined, and in none was a new interpretation made.
|Microgyria and cortical dysplasia||1|
|Gliotic scar with hemosiderin||1|
Table 3 relates pathology to surgical-outcome class. More than half of the patients with class I and class II outcomes had no detectable pathology.
|Class I (15)|
|Classes III and IV|
|Class II (5)|
When the MRI is normal, localization of the brain tissue giving rise to the seizures is very challenging. This is reflected in reports of less favorable surgical outcome in these patients (4,31). For example, in a series of 135 patients with temporal lobectomy, 67 (60%) of 111 patients with MRI-identified structural abnormalities became seizure free, whereas only seven (29%) of 24 patients with normal MRI were seizure free postoperatively (4). Thus it is not surprising that the majority of recent surgical series consist mainly of epilepsy patients with MRI-identified structural abnormalities (12,19,20,31,32). Whether patients with normal MRI should even be considered for epilepsy surgery has been seriously questioned (17).
In our series of 43 patients with normal MRI, all were evaluated with long-term intracranial EEG monitoring. Of these, four had more than one invasive evaluation, with one patient having three (33). Twenty-five (58%) patients with seizure origin localized during invasive monitoring were offered resective surgery. In the 24 patients who underwent surgery and had sufficient follow-up, surgical results with respect to seizure control were satisfactory in 20 (83%), with 15 of these becoming seizure free. These results compare very favorably with reports of patients with MRI abnormalities and are considerably better than previous reports of patients with normal MRIs (4,17). The reason for these good results is almost certainly related to our aggressive approach to localizing the region of seizure origin, using both ictal SPECT and intracranial EEG, sometimes more than once.
Clinical seizure characteristics are a critical part of the presurgical evaluation in these MRI-negative patients. Familiarity with the clinical features of various types of seizures often allows correct diagnosis early in the course of evaluation (2,34). Localization based on history alone was correct in 64% of the patients whose seizure origins were ultimately localized. Clinical seizure characteristics, combined with results of other noninvasive studies, are essential to help plan intracranial electrode evaluations.
Localization failures were more common than surgical failures (42 vs. 17%). Although a localization success rate of 58% is comparable with other reports of patients studied with intracranial electrodes (35), it is obviously important to try to reduce the failure rate further. Seizure onsets in nine patients were initially not localizable because of presumed sampling error (i.e., the region of seizure origin was not adequately covered by the implanted electrode array). Three of these patients were successfully localized with a second intracranial electrode study. Another patient from this group with nonlateralized frontal lobe seizures underwent an anterior callosal section with an excellent result. Therefore sampling error ultimately accounted for five localization failures. Independent multifocal seizure origin was documented in 12 patients during intracranial monitoring. Clinically variable seizures and bilateral independent electrographic seizure build-up, or both, seen in individual patients during noninvasive video and scalp EEG monitoring, could have predicted multifocal epilepsy in seven patients. Ruling out invasive studies in these patients would have significantly reduced the rate of subsequent localization failures. The reason for pursuing invasive evaluations in these patients stems from the observation that patients with single regions of seizure origin can have multiple seizure-spread patterns and thus present a clinical pattern of multifocality (19,20). This was associated with seizure origin in the posterior neocortex. The vast majority of such patients had structural brain abnormalities on MRI and/or signs and symptoms of seizures of parietal or occipital origin (19,20). All of our patients with multifocal epilepsy had a normal MRI, and none had signs or symptoms suggesting seizure origin in the parietal or occipital lobes. Therefore, in the absence of compelling evidence for a single region of seizure origin, presumed multifocal epilepsy based on the results of noninvasive studies should be considered a contraindication to proceeding with intracranial electrode studies.
If MRI is normal, ictal SPECT scanning may provide additional data in the presurgical evaluation of epilepsy surgery candidates (36–39). Ictal SPECT in temporal lobe epilepsy has revealed hyperperfusion in 90% of cases in which the site of seizure origin has been unequivocally confirmed by temporal lobectomy (37). The localizing value of ictal SPECT has also been documented in cases of frontal lobe epilepsy (36), parietal lobe epilepsy (40), and occipital lobe epilepsy (38).
A number of studies used quantitative volume measurements of mesial temporal structures (41,42). We did not routinely perform quantitative volumetric analyses on patients with suspected mesial temporal lobe epilepsy whose MRIs were visually interpreted as normal. However, superiority of quantitatively measured volumes of mesial temporal structures over visual analysis in MRIs from patients who ultimately have successful temporal lobe surgery has not been demonstrated (41). In our series, volumetric analysis in two patients, including the one with the pathologic diagnosis of MTS, gave inconclusive results, and did not eliminate the need for intracranial EEG.
Other techniques for functional imaging consist of positron emission tomography (PET) and magnetic resonance spectroscopy (MRS). In patients with mesial temporal lobe epilepsy, interictal 18-fluordesoxyglucose (18FDG) PET often shows focal or regional hypometabolism even if MRI findings are absent (43–45). In these patients, successful seizure surgery can be performed without invasive EEG recordings when there is concordance between seizure characteristics, scalp EEG, and PET scan (46). In neocortical epilepsy, however, normal glucose metabolism is usually seen. Finally, preliminary data suggest that MRS may be helpful to localize seizure foci in patients with temporal lobe epilepsy and extratemporal epilepsy (42,47,48).
Acknowledgment: Adrian M. Siegel was supported by the Schweizerischer Nationalfonds (Swiss National Research Foundation) and the Kommission zur Förderung des akademischen Nachwuchses des Kt.Zürichs.