| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Neuroradiology |
1 From the Department of Radiology, Section of Neuroradiology (K.R.M., M.E.F., G.L.K., J.D.L.) and Department of Neurology (T.C.), University of Utah School of Medicine, 50 N Medical Dr, 1A71 SOM, Salt Lake City, UT 84132; and Department of Psychology, University of New Mexico, Albuquerque (J.D.L.). From the 2001 RSNA scientific assembly. Received September 27, 2001; revision requested December 3; revision received March 4, 2002; accepted April 2. Address correspondence to K.R.M. (e-mail: kevin.moore@hsc.utah.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Twenty patients with neocortical epilepsy were evaluated with MEG, conventional brain MR imaging with a head coil, and high-spatial-resolution MR imaging with either a surface coil (n = 17) or a high-spatial-resolution birdcage coil (n = 3). Abnormal MEG foci were compared with corresponding anatomic areas on conventional and high-spatial-resolution MR images to determine the presence (concordance) or absence (discordance) of anatomic lesions corresponding to foci of abnormal MEG activity.
RESULTS: Forty-four epileptiform MEG foci were identified. Twelve foci (27%) were concordant with an anatomic abnormality at high-spatial-resolution MR imaging, and 32 foci (73%) were discordant. Results of high-spatial-resolution MR imaging were normal in eight patients, and 23 lesions were detected in the remaining 12 patients. Twelve lesions (52%) were concordant with abnormal MEG epileptiform activity, and 11 (48%) were discordant (ie, there was normal MEG activity in the region of the anatomic abnormality). At retrospective reevaluation of conventional MR images with MEG guidance, four occult gray matter migration lesions that had initially been missed were observed. An additional patient with MEG-concordant postoperative gliosis was readily identified with high-spatial-resolution MR images but not with conventional MR images.
CONCLUSION: Review of MEG-localized epileptiform areas on high-spatial-resolution MR images enables detection of epileptogenic neocortical lesions, some of which are occult on conventional MR images.
© RSNA, 2002
Index terms: Brain, MR, 13.121411, 13.121412 • Epilepsy • Magnetic resonance (MR), high-resolution, 13.121411, 13.121412
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Intractable lesional epilepsy is most commonly associated with gray matter migrational disorders (eg, polymicrogyria, cortical dysplasia, localized pachygyria, gray matter heterotopia [GMH]), gliosis, vascular malformations, or neoplasm. Conversely, patients with nonlesional epilepsy may or may not truly have no lesions, since resected tissue samples may show abnormalities at histopathologic review that are occult at MR imaging. In cases of nonlesional epilepsy in which a lesion is present but difficult to detect, more precise epileptogenic source localization may focus the MR imaging evaluation and improve the likelihood that a reader will detect the abnormality.
Clinical seizure characteristics (semiology) and surface electroencephalographic (EEG) localization are traditionally used for localization of epileptogenic zones, but these techniques offer limited temporal and spatial resolution. Additionally, both techniques are more definitive for mesiotemporal localization than neocortical localization, so invasive confirmation of a putative neocortical seizure focus with depth electrodes and subdural grids is frequently required before formal resection. The use of magnetoencephalography (MEG), a sensitive modality with excellent temporal and spatial resolution, addresses these limitations, and MEG results can be obtained and interpreted simultaneously with EEG results (4,5). The expanding availability of MEG in clinical medical centers has made it important to examine its clinical value for evaluating common neurologic disorders.
Interactive review at an independent workstation of conventional volumetric MR imaging data sets obtained with use of a head coil substantially improves the lesion detection rate by permitting directed scrutiny in optimal planes (6), but this technique is time consuming and its effective yield diminishes with study of larger brain areas. High-spatial-resolution MR imaging with surface coils (7,8) may enable detection of subtle lesions that are occult even on interactively reviewed conventional MR images, but this technique is limited by time constraints and coil coverage, making it impractical to scrutinize the entire brain in a single sitting. Therefore, use of results of a functional technique like MEG to focus evaluation of high-spatial-resolution MR images should produce the highest probability of detecting subtle anatomic abnormalities.
The purpose of our study was to determine whether (a) interictal epileptiform activity seen at MEG corresponds to anatomic abnormalities seen at MR imaging, (b) high-spatial-resolution MR imaging depicts anatomic aberrations in regions without MEG spike activity, (c) MEG-directed review of high-spatial-resolution MR images enables detection of abnormalities not apparent on conventional MR images, and (d) information provided by MEG results in a greater number of diagnosed lesions at re-review of conventional MR images.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
We evaluated a total of 160 patients with epilepsy by using MEG between November 1997 and June 2000. MEG and MR imaging were performed according to a protocol approved by the institutional review board, and all patients gave formal consent. At the time of this study, the institutional review board did not require formal approval or informed consent for a retrospective review of patient images or records. These patients’ records were retrospectively reviewed to identify those patients who had also undergone high-spatial-resolution MR imaging with either a surface coil or a proprietary high-spatial-resolution volume ("Endcap") coil to exclude coexisting mesial temporal lobe disease.
Only patients who had undergone both high-spatial-resolution MR imaging and MEG were considered for study inclusion; 101 patients who had not undergone high-spatial-resolution MR imaging were excluded. No patients were excluded because of an unacceptable MR imaging examination or an MEG examination in which artifacts precluded data interpretation. Patients who did not have a seizure during MEG (n = 3), patients in whom MEG results were nonlocalizing (n = 2), and patients who had clear-cut anteromesiotemporal foci (ie, they had the characteristic clinical seizure pattern or surface EEG findings and/or showed hippocampal sclerosis at MR imaging; n = 34) were excluded.
Twenty patients (eight men, 12 women) fulfilled all criteria. Patient clinical information is summarized in Table 1. Mean patient age was 27.4 years ± 10.0 (SD) (range, 10–45 years). Mean age at onset of seizure disorder was 11.0 years ± 6.7 (range, 4 months to 26 years), and mean duration of seizure disorder was 16.8 years ± 8.4 (range, 2–36 years). All patients were being treated with regimens of multiple antiepileptic drugs. Seizure frequency ranged from several seizures per month to multiple seizures per day. Primary clinical seizure type was partial in 18 patients and generalized tonic-clonic (from secondary generalization) in two. Four patients also had a secondary generalized tonic-clonic seizure semiology. Conventional surface EEG results (available in 18 patients) revealed focal abnormal activity in 14 patients and were normal or nonlocalizing in four patients.
|
The spontaneous electromagnetic data were visually inspected by an experienced magnetoencephalographer (M.E.F.) for epileptiform transients, and spike-wave complexes were analyzed with a multiple-dipole, spatiotemporal model. Briefly, this modeling procedure identifies the minimal set of dipoles that best account (in a statistical sense) for the measured magnetic field pattern. Each dipole was assigned an activation time course. The location of the first dipole to become active in a multiple-dipole sequence provided the spatial coordinates of the relevant trigger zone of cortical activity, with subsequent dipole sources indicating propagation sites. Only focal spike or sharp wave activity was considered epileptiform. To exclude areas of unlikely clinical importance, we required five or more recorded episodes of epileptiform activity before an area was labeled an epileptogenic focus.
MEG and MR imaging data were combined to form composite magnetic source localization images by means of alignment of MEG and MR imaging coordinate frames that were defined by a common set of fiducial marks to demonstrate the anatomic region believed responsible for generating each epileptiform event (9). The accuracy of this integration procedure has been shown to be within better than 3 mm (5). Given this small source of error, combined with the limitations of dipole modeling procedures, the accuracy with which magnetic source images enable localization of the point of generation for an interictal spike is estimated to be within 0.5–1.0 cm. In cases in which a structural lesion was identified at MR imaging, the composite MEG and MR imaging data were classified by the principal investigator (K.R.M.) as concordant if the MEG spike activity localized to within 1 cm of the lesion border and discordant if MEG activity localized to an area more than 1 cm from the nearest lesion.
Conventional brain MR imaging in all patients, performed with a 1.5-T Signa MR imaging unit (General Electric, Waukesha, Wis) and a standard head coil, included sagittal T1-weighted spin-echo, transverse T2-weighted fast spin-echo, transverse fluid-attenuated inversion-recovery, and transverse gradient-echo sequences.
High-spatial-resolution surface-coil MR imaging was performed with the same imaging unit used in conventional MR imaging, with either a dedicated 6-inch, 4-channel phased-array radio-frequency surface coil (Pathway MRI Multipurpose Flex Array coil; IGC Medical Advances, Redmond, Wash) (in 17 patients) or an Endcap birdcage high-spatial-resolution coil (in three patients). The phased-array surface coil provides excellent signal-to-noise ratio at a depth sufficient to permit the examination of deeper neocortical structures, as well as the hippocampus (10). The high-spatial-resolution Endcap coil provides a 24% higher signal-to-noise ratio at the image center and up to a 50% higher signal-to-noise ratio at the superior coil margin compared with standard birdcage head coils (11).
Two high-spatial-resolution MR sequences were performed: (a) a coronal volume T1-weighted spoiled gradient-echo sequence (repetition time msec/echo time msec, 25/9) through the whole brain, with one signal acquired, a 256 x 192 matrix, a 16-cm field of view, an in-plane resolution of 0.63 x 0.83 mm, and a total of 124 sections with a section thickness of 1.2 mm and an intersection gap of 0.0 mm; and (b) an oblique coronal T2-weighted fast spin-echo sequence (5,000/102) through the temporal lobes, with one to two signals acquired, an echo train length of eight, a 512 x 512 matrix, a 16-cm field of view, an in-plane resolution of 0.31 mm, a total of 10 sections with a section thickness of 4.0 mm and an intersection gap of 0.0 mm, peripheral gating, and flow compensation.
The conventional MR images were initially interpreted as part of clinical epilepsy examinations performed by an attending neuroradiologist. All conventional MR images were later retrospectively re-reviewed by a neuroradiologist (K.R.M.) with a Certification of Added Qualification in neuroradiology who used MEG localization data to direct his review. The high-spatial-resolution MR imaging data were reformatted and interactively reviewed by the same neuroradiologist at an independent workstation. Anatomic regions corresponding to areas of abnormal MEG spike activity were examined in multiple planes, including planes parallel and perpendicular to all suspicious gyri; this examination was followed by a general survey of the remaining brain. Review time ranged from 35 to 75 minutes per patient. All abnormal MR imaging findings were confirmed by a second neuroradiologist (G.L.K.) with a Certification of Added Qualification in neuroradiology.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Third and fourth, we sought to determine whether MEG-directed high-spatial-resolution MR imaging depicts abnormalities that are not apparent at conventional MR images and if MEG information results in a greater number of diagnosed lesions during re-review of conventional MR images. Conventional MR images in nine patients were initially interpreted as normal, and conventional MR images in eleven patients were initially considered abnormal. Retrospective reevaluation of the conventional images, in which abnormal MEG data were used to direct review, revealed four gray matter migrational abnormalities in four patients that had initially been overlooked. In eight of the nine patients in whom conventional MR images had been interpreted as normal, high-spatial-resolution MR images were also considered normal. The ninth patient (patient 20) had cortical thinning and gliosis over the middle and inferior temporal gyri at the site of surgical repair of an encephalocele that were not apparent on conventional MR images but were readily identified on high-spatial-resolution MR images (Fig 1).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Although mesial temporal resection for hippocampal sclerosis yields surgical success rates higher than 80% (2), reported outcomes for neocortical resection are comparatively disappointing (3). Some researchers attribute these relatively poorer outcomes to the inherent greater difficulty of localizing neocortical epileptogenic foci (3,12). Surface EEG is frequently nonlocalizing in patients with extratemporal epilepsy, and many such patients do not show concordant lesions at conventional MR imaging. In these patients, MEG may enable more precise localization of the seizure focus, and higher-spatial-resolution MR imaging may better delineate the actual margins of the malformation.
Studies in which poor surgical outcomes are observed are primarily of populations with nonlesional epilepsy; outcomes improve if a concordant lesion is present (3). In one study in which a disproportionately high percentage (91%) of patients had lesional epilepsy, 73.3% of 294 patients had good outcomes after neocortical resection (13). Conversely, Rasmussen (14–17) has reported postsurgical seizure cure rates ranging from 23% to 31% in 381 patients with nonlesional epilepsy. In other studies that involved substantial proportions of patients with nonlesional neocortical epilepsy, lower postsurgical seizure cure rates of 0% to 20% were observed (18–21).
Epileptogenic lesions may be acquired (postsurgical lesions, strokes, tumors) or represent congenital malformations of cortical development (22). Congenital malformations of cortical development, identified in 20% of adults and more than 75% of children who undergo epilepsy surgery (23), are recognized as a frequent cause of epilepsy (24). Not all structural lesions are epileptogenic, however, so it is imperative that the best functional localization information possible be obtained before resection of an anatomic abnormality is contemplated.
Conventional surface EEG remains the first-line modality for noninvasive evaluation of epilepsy. Results of a single interictal EEG examination are abnormal in approximately 50% of patients with epilepsy, with sensitivity rising to 80% with three or four separate recording sessions (1). Continuous inpatient video and EEG monitoring permits contemporaneous comparison of EEG tracings to observed clinical seizure patterns. This kind of monitoring can help confirm the diagnosis of epilepsy and may suggest the putative focus of seizure onset in some cases, but it requires inpatient admission.
When observations of seizures and conventional EEG data are inconclusive or nonconcordant, the standard evaluation method at most epilepsy centers is invasive monitoring, in which surgically placed subdural grids and depth electrodes are used to confirm the presumed focus of seizure onset. Invasive intracranial monitoring provides more accurate electrographic demonstration of epileptic foci than noninvasive evaluation but has inherent vascular and infection risks, which have led investigators to develop noninvasive modalities such as MEG to provide the same information (4,5).
MEG enables detection of weak magnetic signals produced by neuronal currents that may be used to localize interictal epileptogenic zones with millisecond real-time temporal resolution and 1-cm spatial resolution (4,5,25). The MEG data are co-registered onto anatomic MR imaging sections to create a composite magnetic source image in which the spike waves are projected onto the anatomic region proposed to have produced the spikes (9). Results of previous studies have established a role for MEG in the preoperative examination of patients with neocortical epilepsy (4,26,27).
Conventional multiplanar MR imaging with a standard head coil is used at many centers for detection of epileptogenic anatomic abnormalities during presurgical evaluation. However, resolution limitations make it inherently difficult to identify subtle migrational abnormalities with conventional MR imaging, even when pertinent clinical and electrophysiologic localization data are available to direct the search. Also, the margins of large malformations of cortical development may not be readily distinguishable at conventional MR imaging; this may in part explain the poorer surgical outcomes reported in patients with neocortical epilepsy.
Compared with MR imaging performed with conventional quadrature head coil designs, high-spatial-resolution MR imaging with a phased-array surface coil (8) or high-spatial-resolution Endcap volume coil (11) provides improved spatial resolution and better signal-to-noise ratio (8), enabling the reviewer to better detect focal lesions (7). Subsequent interactive review of volume MR imaging data at an independent workstation may enable the identification of subtle anatomic abnormalities that were occult on conventional MR images in standard imaging planes (6). We designed our study so that these proven methods of data analysis were used.
We recognize several possible limitations of our study. First, this was a retrospective study of a highly selected subset of patients with epilepsy. Only patients with medically intractable neocortical partial epilepsy were examined, and our results may not be applicable to the entire population of patients with epilepsy. Nevertheless, we believe this subgroup of patients with epilepsy was a particularly valid group to evaluate because these patients represent the majority of patients referred for surgery at epilepsy centers and have the highest rate of treatment failure.
Second, MEG is primarily an interictal modality. Although many patients with neocortical epilepsy demonstrate interictal activity, truly definitive localization requires identification of characteristic electrographic changes at ictal onset. Similar arguments pertain to the relevance of interictal activity detected at ambulatory and inpatient video EEG monitoring. Nevertheless, interictal data provide a good index of the epileptic "irritative" zone and are considered useful adjuncts to data gleaned at traditional noninvasive and invasive epilepsy evaluations (4,26,28).
Third, a caveat regarding the importance of peri-Rolandic MEG activity is warranted. Magnetoencephalographers disagree on the importance of these MEG spikes, especially in the absence of EEG or clinical correlates. On the basis of our observation of peri-Rolandic MEG spikes without strong EEG correlates in approximately 30% of the patients with epilepsy evaluated with MEG at our institution who do not manifest clinical signs of Rolandic seizure onset, we believe this type of activity usually represents general cortical hyperexcitability and is not necessarily an indication of a seizure onset zone. We identified 10 instances of peri-Rolandic activity in the patient cohort in the present study. Only patient 17 demonstrated peri-Rolandic activity that the MEG team considered to be of clinical import because it correlated strongly with EEG findings. It is noteworthy that during subsequent review of high-spatial-resolution MR images, this patient was found to have a concordant structural lesion.
Fourth, interactive review at a workstation is extremely time consuming. Each gyrus must be evaluated individually in nonconventional orthogonal planes, and time commitment increases and diagnostic yield diminishes with larger areas of brain studied (6). This technique is most useful when clinical or electrophysiologic data direct the search to a well-localized region (6). Finally, our results will be difficult to replicate directly at epilepsy centers that do not have access to MEG. However, growing numbers of medical centers, particularly tertiary centers where most resections of neocortical epilepsy occur, are installing biomagnetometers, making our results pertinent to all physicians caring for patients with epilepsy. Although MEG is more precise than EEG for localization of seizure onset zones, centers without MEG could use a similar imaging approach, in which results of EEG and observed clinical seizure patterns are used to direct review of high-spatial-resolution MR images. Such an approach should enable the identification of more abnormalities than imaging without such guidance.
In conclusion, surgical cure rates for neocortical epilepsy are substantially higher if a correlative anatomic lesion is detected preoperatively. MEG is a specific modality for detecting epileptiform activity that has excellent spatial and temporal resolution. Review of MEG-localized epileptiform areas on high-spatial-resolution MR images may enable detection of epileptogenic neocortical lesions, some of which are occult on conventional MR images.
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, K.R.M.; study concepts and design, K.R.M.; literature research, K.R.M.; clinical studies, K.R.M., M.E.F., T.C., J.D.L.; data acquisition, M.E.F., K.R.M.; data analysis/interpretation, all authors; manuscript preparation and definition of intellectual content, K.R.M.; manuscript editing and revision/review, K.R.M., T.C., G.L.K., J.D.L.; manuscript final version approval, K.R.M., T.C., G.L.K.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  K. Hara, F.-H. Lin, S. Camposano, D.M. Foxe, P.E. Grant, B.F. Bourgeois, S.P. Ahlfors, and S.M. Stufflebeam Magnetoencephalographic Mapping of Interictal Spike Propagation: A Technical and Clinical Report AJNR Am. J. Neuroradiol., September 1, 2007; 28(8): 1486 - 1488. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
) in the right frontal lobe, right peri-Rolandic region, and left peri-Rolandic region. (b) Transverse magnetic source image obtained caudal to a further delineates the extent of abnormal bilateral peri-Rolandic spike activity (
