Subtraction SPECT Coregistered to MRI in Focal Malformations of Cortical Development: Localization of the Epileptogenic Zone in Epilepsy Surgery Candidates

Authors


Address correspondence and reprint requests to Dr. E.L. So at Department of Neurology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, U.S.A. E-mail: eso@mayo.edu

Abstract

Summary: Purpose: To determine the extent to which periictal subtraction single-photon emission computed tomography (SPECT) may improve detection and definition of the epileptogenic zone in patients with focal malformations of cortical development (MCDs).

Methods: Subtraction SPECT coregistered to magnetic resonance (MR) images (SISCOM) were constructed for 22 consecutive patients with focal MCDs who underwent periictal SPECT injection (18 ictal and four postictal). In the 17 patients who had epilepsy surgery, concordance between the site of SISCOM localization and site of surgical resection was determined by coregistration of SISCOM images with postoperative MRIs.

Results: SISCOM images were localizing in 19 (86%) patients, including eight of the 10 with nonlocalizing MRI. Concordance of SISCOM localization was 91% with MRI localization, 93% with scalp ictal EEG localization, and 100% with intracranial EEG localization. Eight patients whose SISCOM localization was concordant with the surgical resection site had lower postoperative seizure frequency scores (SFSs; p = 0.04) and greater postoperative improvement in SFSs (p = 0.05) than the nine patients whose SISCOM was either nonconcordant or nonlocalizing. On multiple regression analysis, a model combining SISCOM concordance with surgical resection site and extent of MRI lesion resection was predictive of postoperative SFS (R2= 0.47; p = 0.03).

Conclusions: Periictal subtraction SPECT using the SISCOM technique provides useful information for seizure localization in patients with focal MCDs, even when MRI is nonlocalizing.

The importance of malformations of cortical development (MCDs) as a common cause of medically refractory focal epilepsy has increasingly been appreciated, particularly since the development of high-resolution magnetic resonance imaging (MRI) (1). The derangement of cortical structure associated with MCDs can be relatively subtle and difficult to detect with computed tomography (CT) or earlier MRI methods (2–5). Moreover, even with state-of-the-art “seizure protocol” MRI, focal MCDs may remain undetected until histologic examination (1,6). Evidence is mounting that structural abnormalities visible on MRI often underestimate the extent of MCDs (7).

MCDs have been classified according to MRI findings as “generalized,”“unilateral hemisphere,” and “focal” (8). Focal MCDs, including focal cortical dysplasias (FCDs), focal subependymal heterotopias, polymicrogyria, and schizencephaly, are the malformations most frequently associated with medically refractory focal epilepsy (2,3,8). Tuberous sclerosis lesions are classified as focal MCDs because clinical information and imaging studies are not useful in differentiating them from FCDs, and it is difficult to distinguish between the two types of lesions pathologically (9). Evidence indicates that patients who have resective epilepsy surgery for focal MCDs have a worse seizure outcome than do patients who have surgery for other focal lesional epilepsy syndromes (10). This may be due partly to dysplastic tissues being more extensive than is apparent on MRI (1). Conversely, the epileptogenic zone may be more restricted than the extent of MRI abnormalities because some patients become free of seizures even when their large or multifocal MCDs are not resected totally (1,2,11). Because the concordance between MRI-defined lesion and epileptogenic zone is not absolute in patients with MCDs, many patients require the implantation of intracranial electroencephalographic (EEG) electrodes before proceeding to surgical resection. Noninvasive methods of defining the epileptogenic zone would be a major advance in the preoperative evaluation of these patients.

Periictal single-photon emission CT (SPECT) is used widely in preoperative localization of intractable focal epilepsy. SPECT has several advantages over positron emission tomography (PET). SPECT is relatively inexpensive to perform, is widely available, and is more practical for periictal studies. Ictal SPECT particularly is highly sensitive and specific in lateralizing seizure onset in medically refractory temporal lobe epilepsy, especially when a focal epileptogenic structural abnormality is present (12–14). More recent studies have shown that subtraction SPECT coregistered to MRIs is valuable for evaluating nonlesional and extratemporal epilepsies in which seizure onset is typically difficult to localize (15–18).

The value of subtraction periictal SPECT in the preoperative evaluation of patients with focal MCDs includes confirming that an MRI lesion is epileptogenic, localizing the epileptogenic zone in patients who have multifocal lesions or no obvious focal MRI lesion, and guiding the extent of intracranial electrode implantation and surgical resection to optimize surgical outcome. However, the clinical usefulness of periictal SPECT has not been evaluated in a large series of patients with focal MCDs. The purpose of our study was to evaluate the localization rate and prognostic significance of subtraction SPECT coregistered to MRIs (SISCOM) in patients with focal MCDs who were evaluated for medically refractory epilepsy.

METHODS

Patient selection

The patients in our study were part of a cohort of 257 patients who had periictal SPECT injections performed as part of their evaluation for intractable partial epilepsy over a 5-year period (1993–1997) at Mayo Clinic, Rochester, Minnesota (17,19,20). The study was approved by the Mayo Foundation Institutional Review Board. During the study period, 22 patients were identified to have focal MCDs. Patients were considered to have an MCD on the basis of either a pathology examination of a resected specimen obtained at epilepsy surgery or, if the MCD was not resected, a typical MRI appearance. MCDs could include FCD, focal subependymal heterotopias, polymicrogyria, or schizencephaly (2,3,8). Patients with tuberous sclerosis were included because the imaging and histopathologic features of cortical tubers are similar to those of FCD, and distinguishing between them is often difficult (8,21).

Accepted diagnostic criteria were used to determine that a patient had an MCD on MRI or pathology examination (or both). The MRI features were focal or multifocal lesions with thickening of the cortical gray matter, blurring of the gray–white interface, disruption of normal gyral patterns, and increased T2 signal in the gray matter (6,7). On pathology examination, the specimens were required to show thickening of the cortical gray matter, blurring of the gray–white interface, with the histologic findings of large atypical neurons in the cortex, usually with haphazard polarity, disruption of the normal cortical lamination, and white matter neurons. As well, in patients determined to have FCD, “balloon cells” with large, round nuclei and prominent nucleoli and abundant pale eosinophilic cytoplasm were seen (9).

SPECT injection and acquisition methods

Trained EEG technicians performed the radiotracer injections during typical seizures while the patients were undergoing inpatient video-EEG monitoring. Interictal studies were performed in standard ambient room lighting with the patient's eyes open and ears unplugged after the patient had been seizure free for ≥24 h. The radiopharmaceutical used was 99mTc-HMPAO in nine (41%) patients and 99mTc-bicisate in 13 (59%). The injected radioisotope dose for both studies was ∼20 mCi.

The SPECT images were acquired with the same camera within 2–3 h after radioisotope injection, with an identical protocol used for both studies. A dual-headed gamma camera system (Helix system; Elscint Inc., Haifa, Israel) equipped with ultrahigh-resolution fan beam collimators was used. The data were acquired in a 128 × 128-byte matrix over 360 degrees, with 120 views obtained at 3-degree intervals by using a circular orbit. For all studies, the gamma camera performed 5 × 360-degree orbits with 3 s per view. The energy setting was 140 keV, with a 15 to 20% window. Transaxial images were reconstructed with a Metz filter [power, 3; full width at half maximum (FWHM), 6 mm] rebinned into a 64 × 64 matrix with a ×2 zoom. Attenuation correction (μ= 0.12 cm) was applied, and a standard series of contiguous images was created in the transaxial, coronal, sagittal, and transtemporal planes. The reconstructed system resolution was ∼7 mm FWHM and was composed of cubic voxels with dimensions of 3.6 or 4.4 mm (depending on image size).

The onset and end of a seizure were defined as in our previous studies (17,19,20). Seizure onset was defined as the time of the earliest indication of a warning (verbalized or pushing the call button) or abnormal movements, behavior, or impaired awareness. The end of a seizure was defined as the time when ictal movements or behavior ceased. When the start and the end of the seizure could not be established confidently from clinical features, the ictal EEG was reviewed for the beginning and end of the rhythmic seizure discharge. The time of the injection was defined as the time when the plunger on the syringe containing the radiotracer was fully depressed.

MRI acquisition

MRI was performed according to a standardized seizure protocol with a 1.5-T Signa scanner (GE Medical Systems, Milwaukee, WI, U.S.A.) (5). This protocol included a spin-echo T1-weighted “whole brain” volumetric series consisting of 124 contiguous slices at 1.5- or 1.6-mm thickness, acquired either coronally or perpendicular to the long axis of the hippocampal formation. Axial T1 images, coronal and axial T2 and proton density images, and coronal fluid-attenuated inversion recovery (FLAIR) images were acquired with a slice thickness of 3 mm, with a 2-mm interslice gap.

A baseline scan was obtained for all patients, and those who proceeded to surgery had an MRI study 3 to 6 months after surgery to assess the anatomic site of the resection and the completeness of lesion removal.

Postacquisition imaging processing

Postacquisition processing of SPECT and MR images was performed off-line on a Unix-based workstation with the aid of commercially available image-analysis software packages (ANALYZE 7.5 and ANALYZE/AVW; Biomedical Imaging Resource, Mayo Foundation, Rochester, MN, U.S.A.). Hyperperfusion SISCOM images were created according to the method previously described and validated by our group (22,23). For this method, the mean intensities of the periictal and interictal SPECT images are normalized to a standard value (i.e., 100). The interictal image is then coregistered and transformed to the three-dimensional space as the ictal image by using a voxel-based matching technique (automatic image registration) (24). The two normalized and coregistered SPECT images are then subtracted pixel by pixel to create a “subtraction” image that then is thresholded to 2 SDs. Next, the thresholded subtraction SPECT image is coregistered to the preoperative MRI by using a surface-matching technique to allow anatomic localization of the regions of periictal blood-flow change (25).

For patients given the radiopharmaceutical injection postictally, hypoperfusion images were created (in addition to the hyperperfusion images) by multiplying the pixel values of the subtraction SPECT by “−1” before thresholding, so that the highest pixel values in these images were those with the greatest decrease in intensity from the interictal study to the ictal study (i.e., maximal relative hypoperfusion) (20). The hypoperfusion images were thresholded to both 1 and 2 SDs (20).

For patients who subsequently had epilepsy surgery, the preoperative thresholded subtraction SPECT images also were coregistered to the postoperative MRI (17). This allowed an assessment of the anatomic correlation between the site of the subtraction SPECT cortical focus and that of the surgical resection.

Blinded review of SISCOM images

The SISCOM images were reviewed independently for regional cortical localization by two primary reviewers (B.P.M. and E.L.S) who were blinded to the clinical data, results of other tests, and surgical outcome. Reviewers also were unaware that the patients in this study had MCD. The reviewers were required to localize the images to one of 18 possible cerebral regions (i.e., either right or left frontal, frontotemporal, temporal, temporoparietal, parietal, occipitoparietal, occipitotemporal, or occipital region) or to classify the images as “nonlocalizing.” The final determination was based on the agreement of the reviewers on the localization or nonlocalization of the images. If the two reviewers disagreed, a third blinded reviewer (M.F.H.) analyzed the images. Agreement of the third reviewer with one of the primary reviewers served as the final determination. If the third reviewer failed to agree with either primary reviewer, the study was considered nonlocalizing.

The reviewers were told the duration of the seizure and the timing of the injection. If the injection was ictal, only the 2-SD hyperperfusion images were reviewed. If the injection was postictal, both the 1- and 2-SD hypoperfusion images and the 2-SD hyperperfusion images were reviewed (20).

EEG and MRI localizations

An ictal scalp EEG was recorded from all patients by using a 32-channel system with the electrodes arranged according to a modified 10–20 system, which included subtemporal electrodes. Thirteen (59%) patients had video-EEG monitoring after long-term intracranial electrodes were implanted (subdural strip/grid electrodes in seven patients, bitemporal depth electrodes in one, and both bitemporal depth and subdural strips in five). Localization by scalp EEG, intracranial EEG, and MRI was determined by retrospective review of the reports. The rate of agreement between the site of localization by each of these tests and by SISCOM also was assessed.

Subgroup analysis for the localization rate of the SISCOM images was performed in patients whose MRI study was nonlocalizing (i.e., no lesion or multifocal lesions) and in those whose EEG was nonlocalizing. The purpose of these analyses was to determine whether SISCOM provided localizing value when more standard tests did not.

Correlation of SISCOM findings with surgical outcome

Seizure outcome for patients who subsequently had epilepsy surgery was determined by review of the patients' medical records for the degree of seizure control at latest follow-up. As in our previous studies, the outcome was classified according to a 12-point seizure frequency score (SFS) (Table 1) (26,27]. An SFS of ≤4 (i.e., seizure free or nondisabling seizures only) was considered an excellent outcome.

Table 1. Seizure frequency scoring system used to assess postoperative seizure outcome

Seizure frequency
Seizure-frequency
scorea
  1. aPatients with a seizure frequency score of ≤4 were considered to have an excellent outcome. Modified from Engel et al. (27), with permission.

Seizure free, not taking antiepileptic drugs 0
Seizure free, need for antiepileptic drugs unknown 1
Seizure free, requires antiepileptic drugs to remain so 2
Nondisabling simple partial seizures 3
Nondisabling nocturnal seizures only 4
1–3 disabling seizures yearly 5
4–11 disabling seizures yearly 6
1–3 disabling seizures monthly 7
1–6 disabling seizures weekly 8
1–3 daily 9
4–10 disabling seizures daily10
>10 disabling seizures daily11
Status epilepticus12

For the analysis of the relation between SISCOM localization and surgical outcome, patients were divided into two groups. In the first group, SISCOM localization was anatomically concordant with the site of surgical resection (Fig. 1). In the second group, SISCOM localization and site of resection were nonconcordant. To determine whether the localizing SISCOM images were concordant with the site of surgical resection, the SISCOM images coregistered to the postoperative MRI were used. This registration was performed by first matching and transforming the postoperative MRIs into the three-dimensional space of the preoperative MRI by using a surface-matching technique. The amount of cortex removed during the surgery is small compared with the entire brain surface, and the algorithm can perform a highly accurate registration by matching the common brain surfaces between the images. These images were then presented separately to the two blinded primary reviewers. When the reviewers disagreed, the third reviewer was given the images. Postsurgical outcome was compared between the two groups according to the rate of excellent outcome, the postoperative SFS, and the degree of postoperative improvement in SFS.

Figure 1.

Left, Subtraction ictal single-photon emission computed tomography (SPECT) coregistered to the postoperative magnetic resonance image (MRI), demonstrating that the site of the surgical resection (arrow) was concordant with the site of the left frontal SPECT hyperperfusion focus. Right, Subtraction ictal SPECT coregistered to the postoperative MRI, demonstrating that the left occipital SPECT hyperperfusion focus was nonconcordant with the site of surgical resection (arrow).

The SISCOM images were all analyzed blindly after the patient had already undergone surgery. The surgical decisions, including the selection of the site of surgery, were based on information from the clinical history, MRI, EEG, ictal video-EEG, and, in selected cases, invasive EEG. SISCOM images were available during the clinical evaluation for only the last year of the study period (1997) and involved only two of the 17 patients in this series who underwent surgery (patients 8 and 14).

For patients in whom a focal lesion was detected on the preoperative MRI study, the blinded reviewers used the postoperative MRIs to determine whether the focal MRI lesion was resected completely or incompletely. Postsurgical outcomes of the two groups were compared.

Statistical methods

The Mann–Whitney U test (two-tailed) was used to detect significant differences between two ordinal or continuous variables. The Fisher exact test was used when variables were dichotomous. Multivariate regression analyses were conducted to determine whether SISCOM results (concordance vs. nonconcordance with the surgical site) or the extent of resection of the MRI-defined lesion (complete vs. incomplete resection or multifocal lesions) was independently predictive of postoperative seizure outcome. Patients included in the multivariate analyses were those who had localizing SISCOM findings and resective epilepsy surgery. Postoperative seizure-outcome variables were the postoperative SFS and the degree of improvement in SFS over the preoperative score.

RESULTS

Clinical, EEG, and MRI results

The demographic, MRI, and ictal EEG findings for the 22 patients are summarized in Table 2. The median age was 23 years (range, 1.5–56 years). Fourteen were male, and eight, female patients.

Table 2. Demographic, MRI, EEG, and SPECT details of 22 patients with MCDs who had periictal SPECT

Patient
Age
(yr)/sex
Nature of MCD
on MRI
Site of MCD
on MRI
Ictal EEG
localization
Intracranial EEG
localization
Duration of injected
seizure (s)
Injection
latency (s)
Injected
seizure type
Blinded SISCOM
localization
  1. CPS, complex partial seizure; FCD, focal cortical dysplasia; GTCS, secondarily generalized tonic–clonic seizure; HS, hippocampal sclerosis; MCDs, malformations of cortical development; MRI, magnetic resonance imaging; NH, nodular heterotopia; NL, not localizing; SPECT, single-photon emission computed tomography; SPS, simple partial seizure; TS, tuberous sclerosis.

  2. aPatient had dual pathology with coexistent MRI features of HS.

 121/MPolymicrogyriaBifrontoparietalR frontalR frontal338 38GTCSR frontal
 229/FNHR frontalNLR frontal 45 39CPSR frontal
 337/MFCDR parietalNL 50 10CPSNL
 44/FFCDL frontalL frontal 43172CPSL frontal
 523/FFCDL occipitoparietalL occipitoparietal268182GTCSR temporoparietal
 620/MFCDL temporalL temporal 91 41CPSL temporal
 734/MFCDR temporalR temporal 73 32CPSR temporal
 824/FFCDL parietalL occipitoparietalL parietal 60 30CPSL temporoparietal
 934/MFCD and NHR temporo-occipitalR temporalR temporo-occipital 76 52CPSR temporooccipital
1045/FFCD and HSL parietalaL temporal 19 29CPSL temporal
1110/MFCD and HSL occipitoparietalaNLL temporal112142CPSL temporoparietal
1218/MNone detectedNone detectedNLL frontoparietal 33 17CPSL frontoparietal
1356/MNone detectedNone detectedR frontalR frontoparietal 91 45GTCSR frontal
1416/MNone detectedNone detectedL centralL parietal 72 18CPSNL
1520/FFCDR parietalR parietalR parietal 72 63CPSR parietal
1648/FFCDL frontoparietalL temporocentralL frontoparietal 93 33CPSL frontoparietal
1724/MFCDR parietalR parietalR parietal 66 12SPSR parietal
1826/MFCDL temporoparietooccipitalL frontocentrotemporal 56 70CPSL temporoparietal
191.5/FFCD (TS)MultifocalNL 80 28CPSR temporal
2023/MFCD (TS)MultifocalL temporalL temporal 32 23CPSL temporal
212/MFCD (TS)MultifocalNLR frontal 49 49CPSNL
2211/MFCD (TS)MultifocalNL 69 14CPSL occipital

MRI detected MCDs in 19 (86%) patients, of whom four (21%) had tuberous sclerosis. Twelve patients had localizing MRI findings, with a single MCD localized to a topographically restricted brain region (10 FCD, one nodular heterotopia, and one FCD with underlying nodular heterotopia). Ten patients had nonlocalizing MRI findings, with either multifocal lesions (seven patients) or no lesion detected (three patients). Two of the patients with multifocal lesions (patients 10 and 11) had dual pathology, with both a region of FCD and hippocampal sclerosis; one patient had bifrontal regions of polymicrogyria; and four had multiple lesions consistent with FCD associated with tuberous sclerosis. In the three patients without MRI-detected abnormality, histopathologic examination of the resected specimens demonstrated changes typical of MCD (28).

Seizure onset on the scalp EEG was localizing to a focal brain region in 15 (68%) patients. Intracranial EEG recordings were localizing in all 13 patients who had implanted intracranial electrodes.

SPECT injection details

The details of the seizures in which the periictal SPECT radiotracer was injected are summarized in Table 2. Injections were ictal (i.e., during the clinical or electrographic seizure activity) in 18 patients and postictal in four. The median length of the “injected seizure” was 67.5 s (range, 19–338 s), and the median time from seizure onset to injection was 35.5 s (range, 10–182 s). The injected seizure types were complex partial in 18 patients, secondarily generalized in three, and simple partial in one.

SISCOM localizations

The first blinded reviewer determined that SISCOM images were localizing in 20 (91%) patients, and the second reviewer, in 16 (73%). After the disputed cases were analyzed by the third blinded reviewer, the images were classified as localizing in 19 (86%) patients. SISCOM interpretations of the two primary reviewers were in agreement for 17 (77%) patients, with localization to the same brain region in 15 patients and nonlocalization in two. The primary reviewers did not agree on the interpretation of SISCOM findings in five patients. For four of these five patients, one reviewer determined that SISCOM images were localizing, and the other reviewer determined that they were nonlocalizing. In the other patient, the reviewers localized the SISCOM abnormality to different brain regions (temporal vs. temporoparietal).

The sites of the final SISCOM localizations for each of the 22 patients are detailed in Table 2. Of the 12 patients with a single focal MCD localized to a topographically restricted brain region, SISCOM images were localizing in 11 (92%). In 10 (91%) of the 11 patients, the site of SISCOM localization was anatomically concordant with the site of the MCD. Of the 10 patients with nonlocalizing MRI findings (i.e., multifocal lesions or nonlesional), SISCOM images were localizing in eight (80%). Two patients (patients 10 and 11) had dual pathology (i.e., both MCD and mesial temporal sclerosis) on MRI. In one, the SISCOM was localizing to the left mesial temporal lobe rather than the site of the MCD (left parietal). The other had a large region of localization involving both the left mesial temporal lobe and the site of the MCD (left occipitoparietal).

SISCOM images were localizing in 14 (93%) of the 15 patients who had focal seizure onset identified on scalp ictal EEG. In 13 (93%) of the 14 patients, the site of the SISCOM localization was concordant with that of the ictal EEG. In the seven patients with nonlocalizing scalp ictal EEG, SISCOM images were localizing in five (71%). Of the 13 patients who were studied with intracranial electrodes, SISCOM images were localizing in 11 (85%), and all were anatomically concordant with the site of seizure onset on the intracranial EEG recordings.

The only patient (patient 5) whose SISCOM finding was discordant with both MRI and scalp ictal EEG had a late SPECT radiotracer injection (182 s after seizure onset) (Table 2).

Epilepsy surgery outcomes

Seventeen patients had resective epilepsy surgery and have had ≥12 months of postoperative follow-up. Eight patients had a single focal lesion identified on MRI, six had multifocal lesions, and three had no lesion detected (Table 3). The site of the resection was the frontal area in five, temporal in six, parietal in four, frontoparietal in one, and occipital in one. Median postoperative follow-up was 22 months (range, 12–71 months). Eight (47%) patients had an excellent outcome with respect to seizures, whereas 14 (82%) had an improvement of ≥2 points on the SFS scale, which equates to a 75% reduction in seizure frequency.

Table 3. Site of surgical resection, histopathology, and seizure frequency for 17 patients who had resective surgery
Patient no.Site of surgica resectionConcordant with SISCOM localizationHistopathology of resected specimenSFSOutcomea
PreopPostop
  1. ATL, anterior temporal lobectomy; FCD, focal cortical dysplasia; HS, hippocampal sclerosis; NL, nonlocalizing; Preop, preoperative; Postop, postoperative; SFS, seizure frequency score; SISCOM, subtraction single-photon emission computed tomography coregistered to magnetic resonance images; TS, tuberous sclerosis.

  2. aExcellent outcome: postoperative SFS of ≤4; improved outcome: improvement in SFS of 2 (i.e., >75% improvement in seizure frequency) (26,27).

  3. bPatient 10 had dual pathology, with signs on magnetic resonance imaging of both HS and a parietal FCD. The HS, but not FCD, was resected.

  4. cPatient 11 had dual pathology (HS and a temporal neocortical FCD), and both were resected.

 1R frontalYesMultifocal polymicrogyria 74Excellent
 2R frontalNoHeterotopic tissue with gliosis105Improved
 3L parietalNLFCD 82Excellent
 4L frontalYesFCD 95Improved
 5L occipitalNoFCD 71Excellent
 6L temporalYesFCD with “bizarre” cells 81Excellent
 7R temporalNoFCD 86Improved
 8L parietalNoFCD 71Excellent
 9R temporalYesFCD 97Improved
10L temporal (ATL)YesHS and neocortical gliosisb 96Improved
11L temporalYesFCDc 80Excellent
12L parietalYesFCD with rare ganglion cells105Improved
13R frontoparietalYesFCD 83Excellent
14L parietalNLFCD 91Excellent
19L frontalNoFCD (TS) 88Poor
20L temporal (ATL)NoFCD (TS) 88Poor
21R frontalNLFCD (TS) 88Poor

Of the 17 surgical patients, 14 had localizing SISCOM images. When the SISCOM images were coregistered to the postoperative MRI, the region of hyperperfusion was determined to be concordant with the site of surgical resection in eight patients. In two, the hyperperfusion region was completely included within the surgical-resection site, and in the other six, it extended beyond the resection margins. In six patients, the SISCOM hyperperfusion region did not overlap the resection site at all (i.e., “nonconcordant”). All these “nonconcordant” patients were operated on before SISCOM images were available for the surgical decision-making process at Mayo Clinic. In Table 4, seizure outcome is compared between the eight concordant patients and the six “nonconcordant” patients. Those with concordant SISCOM localization had a lower postoperative SFS (p = 0.04) and tended to experience greater postoperative improvement in the SFS (p = 0.05). The rates of “excellent” did not differ significantly between the two groups (p = 0.14). The median duration of postoperative follow-up did not differ significantly between the groups (19 vs. 24 months, p = 0.89, Student's t test).

Table 4. Postoperative outcome of 14 patients with localizing SISCOM according to concordance or nonconcordance between SISCOM and site of surgical resection


SISCOM vs.
surgical site

Postoperative
SFS, median
(range)

Excellent
outcome, no.
patients (%)
Postoperative
improvement in
SFS,a median
(range)
  1. SFS, seizure frequency score; SISCOM, subtraction single-photon emission computed tomography coregistered to magnetic resonance images.

  2. aDecrease in SFS from preoperative to postoperative assessment.

  3. bp = 0.04, Mann–Whitney U test.

  4. cp = 0.14, Fisher's exact test.

  5. dp = 0.05, Mann–Whitney U test.

Concordant (n = 8)3.5 (0–6)b5 (63)c5 (3–5)d
Nonconcordant (n = 6)6.5 (1–8)b1 (17)c2 (1–8)d

Outcome according to the concordance or nonconcordance of SISCOM result with the surgical site also was compared in the subgroup of nine patients with nonlocalizing MRI findings. Although the number of patients is small, the results are similar to those for the whole group of MCD patients who had surgical resection. The five patients with concordant SISCOM had a lower median postoperative SFS (4 vs. 8) and a greater postoperative improvement in SFS (5 vs. 0) than the four patients with nonconcordant SISCOM (both p = 0.05). Three of the concordant group and none of the nonconcordant group had an excellent outcome.

Postoperative outcome is compared in Table 5 according to postresection MRI findings. Patients with complete resection of the MRI-defined lesion had a lower postoperative SFS and a greater postoperative improvement in the SFS than did patients whose postresection MRI study showed residual lesion or whose preoperative MRI study showed either no lesion or multifocal lesions (p = 0.05). The difference in rates of excellent outcome was not significant between the two groups.

Table 5. Postoperative outcome according to postresection MRI findings in 14 patients with malformations of cortical development detected on preoperative MRI
 
Postoperative
SFS, median
(range)

Excellent
outcome, no.
patients (%)
Postoperative
improvement in
SFS,a median
(range)
  1. MRI, magnetic resonance imaging; SFS, seizure frequency score.

  2. aDecrease in SFS from preoperative to postoperative assessment.

  3. bp = 0.05, Mann–Whitney U test.

  4. cp = 0.24, Fisher's exact test.

  5. dp = 0.05, Mann–Whitney U test.

Complete resection of lesion (n = 4)1 (1–5)b3 (75)c6.5 (4–7)d
Residual lesion/multifocal lesions (n = 10)6 (0–8)b3 (30)c2.5 (0–8)d

The results of multivariate logistic regression analyses of the 14 patients with localizing SISCOM images are summarized in Tables 6 and 7. A model consisting of SISCOM concordance with the resection site and the postresection MRI finding as independent variables was predictive of both postoperative SFS (R2= 0.47; p = 0.03) and postoperative improvement in SFS (R2= 0.68; p = 0.03). A trend was noted for the partial correlation coefficients for both the SISCOM and postresection MRI results to be significant in postoperative SFS (p = 0.097 and p = 0.066, respectively) and in postoperative improvement of SFS (p = 0.064 and p = 0.111, respectively).

Table 6. Multivariate regression analysis of SISCOM and postresection MRI findings for prediction of postoperative outcome by seizure frequency score in 14 patients with localizing SISCOM findings
 Postoperative seizure
frequency score
(Rmath image= 0.47; p = 0.03)
Independent variablesBetabSEcp Value
  1. MRI, magnetic resonance imaging; SISCOM, subtraction single-photon emission computed tomography coregistered to MRI.

  2. aR2, coefficient of multiple determination of the model.

  3. bBeta, standardized regression coefficients for each independent variable.

  4. cSE, standard error for each independent variable.

  5. dSISCOM findings, concordance with resection site vs. nonconcordance.

  6. ePostresection MRI findings, complete lesion resection vs. postoperative residual lesion, multifocal lesions, or indeterminate (no lesion on preoperative MRI).

SISCOM findingsd0.410.230.097
Postresection MRI findingse0.460.230.066

DISCUSSION

Our study represents the largest series to evaluate specifically the clinical usefulness of and relation of periictal SPECT to surgical outcome in patients with medically refractory epilepsy due to focal MCDs. Previous publications of periictal SPECT in MCDs consisted of isolated case reports (29,30), small series of MCD patients (11,31–33), or a small number of MCD patients included in a larger series of patients with various other cerebral lesions (15,17,34–41). Only one previous study has had >10 MCD patients (6). In that study, ictal hyperperfusion was observed in association with MCDs in 28 (82%) of 34 patients; however, the review of SPECT images was not blinded, no comparison was made between the SPECT findings and those of other techniques (e.g., EEG or MRI), and the relation to postsurgical outcome was not assessed.

Our study provides evidence that periictal SPECT analyzed with the SISCOM technique is highly sensitive and specific for localizing the epileptogenic zone in MCD patients. The 86% rate of SISCOM localization in MCD patients is comparable to the results of our previous studies involving patients with various other cerebral lesions associated with medically refractory epilepsy (15,17). Of particular clinical importance is our finding that periictal SPECT provides localizing information in a high proportion of patients for whom MRI is nonlocalizing (i.e., multifocal lesions or no definite lesion). Patients with nonlocalizing MRI findings often require intracranial electrode implantation and EEG recording before epilepsy surgery can be performed. Before being considered cost effective in the presurgical evaluation of medically refractory epilepsy, ancillary tests such as periictal SPECT must be shown to provide information that is additional to, rather than redundant with, information from more routine tests such as EEG and MRI. The complementary value of SISCOM to MRI and ictal EEG has been demonstrated in our study by the high localization rates in patients whose MRI or scalp ictal-EEG study was nonlocalizing (80% and 71%, respectively). The results attest to the value of periictal SPECT with SISCOM analysis in increasing the number of MCD patients whose epileptogenic zones can be localized noninvasively.

The specificity of SISCOM in localizing seizure onset in MCD patients is supported by the >90% anatomic concordance with the MRI-defined lesion and with seizure localization by scalp and intracranial EEG recordings. The only patient whose SISCOM localization was discordant with both MRI and scalp ictal EEG had a late SPECT radiotracer injection of 182 s after seizure onset. The rhythmic ictal EEG discharge in this patient commenced at the left occipital region, which harbored an FCD visible on MRI. The seizure discharge rapidly propagated to the contralateral temporal lobe, where it remained localized for almost 4 min before becoming generalized. The findings in this patient demonstrate the importance of obtaining early ictal SPECT injection times (15,42) and interpreting SPECT images in the context of video-EEG information derived from the “injected seizure” (43).

The outcome of epilepsy surgery for patients with MCDs generally is reported to be poorer than that for patients with other focal lesional epilepsy syndromes. Only two of the 26 patients reported by Andermann and Palmini (10) became seizure free postoperatively, and only nine had a >90% decrease in seizure frequency. A potential explanation for the poor postoperative outcome for MCD patients is that dysplastic cortex extends beyond the margins of surgical resection. MCDs are often large, multifocal, and more extensive than the area of obvious abnormality visible on MRI. Andermann and Palmini (10) found a correlation between postoperative outcome and the extent of MCD resection. We also found that patients with complete resection of the MRI-defined lesion had a better outcome. Histopathologic examination of FCD specimens in 15 of 18 patients has been reported to show dysplastic tissue in areas beyond the margins of the MRI lesion (7). Therefore, diagnostic modalities that provide information that complements MRI findings would be valuable in preoperative planning and potentially could improve postoperative outcome. In support of this, Kuzniecky et al. (11) described two patients who had extensive CDs and large areas of epileptogenesis identified on scalp and subdural EEG recordings. Resection of localized regions of ictal SPECT hyperperfusion rendered the patients seizure free. In our present study of 22 patients, we were able to prove statistically that the results of periictal SPECT using the SISCOM technique are important for the postoperative prognosis of MCD patients.

The results of our present study provide additional evidence that periictal SPECT analyzed with the SISCOM technique may be valuable in guiding the intralobar extent of surgical resection in MCD patients. Patients whose SISCOM localization was concordant with the surgical site had a lower postoperative SFS than did patients whose SISCOM localization and surgical site were nonconcordant. Moreover, the significance (p = 0.05) for the correlation between the concordance of SISCOM localization and surgical site and greater postoperative improvement in the SFS was borderline. The difference in rates of excellent outcomes did not attain statistical significance, likely because the number of patients in each group was relatively small. Nonetheless, all eight patients whose SISCOM localization was concordant with the surgical site had a ≥75% reduction in seizure frequency after surgery, strongly suggesting that the surgical resection involved part of the epileptogenic zone in all the patients.

The prognostic value of SISCOM localization also is observed in patients with nonlocalizing MRI findings, despite an even smaller number of patients in this subgroup. The value of SISCOM localizing information is demonstrated further by the results of multivariate analyses, which showed that the concordance of SISCOM localization with the site of surgical resection and the extent of lesion resection are independently prognostic of postoperative outcome. Additional studies are needed to determine the role of SISCOM in the delineation of the precise extent of surgical resection required in MCD patients. However, as a result of our experience, SISCOM findings are now routinely used as part of the surgical decision-making process in our epilepsy programs to plan the site of intracranial electrode implantations and surgical resection.

We previously demonstrated that the sensitivity, specificity, and interobserver reliability of SPECT studies analyzed with the SISCOM technique are superior to those of the visual comparison of periictal and interictal SPECT images (15,20). The visual comparison of SPECT images is potentially more difficult in patients with MCDs than in patients with other focal epilepsy syndromes. Baseline interictal images often show prominent focal hypoperfusion in the region of the MCDs (44,45), although interictal focal hyperperfusion has been observed occasionally (46). This variability in interictal perfusion reinforces the importance of closely comparing periictal SPECT images with interictal images—ideally with the aid of subtraction and MRI coregistration techniques such as the SISCOM technique—to minimize misinterpretations.

In conclusion, the results of our study demonstrate that periictal SPECT analyzed with the SISCOM technique provides valuable information for localizing the epileptogenic zone in medically refractory epilepsy due to MCDs, even when MCD lesions are multifocal or not apparent on MRI. In addition, our results suggest that SISCOM may provide prognostically important information about the intralobar site of surgical resection complementary to that of MRI. The surgical approach to optimize postoperative seizure outcome in MCD patients may be to resect as much of the abnormal SISCOM focus as possible, in addition to maximally resecting the MRI lesion. However, because of the limited number of patients, this could not be formally evaluated in our study. Only two patients had a complete resection of the SISCOM focus (patients 1 and 6), of whom only the latter also had a complete resection of the MRI lesion. Both patients had an excellent postoperative outcome.

Ancillary