Neocortical abnormalities of [¹¹C]-flumazenil PET in mesial temporal lobe epilepsy

Patients and controls.

We studied 15 patients (11 women) with refractory mTLE20 who were recruited from the epilepsy clinics of the National Hospital for Neurology and Neurosurgery, Queen Square, London, United Kingdom. The median age at onset of habitual seizures was 7 years (range: 1 to 23 years), the median duration of epilepsy before the PET examination was 24 years (range: 12 to 45 years), and the median age at PET examination was 31 years (range: 19 to 49 years). Eleven of the 15 patients had a history of complex febrile convulsions in early childhood. Antiepileptic medication treatment consisted of mono- or dual-therapy with carbamazepine (12), lamotrigine (6), gabapentin (5), phenytoin (3), topiramate (1), or sodium valproate (1) ( table 1). Patients who were treated with benzodiazepines, barbiturates, or vigabatrin within 2 months of the PET examination were not included in the study because these drugs could interfere with [¹¹C]-FMZ binding.

All patients had a clear MRI diagnosis of unilateral HS according to accepted qualitative and quantitative MRI criteria4,5⇓ with unilaterally reduced hippocampal volumes corrected for intracranial volume, abnormal hippocampal volume asymmetry indices, and increased hippocampal T2 relaxation times. Patient 2 had a marginally increased T2 relaxation time in the contralateral hippocampus and a marked increase in the ipsilateral hippocampus; he is now seizure free 18 months after surgery. Patient 9 had temporal lobe atrophy ipsilateral to the EEG focus on MRI.

Thirteen healthy volunteers (two women) were studied for comparison. The median age at examination was 32 years (range: 23 to 64 years). These individuals had no history of neurologic or psychiatric disorder, were taking no medications, and had normal MRI studies. Individuals did not consume alcohol within the 48 hours preceding PET. Written informed consent was obtained in all cases according to the Declaration of Helsinki, and the approvals of local ethics committees and of the UK Administration of Radiation Substances Advisory Committee were obtained.

Voxel-based data has previously been reported in six of the patients (Patients 1 through 6)13; region-based hippocampal data has previously been reported in 12 of the patients15,21⇓ and in the 13 healthy volunteers.

EEG.

All 15 patients had prolonged video-EEG recordings with scalp and surface sphenoidal electrodes. Scalp EEG showed clear, anterior temporal interictal epileptiform activity consisting of sharp and slow wave components in all patients. Seizures were recorded in all patients. Unilateral anterior temporal ictal EEG changes, at or before clinical seizure onset, were recorded in all cases, ipsilateral to the side of the HS. Patient 2 had bilateral independent temporal interictal epileptiform activity with a left-sided preponderance (left/right ratio, 2:1). Subsequent depth EEG recordings revealed the seizure onset consistently in the sclerotic right hippocampus. Three other patients (Patients 1, 7, and 12) had bilateral temporal interictal epileptiform activity with a lateralized preponderance and unilateral ictal onset (see table 1).

Epilepsy surgery, histopathology, and surgical outcome.

The MRI finding of HS was histologically verified in the excised mesial temporal structures (hippocampus, part of the amygdala, and anterior 3 cm of lateral temporal neocortex) in all patients. Histopathologically, the presence of neurons in the white matter of the resected anterior temporal lobe was graded semiquantitatively as few neurons (unlikely microdysgenesis), a moderate amount of neurons (possible microdysgenesis), and many neurons (microdysgenesis) (see table 1). Median postsurgical follow-up was 3 years (range: 1.5 to 4 years), and outcome according to the classification of Engel22 was class IA (completely seizure free) in nine patients, class IB (occasional auras only) in four patients (Patients 3, 9, 13, and 15), class IIA (initially seizure-free, one seizure after 3.5 years) in one patient (Patient 6), and class IID (nocturnal seizures only) in one patient (Patient 12) (see table 1).

PET methodology.

Image acquisition and processing were performed as described previously.13,16,23⇓⇓ PET scans were performed with transaxial images aligned along the long axis of the hippocampus and coronal images orthogonal to it. Three hundred seventy megabequerels of high specific activity [¹¹C]-FMZ tracer7 was injected IV. Arterial blood was sampled continuously in order to determine a metabolite-corrected plasma input function. Voxel-by-voxel parametric images of [¹¹C]-FMZ volume of distribution ([¹¹C]-FMZ-V_d), reflecting binding to cBZR,24 were produced from the brain uptake and plasma input functions using spectral analysis.25

PET image analysis.

Analyze version 7.526 and Matlab (Mathworks Inc., Sherborn, MA) were used to perform image manipulation and measurements.

VOI analysis.

We aimed to quantify [¹¹C]-FMZ-binding after correction for partial volume effect in the hippocampus, amygdala, and neocortex. We used the same technique as described previously.27 The hippocampus was first delineated on MRI using accepted criteria28 and then geometrically divided along an anterior–posterior axis into three parts of equal length, each 8 to 10 mm long. A predefined template consisting of 28 extrahippocampal cortical VOI, derived from the Montreal Neurologic Institute representative brain (SPM96, Wellcome Department of Cognitive Neurology, London, UK), was first transformed into the individual’s MRI space. The individually outlined hippocampal VOI (three on each side: anterior, middle, and posterior third) were added onto the template. The high-resolution volume acquisition MRI scans were then automatically segmented into probability images of gray matter, white matter, and CSF using a clustering, maximum likelihood “mixture model” algorithm.29 The gray matter, white matter, CSF images, and VOI were coregistered with the parametric images of [¹¹C]-FMZ-V_d30 and then “blurred” to the same spatial resolution as PET by convolving the MRI with the three-dimensional point spread function of the PET scanner.31 This allows estimates of partial volume effect within the multiple VOI of homogenous tracer activity to be obtained. We report only the gray matter contributions to the activity of neocortical regions.

Voxel-by-voxel analysis.

[¹¹C]-FMZ-V_d images were also analyzed using SPM (SPM 96, Wellcome Department of Cognitive Neurology) implemented in Matlab (Mathworks Inc.). Statistical parametric maps are three-dimensional projections of statistical functions that are used to characterize significant regional brain differences in imaging data.32-35⇓⇓⇓ For the purposes of between-group statistical analyses, the [¹¹C]-FMZ-V_d images of patients with left-sided HS were reversed so that the HS appeared on the right side in all patients. The images were then transformed into a standard anatomic space.36 The procedure involves a linear three-dimensional transformation and uses a set of smooth basis functions that allow for normalization at a finer anatomic scale. Images were smoothed using a 12 × 12 × 12 (full width at half maximum) isotropic Gaussian kernel as a final preprocessing step. This spatial filter accommodates interindividual anatomic variability and improves the sensitivity of the statistical analysis.32 Each patient’s MRI scan was coregistered with his or her [¹¹C]-FMZ-V_d image and then transformed into standard space using the transformation matrix derived from the spatial normalization of that individual’s [¹¹C]-FMZ-V_d image.

Statistical analysis.

Voxel-by-voxel analysis.

Individual patients were compared with an age-matched group of 13 normal subjects with the design matrix including global cerebral [¹¹C]-FMZ-V_d as a confounding covariate of no interest; this analysis can therefore be regarded as an analysis of covariance.38 Significant differences between patients and controls were estimated according to the general linear model at each and every voxel.34 Linear contrasts were used to test the hypotheses for specific focal effects. The resulting set of voxel values for each contrast constituted a statistical parametric map of the t-statistic SPM {t}. The SPM {t} were transformed to a normal distribution, SPM {Z}, and thresholded at 3.09 (or p = 0.001 uncorrected).

For the correction for multiple comparisons, the resulting foci of significant differences were then characterized in terms of peak height (μ) and spatial extent (k). The significance of each region of [¹¹C]-FMZ-V_d difference was estimated using distributional approximations from the theory of Gaussian fields.34 For neocortical regions, the corrected threshold chosen was p < 0.05 for peak height and spatial extent. Because the hippocampus and amygdala are small structures, below 2 × full width at half maximum of our scanner resolution, [¹¹C]-FMZ-V_d differences in these regions of special interest were only corrected for peak height but not for spatial extent. Age, sex, number of complex partial seizures per year, outcome after epilepsy surgery, and the interval between last seizure and PET scan were defined as covariates of interest and tested separately for their effect on FMZ binding.

VOI analysis: absolute [¹¹C]-FMZ-V_d with correction for partial volume effect.

Individual patients with HS.

Mesial temporal structures.

After partial volume effect correction, [¹¹C]-FMZ-V_d was unilaterally reduced in the atrophic hippocampus in 14 of 15 patients with mTLE concordant to the side of the EEG focus ( table 3). Thirteen of these 14 patients showed reductions in the anterior third of the hippocampus, five of 14 in the middle third, and five of 14 in the posterior third ( figure 1). These reductions extended into the ipsilateral amygdala in four patients. There were no increases in [¹¹C]-FMZ-V_d in the mesial temporal structures (see figure 1). One patient with anterior HS on MRI (Patient 13) showed reduced [¹¹C]-FMZ binding in the ipsilateral anterior hippocampus with an asymmetry index outside the normal range, but absolute binding was within 2.5 SD of the normal control mean. No contralateral abnormalities were observed.

View this table:

• View inline
• View popup

Table 3.

Partial volume effect–corrected [¹¹C]-flumazenil volume of distribution in 15 patients with unilateral HS

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Partial volume effect–corrected absolute [¹¹C]-flumazenil volume of distribution changes in 15 patients with mesial temporal lobe epilepsy and unilateral hippocampal sclerosis. Increases are indicated in red, decreases are indicated in blue. The intensity of the colors corresponds to the number of patients in whom a change was observed (in gray). AL = anterior lateral temporal lobe; AM = anterior medial temporal lobe; TMI = middle and inferior temporal gyrus; TS = superior temporal gyrus; Fu = fusiform or occipitotemporal gyrus; PH = parahippocampal gyrus; Amyg = amygdala; Han, Hmi, Hpo = anterior, middle, and posterior third of the hippocampus; TLpo = posterior temporal lobe; OL = occipital lobe; FL = frontal lobe; GCan = anterior cingulate gyrus; GCpo = posterior cingulate gyrus; PL = parietal lobe.

Temporal neocortex.

Temporal neocortical increases in [¹¹C]-FMZ binding were observed in Patients 7 and 13 (see figures 1 and figure 2). These involved the anterior lateral temporal lobe and the temporal pole bilaterally in Patient 13 and contralateral to the side of the EEG focus in Patient 7 (see table 3). Patient 7 had additional increases in the extratemporal neocortex (see below).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Example of a patient (Patient 7) with decreased [¹¹C]-flumazenil (FMZ) volume of distribution (V_d), corrected for partial volume effect, in the ipsilateral (right) middle and posterior hippocampus, and increased [¹¹C]-FMZ-V_d in the contralateral anterior lateral temporal neocortex. On the right, a typical control scan is shown. The hippocampal decrease is evident on visual inspection. The anterior lateral temporal neocortical increase is only revealed by quantification.

Extratemporal cortex.

Patients 1 and 5 had decreases in extratemporal cortical [¹¹C]-FMZ-V_d. These involved the occipital lobe contralateral to the EEG focus in Patient 5 and the contralateral frontal lobe in Patient 1 (see table 3). Patients 7 and 8 had extratemporal cortical increases. Patient 7 had increases in the ipsilateral occipital lobe and in the contralateral posterior cingulate gyrus. Patient 8 had increases in the frontal lobe ipsilateral to the EEG focus (see table 3) No patient had both neocortical increases and decreases in [¹¹C]-FMZ binding.

Between-group analysis.

Reductions of [¹¹C]-FMZ-V_d in patients compared with controls were found in the following regions: the anterior third of the hippocampus (mean reduction −30.7%, p < 0.0001, unpaired t-test), the middle third (mean reduction −22.2%, p = 0.002), the posterior third (mean reduction −29%, p < 0.0001), and the amygdala (mean reduction −14.8%, p = 0.01) ipsilateral to the EEG focus. There were no abnormalities in temporal and extratemporal neocortex (at p < 0.001, Bonferroni corrected for multiple comparisons).

Asymmetry indices of partial volume effect corrected [11C]-FMZ-V_d.

Fourteen patients with mTLE had asymmetry indices outside the normal range for the hippocampus (12, five, and nine patients for the anterior, middle, and posterior third). Patient 9 had a further abnormal asymmetry index in the amygdala. Two patients had increased asymmetry indices for structures outside the mesial temporal cortex: Patient 9 in the fusiform gyrus (lower on side of focus) and posterior cingulate gyrus (higher on side of focus), and Patient 7 in the posterior cingulate gyrus. In the latter, the asymmetry index outside the normal range corresponded to a contralateral increase in the absolute value for [¹¹C]-FMZ-V_d, whereas in Patient 9, the asymmetry indices outside the normal range did not correspond to any changes in absolute [¹¹C]-FMZ-V_d.

Voxel-by-voxel analysis (SPM).

Individual patients with HS.

Temporal neocortex.

Temporal neocortical decreases in [¹¹C]-FMZ binding were observed in Patient 9. These involved the parahippocampal and fusiform gyrus and the posterior temporal lobe, concordant with the side of the EEG focus.

Patients 2, 7, and 8 showed increases in temporal neocortical [¹¹C]-FMZ binding. The ipsilateral middle and inferior temporal gyrus were involved in Patients 2 and 7. Patient 7 showed additional increases in the ipsilateral parahippocampal gyrus and superior temporal gyrus as well as in the inferior, middle, and superior temporal gyrus contralaterally. Only contralateral increases were observed in Patient 8 (anterior lateral temporal lobe and superior temporal gyrus).

Extratemporal cortex.

There were significant decreases in extratemporal cortical [¹¹C]-FMZ binding in two patients, with reductions in the occipital lobe ipsilateral to the side of the EEG focus (Patient 9), and bilateral reductions in the medial parietal lobe (Patient 15). No increases in extratemporal cortical FMZ binding were observed.

Covariates of interest.

Age, sex, frequency of complex partial seizures per year, and the interval between last seizure and PET scan had no significant effect on [¹¹C]-FMZ binding. There was a positive correlation between increased [¹¹C]-FMZ binding in the white matter of the temporal lobe contralateral to the resected temporal lobe with outcome poorer than Engel class IA (r = 0.64, p = 0.005, Spearman’s correlation coefficient). There was a positive correlation of the number of white matter neurons in the histology specimen, with increased [¹¹C]-FMZ binding in the white matter of the resected temporal lobe (r = 0.81, p < 0.001) and the contralateral temporal lobe (r = 0.80, p < 0.001; Spearman’s correlation coefficient).

Between-group analysis.

There was a highly significant decrease in ¹¹C-FMZ binding in the patient group involving the entire length of the hippocampus on the side of the EEG focus. No other changes were observed.

Comparison of partial volume effect–corrected VOI results and SPM results. The comparison of neocortical abnormalities is summarized in table 4.

In the two HS cases not identified by SPM (Patients 4 and 10), absolute [¹¹C]-FMZ-V_d after partial volume effect correction was significantly reduced with abnormal asymmetry indices. In Patient 13, no absolute decrease in [¹¹C]-FMZ binding was demonstrated with the VOI approach, but SPM detected significant reductions in voxels located in the anterior hippocampus. Asymmetry indices for this region were also outside the normal range.

Mesial temporal structures.

Changes of [¹¹C]-FMZ binding in mesial temporal structures were detected in 14 of 15 patients using the partial volume effect–corrected VOI approach and in 13 of 15 patients with SPM.

Temporal neocortical structures.

Five patients had significant alterations of [¹¹C]-FMZ binding in the temporal neocortex with VOI, and two of these five plus another three patients with SPM analysis had significant alterations of [¹¹C]-FMZ binding. These changes were concordant in that the patients showed either temporal neocortical increases or decreases with both methods. In only one patient (Patient 7), however, were these changes significant using both methods. Two patients with significant increases (Patients 2 and 8) and one with significant decreases (Patient 9) in SPM only showed corresponding changes in absolute values when the threshold was lowered. Patient 13 had significant increases in the partial volume effect–corrected anterolateral temporal lobe VOI bilaterally, but the corresponding changes in SPM were only significant when uncorrected for spatial extent.

Extratemporal neocortical regions.

Six patients had significant abnormalities of [¹¹C]-FMZ binding in the extratemporal neocortex identified with either VOI or SPM analysis. Four patients had abnormalities in the VOI analysis (Patients 1, 5, 7, and 8; see table 3). In three of these patients (Patients 5, 7, and 8), VOI-detected changes were similar to SPM results when the thresholds were lowered. In Patient 1, no corresponding changes were detected using SPM. In Patients 9 and 15, SPM showed significant decreases that were not reflected by any binding changes in the VOI analysis.

Methodologic considerations.

The use of both a VOI and voxel-based analysis increased the sensitivity of [¹¹C]-FMZ PET for detecting significant mesial temporal abnormalities in HS to 100%.

Partial volume effect correction is especially important when structural abnormalities are present. This is clearly the case for mesial temporal structures in mTLE. Extramesial atrophy, however, is well described.17 In our series, one patient with temporal lobe atrophy ipsilateral to the HS (Patient 9) showed ipsilateral posterior temporal and occipital decreases of [¹¹C]-FMZ binding on the SPM analysis that were not corroborated by the VOI analysis. Because our method also corrects for the thickness of the cortical ribbon, this and other differences between the two analyses can be explained by normal [¹¹C]-FMZ binding per unit of gray matter when partial volume effect correction is used, and highlights the usefulness of partial volume effect correction in extramesial neocortical structures.

For the VOI approach, we used a template of multiple cortical VOI defined in standard stereotactic space and subsequently transformed into individual MRI and PET space. The use of such a template, its automated coregistration, and the subsequent use of partial volume effect correction provides an entirely objective and observer-independent method for defining multiple neocortical VOI and quantifying [¹¹C]-FMZ-V_d.15,31⇓ This approach works satisfactorily for neocortex, but the software did not segment and normalize the basal ganglia or the hippocampus satisfactorily. The hippocampus was therefore subjectively outlined by hand in all individual subjects. Compared with the VOI approach, SPM localizes significant changes in receptor binding on a voxel level. Thus it may detect abnormalities not identified by the VOI method when these abnormalities are small in spatial extent and become averaged in large VOI or do not conform to the anatomic boundaries used in the template. However, owing to the large number of comparisons made, rigorous statistical thresholds for significance must be applied to the amplitude and extent of changes. Furthermore, the final spatial resolution of SPM is lower than in a VOI approach owing to the necessary smoothing, thus it is more affected by partial volume effect. SPM cannot distinguish loss of signal due to tissue atrophy from a true functional abnormality. The two methods are therefore complementary.

Comparison with previous findings.

We have previously demonstrated, using hippocampal VOI outlined on high-resolution MRI and coregistered to PET, that in mTLE due to HS there are decreases of [¹¹C]-FMZ binding in the sclerotic hippocampus.16 Our previous method, however, used only a box-region around the hippocampus, thus averaging low temporal and high occipital [¹¹C]-FMZ binding and leading to an underestimation of true tracer binding in the anterior hippocampus, and not allowing the investigation of extrahippocampal regions.

Using SPM 95, we previously found areas of decreased [¹¹C]-FMZ binding to be restricted to the sclerotic hippocampus in 12 patients with mTLE due to HS.13 In the current study, we found abnormalities of [¹¹C]-FMZ-V_d outside the hippocampus in five of 15 patients with HS using a more recent version of SPM (SPM 96), which contains more advanced methods of spatial normalization and more stringent criteria for correction of multiple comparisons, assessing both maximal height and spatial extent, and is therefore more sensitive. Only one of the six patients included in both this and the previous study13 using SPM 95 (Patient 2) showed abnormalities in the current SPM analysis that had not been detected using the previous version of the software. Besides the advanced statistical methods implemented in SPM 96, differences in the patient populations may thus play a role in the apparent difference in the results obtained.

Neurobiological considerations.

The exact pathologic and pathophysiologic mechanisms underlying mTLE are still under study. Taken as a group, only the reduced [¹¹C]-FMZ binding in the epileptogenic hippocampus, extending into the amygdala, was significant. However, in five of 15 patients with HS, there were additional neocortical abnormalities of [¹¹C]-FMZ binding. In contrast to previous findings,39 these did not follow a specific pattern in the extratemporal areas to suggest a predominant involvement of projection areas.

The only temporal neocortical changes in the VOI analysis were increases in the anterior lateral temporal VOI, including the temporal pole (in Patient 7 contralaterally and in Patient 13 bilaterally). In an autoradiographic study, 11 temporal lobe specimens from patients with HS were compared with six control specimens.40 The authors found decreased [³H]-FMZ binding in most hippocampal layers but diffusely increased binding in the lateral temporal neocortex in the patient group, which reached statistical significance in cortical layers V and VI. Although our group results for the anterior lateral temporal lobe VOI in the patients did not differ significantly from the control group, two patients had unilateral or bilateral increases in [¹¹C]-FMZ binding in this VOI (see figure 1). It is possible that chronic deprivation of input leads to a compensatory increase in cBZR density, or this may be an adaptive response to epileptic activity.

Other possible explanations for increased [¹¹C]-FMZ binding include an increased neuronal density or ectopic neurons bearing cBZR, as for example in microdysgenesis. Areas of increased [¹¹C]-FMZ-V_d have been demonstrated in cortical dysplasia using SPM23,41⇓ and in patients with unremarkable MRI scans.42

Recently, postsynaptic increases in the number of GABA_A receptors underlying inhibitory potentiation in the kindling model have been described.43 Such an increase in available binding sites (B_max) will lead to an increase in [¹¹C]-FMZ-V_d. Our method does not distinguish between changes of B_max and changes of the receptor affinity. Autoradiographic and histopathologic studies of sclerotic human hippocampi obtained during epilepsy surgery have showed reduced neuron counts and cBZR densities44,45⇓ and reduced cBZR density per remaining neuron, whereas receptor affinity was increased in the hilus, dentate gyrus, and subiculum.45 This suggests that functional abnormalities may be greater than structural ones in human mTLE. Similarly, a recent study using the pilocarpine model has found presynaptic and postsynaptic changes of GABA transmission46 involving changes of GABA_A β receptor subunit composition. Thus, increased density or affinity of available receptors per neuron, either on abnormal nerve cells or as an adaptive response to the abnormal neuronal activity in mTLE, may explain the observed extramesial increases of [¹¹C]-FMZ binding.

Clinical considerations.

Earlier studies showed decreases in mesiotemporal [¹¹C]-FMZ binding in eight of eight patients with TLE9 or found mesiotemporal decreases in 10 of 10 patients with mTLE, but detected no neocortical changes.10 These studies concentrated on the mesial temporal lobe and used region of interest–based analyses placing the regions directly on the PET images.

Using the same methods as described here, we have recently found increases and decreases in mesial temporal structures and in the neocortex ipsilateral and contralateral to the EEG focus in eight of 10 patients with TLE and normal structural imaging.27

Recently, [¹¹C]-FMZ-PET findings in a series of 100 patients evaluated for epilepsy surgery were reported.47 Of 35 patients with HS, reduced [¹¹C]-FMZ binding in the mesial temporal lobe extended toward the temporal pole or lateral temporal neocortex in seven patients. Extratemporal areas were not examined in these patients. We found the mesial temporal decreases extended into the amygdala in four of 15 patients in our VOI analysis and in eight of 15 patients in the SPM voxel-by-voxel analysis. We only found one patient (Patient 9) with decreased [¹¹C]-FMZ binding in the temporal neocortex using SPM, which was not corroborated by the VOI analysis. This is most likely explained by partial volume effect due to the temporal lobe atrophy in this patient.

Increases in [¹¹C]-FMZ binding in the white matter of the temporal lobe contralateral to the side of the resection found on the voxel-based assessment were correlated with a poorer chance of an Engel class IA outcome and may indicate epileptogenic abnormalities in the contralateral temporal lobe. Temporal lobe white matter increases of [¹¹C]-FMZ binding were found to be correlated with the presence of an increased number of white matter neurons in the specimen and may therefore indicate microdysgenesis. There was, however, no significant correlation between presence of white matter neurons and outcome.

In the long term, one third of patients will continue to have seizures after epilepsy surgery for HS.2 A longer follow-up than the median of 3 years in this study and a larger series may show more clearly whether patients with extrahippocampal neocortical abnormalities of [¹¹C]-FMZ binding have a poorer prognosis.

Using quantitative postprocessing of preoperative structural MRI, extrahippocampal abnormalities have been reported in 14 of 27 patients with subsequently histologically proven HS, 10 of whom did not become seizure free after anterior temporal lobectomy.48 Our findings of neocortical functional abnormalities of the GABAergic transmission give additional evidence that abnormalities may extend beyond the mesial temporal lobe.

A positive correlation of seizure frequency with the degree of cBZR reduction has been described in patients with daily, severely disabling complex partial seizures compared with patients with weekly or less frequent seizures.49 We did not replicate this finding, and [¹¹C]-FMZ binding was not correlated with the interval between the last seizure and the PET scan. However, all of our patients were medically refractory and may have been too homogenous to find any such correlation.