Abstract
Objective: Fluorodeoxyglucose positron emission tomography (FDG-PET) studies of temporal lobe epilepsy (TLE) generally report interictal hypometabolism in the vicinity of the seizure focus. Yet, other evidence suggests that interictal metabolic abnormalities might extend to remote brain areas. We used FDG-PET to evaluate metabolism in selected regions distant from the focus in TLE. Subjects: Twenty adult patients with medically intractable TLE were selected by criteria favoring a unilateral mesiobasal temporal focus. Structural imaging in this sample was normal except for medial temporal sclerosis in 13 patients. Twenty normal volunteers were controls. Design: PET imaging was performed interictally. Regional glucose metabolism normalized by global metabolism was analyzed using t tests and correlation analysis. Results: Ipsilateral to the seizure focus, metabolism was depressed compared with normal in the temporal pole (p equals 0.001), but relatively elevated in the mesiobasal region (p equals 0.005). Contralateral to the focus, metabolism was elevated in lateral temporal cortex (p equals 0.0003) and mesiobasal regions (p equals 0.0001). Metabolic correlation between ipsilateral and contralateral mesiobasal regions was similar in normal subjects (r equals 0.74) and patients (r equals 0.68). In contrast, correlations were abnormal between temporal poles and other temporal lobe subregions, both ipsilateral and contralateral to the seizure focus. Conclusions: Relative to normal values, both elevations and depressions of metabolism exist interictally in TLE. Such abnormalities, and accompanying changes in interregional correlations, may have wide spatial distribution. These findings are atypical among PET studies but are consistent with other physiologic, anatomic, and neuropsychological investigations of TLE.
NEUROLOGY 1995;45: 2212-2223
The pathophysiology of temporal lobe epilepsy (TLE) involves long-term modification of neuronal connections both within the temporal lobe and between the temporal lobe and other parts of the brain. Persistent abnormalities are evident within the epileptic temporal lobe in terms of cell number, axon pathways, and synaptic connections. [1-3] These cellular changes appear related to larger-scale ictal and interictal abnormalities of the diseased temporal lobe seen in EEG activity, cerebral metabolism and blood flow, and neuropsychological function. [4] Less impressive evidence so far has been found for persistently altered function in the epileptic brain beyond the seizure focus. Abnormal electrical activity originating in the focus obviously affects remote areas acutely during seizure generalization. The separate and clinically important issue of whether the transmission of such abnormal activity between the focus and recipient regions may lead to prolonged alterations in remote areas has received only preliminary attention. The most specific evidence for such distant effects in TLE comes from neuropsychological studies. Reports by Milner, [5] Novelly et al, [6] and Takahashi et al [7] suggest that in unilateral TLE, the interictal function of the temporal lobe contralateral to the focus is impaired. Interictal performance of frontal lobe tasks also worsens in the presence of a temporal lobe focus. [8] These studies found that such neuropsychological deficits may improve following effective excision of the temporal focus, raising the possibility of both reversible and irreversible alterations in neural circuits that include the seizure focus.
Fluorodeoxyglucose positron emission tomographic (FDG-PET) studies also support the presence of functional abnormalities in TLE beyond the seizure focus. Numerous PET studies have reported interictal hypometabolism in TLE not only in the vicinity of the electrographically defined seizure focus, but also more widely within the affected temporal lobe and in nontemporal regions of the same hemisphere (see Henry et al, [9] for example). Areas of evident hypometabolism tend to extend further than any evident anatomic abnormality; this is particularly true of medial temporal sclerosis (MTS), the lesion most characteristically associated with temporal lobe partial complex seizures. [10] Some PET studies suggest that in TLE, specific temporal lobe regions are especially prone to exhibit hypometabolism, often independently of the actual location of the focus within the temporal lobe [11,12] (see also Sloviter [1]), a phenomenon that could contribute to expansion of metabolic abnormalities. Taken together, these findings suggest that the TLE process, the etiology of which remains unknown and possibly heterogeneous, may evoke characteristic, spatially distributed metabolic abnormalities within cerebral circuits that include the diseased temporal tissue. Such responses presumably are shaped both by the intrinsic properties of each region in the network and by synaptic interconnections among the regions.
We hypothesized that in TLE, regions remote from the focus but implicated as dysfunctional by prior neuropsychological studies would display metabolic abnormalities. Thus, beyond the predominant abnormality seen in most interictal FDG-PET studies (hypometabolism ipsilateral to the focus), we suspected that metabolic abnormalities might also exist in remote areas such as the frontal cortex and/or the contralateral temporal lobe. As some temporal structures may be more prone than others to manifest metabolic abnormalities in TLE, we studied subregions within the temporal lobe. Twenty adults with strongly lateralized TLE were selected as subjects, and regional interictal metabolism measured with FDG-PET was compared with metabolism in 20 normal volunteers. The restriction of our clinical sample to cases of unilateral TLE allowed us to distinguish interictal effects exerted ipsilaterally from those exerted contralaterally to the seizure focus.
Methods.
Subjects.
Ethical permission for these studies was obtained from the Institutional Review Board of North Shore University Hospital/Cornell University Medical College. Written consent was obtained from each subject following a detailed explanation of the procedures.
Controls.
Twenty normal volunteers (10 women, 10 men; mean age plus minus SD equals 47.0 plus minus 17.1 years) were recruited as previously described. [13] These subjects were free of prior neurologic or psychiatric illness, prior neuroleptic or illicit drug use, hypertension, cardiovascular disease, or diabetes mellitus. All had a normal neurologic examination.
Patients.
Twenty outpatients (11 women, nine men; mean age 34.0 plus minus 9.8 years) with medically intractable epilepsy were selected from those referred to our PET facility from three university hospitals in the course of assessment for neurosurgical intervention (see Table 1). Twelve had no known personal risk factors for epilepsy, six had febrile seizures in infancy (one with meningitis), one had a history of closed head injury without loss of consciousness, and one had a concussion with brief loss of consciousness 18 years prior to his first seizure. The mean age at onset of seizures was 16 years (range, 2 to 44), with mean duration of illness 20 years (range, 2 to 38). All patients were classified clinically and by video-EEG criteria as having complex partial seizures of temporal lobe origin as their predominant seizure type, and adequate trials of standard medications had failed in all. All patients had MRI studies (coronal double-echo inversion recovery, GE Signa 1.5-tesla field strength) that either were normal or had findings restricted to unilateral MTS (presumptive radiologic diagnosis in 13 of 20 cases, as judged by evidence of hippocampal atrophy with diffusely increased signal). Histopathology confirmed this diagnosis in six of six cases where surgical specimens were available. To date, eight patients have progressed to surgery.
Table 1. Patient characteristics (all with partial complex seizures as predominant seizure type)
The TLE cases were further selected for evidence of seizure origin in a single temporal lobe. The final sample comprised 11 patients with left and nine with right temporal foci meeting the following criteria: Combined interictal and ictal data from video-EEG monitoring in all cases had to indicate a unilateral temporal lobe focus and agree with any lateralizing semiology of seizure onset. When available (seven cases), data from surface corticography and/or depth electrodes also had to localize the seizure focus to one temporal lobe and be in agreement with clinical and scalp EEG lateralization; in each such case, the intracranial data indicated a mesiobasal (MB) focus. For patients with unilateral MTS, this MRI abnormality had to lateralize concordantly with clinical and electrophysiologic data. Finally, where neuropsychological data were available (13 of 20 cases), predominant findings either had to match the lateralization indicated by electrographic and other clinical data or be nonlateralizing.
All patients selected according to these criteria had definite metabolic asymmetries on at least two contiguous planes of their interictal FDG-PET studies, with areas of apparent hypometabolism concordant with other lateralizing data. This semiquantitative assessment in our laboratory corresponds to detection of regional metabolic reductions of approximately 15% or greater, which generally exceed the normal range. [14]
Imaging procedure.
Patients were free of clinically observable seizures during the PET procedure and had been so for at least 24 hours beforehand (by self report and family report). They were taking their routine medications at the time of the imaging. Studies were performed on the Superpett 3000 tomograph (Scanditronix; Essex, MA). The characteristics of this instrument have been previously described. [15] This four-ring BaF2 time-of-flight, whole-body tomograph acquires 14 PET sections with Z-axis translation. Each section was 8 mm thick and reconstructed with a transaxial resolution of 8 mm (full width at half maximum). Sections were acquired parallel to the O-M line, with eyes open in a dimly lit room and minimal auditory stimulation. A cylindrical tube filled with68 Ge was placed in the field of view for internal calibration of each section. Production of FDG and arterial sampling were performed as previously described. [15,16] All images for this study were obtained between 1991 and 1993.
Regional metabolic measurements.
Arterial samples to measure plasma18 F radioactivity were collected only during the imaging of normal subjects. To quantify data from subjects with TLE, ``raw'' counts from a single autoradiographic study (acquired for 20 minutes beginning 35 minutes postinjection) were calibrated against the68 Ge tube source visible in each section, and metabolic rates for glucose were determined using the method of Rhodes et al [17] as adapted by Lammertsma et al. [18] This procedure was also applied to autoradiographic studies of the 20 normal subjects, for comparison with measurements based on plasma sampling. For all 40 subjects, regional metabolic rates were normalized by each individual's whole brain rate (see below).
Region of interest (ROI) analysis was performed on 256 times 256 reconstructions using a SUN microcomputer (490 SPARC Server; Sun Microsystems, Mountain View, CA) with Scan V/P Software. [19] Anatomic ROIs were defined interactively on reconstructed PET sections with reference to a standard neuroanatomic atlas. [20] MRI data were not used in placing ROIs, beyond determining that the TLE subjects did not have gross structural abnormalities. [21] Each ROI initially consisted of a rectangular or elliptical ``frame'' fitted generously around the presumed cerebral structure of interest. The frame for a given ROI was placed first within the right hemisphere in control subjects or within the hemisphere contralateral to the focus in TLE patients, and then with no change in dimension reflected across the midline to a symmetric position in the opposite hemisphere. To select pixels representing gray rather than white matter metabolism and to reduce partial volume effects, we identified the upper 20% of local pixel values within each ROI frame, and derived a peak metabolic value for the ROI by averaging across these selected pixels. [21,22] This automated thresholding procedure determined the final shape of each ROI. Whenever anatomic regions straddled contiguous PET sections, full regional counts were calculated by weighting component ROI values from each section by the number of pixels selected by thresholding. Regional values were determined in this way for 13 standardized cortical and subcortical sites per hemisphere; these values were then normalized by dividing by the whole-brain (global) value for each patient, yielding a metabolic index, or MI. [23] The global metabolic value was calculated as the pixel-weighted mean of all supratentorial regional metabolic values. Mean values for MI in our normal sample ranged from 0.70 in medial temporal lobe to 1.58 in occipital lobe.
Arterial measurements from our normal subjects allowed us to compute plasma glucose input functions. Therefore, in this group, we could examine the relationship between MI at each ROI to analogous measurements for the same regions obtained with conventional quantification (ie, rCMRglc normalized by global metabolic rate [GMR] [24]). Regression analysis of rCMRglc/GMR versus MI showed excellent correlation between the two methods within each normal subject, and across the entire group of normal subjects (see Figure 1; compare Eidelberg et al [25]). Thus, with minimal invasiveness, we derived a measure of regional metabolism linearly related to normalized rCMRglc, at least over the range of this parameter found in our normal subjects.
Figure 1. Linear relationship between metabolic index (ratio of raw regional:raw global values, derived without plasma sampling) and the analogous measure using absolute values of glucose metabolism (rCMRglc:GMR, based on arterial plasma data; see Phelps et al [24]) in 20 normal subjects. The regression line has the following formula (plus minus SE): y equals 1.23(plus minus 0.03)x minus 0.24(plus minus 0.03), with r2 equals 0.99 (p less than 0.00001). Values shown are means from the left hemisphere; similar results were found in the right hemisphere. Plots from individual subjects had the same parameters and r2 as the group mean plot. The relative positioning of regional values along the regression line indicates the metabolic profile for the normal sample; this topography is comparable with either method of quantitation and matches that reported in other studies of normal subjects (see text). BG equals basal ganglia, CBEL equals cerebellum, CUNE equals cuneate cortex, LAT TEMP equals lateral temporal cortex, LF equals lateral frontal cortex, MB equals mesiobasal temporal lobe, MF equals medial frontal cortex, OCCIP equals occipital cortex, OPERC equals opercular frontal cortex, PAR equals parietal cortex, SUP TEMP equals superior temporal cortex, TEMP POLE equals temporal pole, THAL equals thalamus.
In keeping with the hypotheses of this study, we confined our analysis primarily to three temporal regions (MB temporal lobe, lateral temporal cortex, and temporal pole [anterior-inferior temporal lobe]), as well as to thalamus, medial frontal, and lateral frontal cortex. Figure 2 illustrates the temporal lobe ROIs. For comparrson with prior studies, we also examined the effect of using much larger temporal lobe ROIs. These ROIs [9,21] divided the temporal lobe into lateral and medial half-lobes, with each area extending the length of the lobe from base to pole. Counts for these large ROIs were otherwise obtained as above, including 20% thresholding and the weighted addition of components of each ROI across all relevant sections.
Figure 2. Contiguous transverse FDG-PET images from a normal subject illustrating the temporal lobe regions of interest (ROIs) used in this study. Sections are oriented parallel to the O-M line and proceed superiorly from image A; anterior is up, and the subject's right is on the right side of the image. Dashed rectangles within each section illustrate the initial ROI ``frame'' (here demonstrated only on the right side). The blackened interior areas show the final group of pixels selected by applying the 20% thresholding procedure within each ROI frame. Weighted contributions from each section were combined when ROIs straddled more than one section (see Methods). The temporal pole ROI is shown in the two most inferior sections shown here (A and B); selection for this ROI always began superior to the first plane containing temporal pole tissue (not displayed). Sections C and D illustrate the areas selected for mesiobasal (smaller, medial rectangles) and lateral temporal ROIs.
Statistical analysis.
Our selection criteria maximized the likelihood that patients' seizures originated within one temporal lobe. As the subgroups consisting of only left-sided or only right-sided focus cases were relatively small, we considered all 20 TLE cases as a single group, referencing regions of interest as either ``ipsilateral'' or ``contralateral'' to the site of seizure origin. For comparison, we used normal values, defined as the average MI from left and right homologous regions in the 20 control subjects. The significance of these comparisons was tested with unpaired two-tailed t tests, using the Bonferroni correction for multiple comparisons, and an alpha level of 0.05 (all p values reported prior to correction). Unless otherwise noted, significant findings remained so after correction.
Metabolic asymmetry was calculated for each region. For normal subjects, this index was defined as (L minus R)/[(L plus R)/2] times 100%, where L represents MI from the left and R from the right homologous brain regions. [14] For the TLE sample, patients with left and right foci were considered together and the asymmetry index calculated as (ipsi minus contra)/[(ipsi plus contra)/2] times 100%, where ``ipsi'' designates the regional MI ipsilateral to the seizure focus, and ``contra'' designates MI for the contralateral homologue.
Pearson product-moment correlations were used to examine metabolic relationships between pairs of bran regions. Correlation of metabolic activity between two regions may indicate a significant functional linkage, or joint modulation of the two regions by a common input. [23] In this study, correlation coefficients more than equals 0.44 were significant at p less than 0.05; those more than equals 0.52 were significant at p less than 0.01. Interregional correlations for TLE were compared with values from the normal sample using a t test on Fisher z-transformed r values. [26] Left- and right-focus TLE cases were again combined for this analysis, with regions designated as ipsilateral or contralateral to the seizure focus. In the case of correlations between bilaterally paired structures, the correlation in the TLE sample was compared with the single corresponding correlation in the normal sample. On the other hand, each pairing of nonhomologous regions in the TLE group could conceivably be compared with two pairs of regions (and thus two correlations) in the normal sample, generated by taking either the left or the right normal hemisphere as comparable to the ipsilateral hemisphere in the TLE sample. Therefore, each correlation between nonhomologous regions in the TLE group was compared with the average of the two pertinent normal correlations.
Results.
Global metabolism.
Mean global metabolism (plus minus SD) was the same for the normal (1.06 plus minus 0.38) and TLE (1.06 plus minus 0.30) groups (NS by two-tailed t test).
Regional metabolism within the temporal lobe.
Ipsilateral to the seizure focus, the most visually apparent abnormality in the TLE cases was decreased metabolism in the temporal pole. This impression was confirmed quantitatively (mean MI for TLE equals 0.75 plus minus 0.11 versus mean in normals equals 0.84 plus minus 0.05; p equals 0.001; see Figure 3). Although visually hypometabolism often appeared to extend beyond the temporal pole (into more proximal temporal lobe and/or adjacent frontal lobe), on average metabolism in the ipsilateral lateral temporal cortex did not differ significantly from normal (TLE equals 0.93 plus minus 0.09 versus control equals 0.95 plus minus 0.04; p more than 0.4). On the other hand, the ROI covering the MB region of the temporal lobe ipsilateral to the seizure focus showed a relatively elevated MI (TLE equals 0.89 plus minus 0.11 versus control equals 0.80 plus minus 0.06; p equals 0.005).
Figure 3. Metabolic index in temporal lobe subregions; comparison of means (plus minus SD) from 20 temporal lobe epilepsy (TLE) cases and 20 normal subjects (NORM). Homologous regions in the two hemispheres are sorted by their laterality relative to the seizure focus in the TLE cases, with ipsilateral regions shown in the left half of the chart and contralateral regions on the right. Values for the normal subjects are means of left and right hemispheres. p Values indicate significance of two-tailed t tests (normal versus TLE at each region). Temporal pole (POLE) is relatively hypometabolic ipsilateral to the focus. Mesiobasal (MB) metabolism is elevated both ipsilateral and contralateral to the focus, compared with normal. Contralateral lateral temporal cortex (LAT TEMP) also displays significant relative hypermetabolism.
Contralateral to the seizure focus, a different average profile emerged. The MI for the temporal pole on this side of the brain was not significantly different from normal (TLE mean equals 0.86 plus minus 0.12 versus control mean equals 0.84 plus minus 0.05; p more than 0.5). Lateral temporal cortex contralateral to the focus showed an average increase in mean MI relative to normal (TLE equals 1.04 plus minus 0.09 versus control equals 0.95 plus minus 0.04; p equals 0.0003). Metabolism in the contralateral MB also was significantly elevated relative to normal (TLE equals 0.92 plus minus 0.10 versus normal 0.80 plus minus 0.06; p equals 0.0001).
Findings with large temporal lobe ROIs.
We also examined interictal metabolism in the temporal lobe using less selective ROIs, which divided the entire lobe longitudinally into lateral and medial halves. MI for the large lateral ROI ipsilateral to the focus was not significantly different in TLE (0.84 plus minus 0.08) compared with normal (0.89 plus minus 0.05; p equals 0.03, NS following Bonferroni correction). The large medial area also failed to show significant metabolic differences between TLE (0.74 plus minus 0.06) and normal (0.76 plus minus 0.05; p equals 0.12). Contralateral to the focus, the large lateral area showed an apparent elevation of metabolism in TLE (0.94 plus minus 0.05) compared with normal (0.89 plus minus 0.05; p equals 0.0005), while MI for the contralateral large medial ROI was not different in TLE (0.80 plus minus 0.04) versus normal (0.76 plus minus 0.05; p equals 0.02, NS following Bonferroni correction).
Metabolism in thalamic and frontal regions.
The regional metabolic values for thalamus and medial frontal and lateral frontal cortex in the TLE sample were not significantly different from normal, either ipsilateral or contralateral to the seizure focus.
Analysis of regional asymmetry.
Temporal pole showed the greatest asymmetry index (minus 13.32 plus minus 10.43%), significantly different from the mean value for the same region in normal subjects (minus 2.08 plus minus 8.08%; p equals 0.0005; see Figure 4). Lateral temporal cortex had a mean asymmetry index of minus 10.76 plus minus 10.18%, significantly different from the normal value of minus 3.25 plus minus 5.76% (p equals 0.007). Mean asymmetry index for the MB region in TLE was minus 4.00 plus minus 10.23%, not significantly different from the normal value of minus 0.91 plus minus 6.07% (p equals 0.26).
Figure 4. Asymmetry index in temporal lobe subregions; comparison of means (plus minus SD) from 20 temporal lobe epilepsy (TLE) cases and 20 controls. Left- and right-focus TLE cases are considered as a single group, with the index calculated based on location relative to the focus (see Methods). Negative values for asymmetry index indicate ipsilateral less than contralateral metabolic index for a given region of interest. p Values show significance of two-tailed t tests. THAL equals thalamus; MED FRONT equals medial frontal cortex; LAT FRONT equals lateral frontal cortex; other abbreviations as in Figure 3
The negative values for asymmetry index in temporal pole and lateral temporal cortex both indicated lower metabolism ipsilateral to the focus relative to the contralateral regions. For the temporal pole, this reflected lower-than-normal metabolism ipsilateral to the focus in 19 of the 20 TLE subjects. By contrast, the negative asymmetry index for lateral temporal cortex stemmed from an MI greater-than-normal contralateral to the focus (18 of 20 TLE subjects; see Figure 3).
A parallel analysis with the larger ROIs showed that the region encompassing the entire lateral half of the temporal lobe had an asymmetry index in TLE (minus 11.36 plus minus 10.01%) that differed significantly from normal (minus 2.77 plus minus 7.09%; p equals 0.003). The asymmetry index for the large medial temporal area was also significantly different in TLE (minus 8.32 plus minus 8.08%) compared with normal (minus 0.15 plus minus 6.15%; p equals 0.0009). The negative asymmetry indices in TLE indicated relatively lower metabolism ipsilateral to the focus for both the large lateral and large medial temporal ROIs, although the regional means (see above) indicated that contralateral metabolic elevations had an important influence on these results.
There were no significant differences in asymmetry between TLE and control cases for thalamus or lateral frontal cortex. The asymmetry index for medial frontal cortex showed a trend toward differing from normal (TLE equals minus 1.12 plus minus 2.91%; normal equals 1.10 plus minus 2.58%; p equals 0.015), eliminated by Bonferroni correction.
Metabolic correlations among regions of interest in TLE.
As the MB region ipsilateral to the EEG focus was the likely site of epileptogenesis in our patient sample (see Discussion), we centered our correlation analysis on the metabolic relationships between MB and other regions of the ipsilateral and contralateral hemispheres. Left- and right-sided TLE cases were considered as a single group, with regions referenced as either ipsilateral or contralateral to the MB region presumed to contain the seizure focus. Correlations among these regions were evaluated against the comparable correlations in the normal sample, averaged from left and right hemispheres.
Temporal lobe.
Figure 5 summarizes correlations among the three temporal ROIs. We first considered relationships between homologous (bilaterally paired) regions. A strong correlation (r equals 0.68) was noted between ipsilateral and contralateral MB in TLE, similar to the corresponding value in normal subjects (r equals 0.74; intergroup t test NS, p equals 0.76). Between the paired lateral temporal regions the correlation in TLE was 0.46, identical to the normal value. The correlation between the paired temporal poles in TLE was 0.74, in contrast to the normal value of 0.30 (intergroup p equals 0.06). Correlations between non-homologous regions in TLE were also examined. First, the four such correlations among bilateral MB and bilateral lateral temporal regions were all statistically indistinguishable from the mean normal value (r equals 0.47); three of these correlations showed a tendency toward larger-than-normal positive coefficients. Second, of the four nonhomologous correlations among the MB regions and the temporal poles bilaterally, one differed significantly from the normal mean value (r equals 0.16), and another showed a similar trend (p equals 0.07). Finally, of the four correlations among the lateral temporal regions and the temporal poles, three were significantly different from the normal mean of 0.37. Considering together the eight correlations between the temporal poles and the other temporal regions (enclosed by bold lines in Figure 5), all five coefficients that were abnormal in TLE (one a trend) were negative. The remaining three correlations, although not significantly different from normal, were shifted in the same direction. Thus in TLE, all the MB and lateral temporal regions retained normal metabolic correlations with one another. As an ensemble, these four regions tended toward negative correlations with the temporal poles. By contrast, the temporal poles tended toward an abnormally positive correlation with one another.
Figure 5. Interregional correlation coefficients for temporal lobe metabolism in temporal lobe epilepsy (TLE). Each cell indicates the correlation of metabolic rates between pairs of temporal lobe regions, as indicated along the edges of the matrix. TLE cases with left- and right-sided foci are displayed together by referencing regions as either ipsilateral (IPSI) or contralateral (CONTRA) to the seizure focus, which in this sample lies in the ipsilateral mesiobasal region. Correlations between pairs of regions that involve the focus are shaded. Individual coefficients with absolute values more than equals 0.44 are significant at p equals 0.05 (shown in bold); those more than equals 0.52 are significant at p equals 0.01. Normal values (not shown, see Results) have the same significance cutoffs. The significance of comparisons (two-tailed t test) between TLE and normal correlations is indicated as follows: dagger p less than 0.10, *p less than 0.05, **p less than 0.01; no symbol, p more than equals 0.10. The heavy outline encloses correlations that describe the relationship between the mesiobasal and lateral temporal regions (as an ensemble) and the two temporal poles; these correlations all tend toward zero or significant negative values in TLE. Other comparisons are discussed in the text.
Extratemporal regions.
In the normal subjects, strong positive correlations were found between the bilaterally paired ROIs for thalamus (r equals 0.72), medial frontal cortex (r equals 0.81), and lateral frontal cortex (r equals 0.74). Correlations in the normal sample between thalamus and frontal cortical regions (average of left and right hemisphere values) were near zero (r equals minus 0.12 for medial and minus 0.19 for lateral cortex); similarly, none of the correlations between MB and thalamus, MB and medial frontal, or MB and lateral frontal cortex differed significantly from zero (within or across hemispheres). Considering the 20 TLE cases together, none of the correlations among these extratemporal regions differed significantly from the values seen in the normal sample.
Discussion.
Patient selection, localization of seizure focus.
Our inclusion criteria were designed to select patients with seizure origin in a single temporal lobe. The ultimate standards for such determination remain depth electrode studies and the efficacy of unilateral temporal lobe surgery, since not all patients with complex partial seizures necessarily have a temporal focus. [27] However, our approach using scalp-sphenoidal EEG (with consideration of both ictal and interictal findings) combined with semiquantitative screening of interictal PET studies, especially when supplemented by other criteria such as seizure semiology and MRI findings, can assign seizure origin to the correct temporal lobe with an accuracy that is not further enhanced by depth electrode studies. [28,29] Moreover, the clinical characteristics, surface EEG, and lack of MRI abnormalities (except for MTS) of our sample conform closely to the profile of TLE patients whose seizures have been confirmed by subsequent intracranial recording as not only originating from a single temporal lobe, but originating explicitly from the MB temporal region. [30] In all seven of our cases where intracranial EEG data were available, MB seizure origin was confirmed, supporting our view of this sample as representing patients with unilateral, MB TLE.
Quantitative method.
We normalized regional metabolic values by each subject's global metabolic value, yielding an MI. Although information about absolute metabolic rates (rCMRglc) is not obtained, normalization in this manner has the merit of reducing the impact of variability from several sources: that due to global metabolic differences among normal subjects; variation plausibly related to TLE (eg, due to heterogeneity of disease severity or spatial extent, effects of medication, [31] or phase of the ictal cycle [32]); or the possibility that metabolic parameters derived from normal brain may not necessarily apply to abnormal tissue. [33] Given the difficulty of estimating for each subject the impact of these factors on absolute measures of cerebral metabolism, normalization by global rate provides a reasonable basis for comparing regional metabolic parameters between TLE patients and controls. We used raw counts in the normalization process, without incorporating an arterial input function for FDG, eliminating the need for plasma sampling in our clinical group. [18] We demonstrated in our normal subjects that this normalized measure of metabolism was very highly correlated with the corresponding absolute CMR measure derived by conventional methods using arterial input functions [24] Figure 1.
One hazard of this type of normalization is that global variation between groups could theoretically generate spurious regional elevations or depressions of normalized metabolism, when in fact absolute rCMRglc is unchanged. However, the mean global metabolic rates in our two groups were statistically indistinguishable. In addition, the metabolic topography for our control sample matches that previously reported by others for both absolute [21,34] and normalized [23,35] glucose metabolism. Nevertheless, because we used the metabolic values from our control group as an external reference to describe the direction of metabolic alterations in TLE, the terms ``hypometabolic'' and ``hypermetabolic'' are used here strictly to indicate that the MI for a given region in the TLE sample differs significantly from the index for that region in the control sample.
Significance of ROIs, likely structures involved.
Our subjects were without localized cerebral lesions or atrophy beyond possible MTS in some patients. On this basis, [21] we referenced our ROIs to a standard atlas. Our ``temporal pole'' ROI encompasses primarily anterior and inferior temporal neocortex; variable (and in terms of relative volume within the ROI, minor) portions of amygdala and hippocampus may be included on the more superior sections contributing to this ROI. The ROI for MB temporal lobe is likely dominated by the amygdalahippocampal complex and parahippocampal cortex. Finally, the ROI for lateral temporal cortex probably samples primarily the middle temporal gyrus, although resolution of gyri is inadequate to confirm this. In all subjects, the same-sized ROI frames were applied to homologous areas in the two hemispheres, and final selection of pixels within the frame was made by an automated thresholding process that selected for gray matter and reduced partial volume effects.
Metabolic differentiation within the temporal lobe.
The normal temporal lobe overall has a lower metabolic rate than other cortical regions. [23,35,36] We confirmed this finding, but noted that MI is not uniform across the lobe. In normal subjects, the MB region had the lowest, temporal pole slightly higher, and lateral temporal cortex the highest MI within the temporal lobe Figure 3. In the TLE sample, metabolic differentiation among these subregions was also apparent. Ipsilateral to the seizure focus, the temporal pole was hypometabolic relative to normal, while contralaterally, temporal pole metabolism did not differ significantly from normal. These results for the temporal pole conform to conventional PET descriptions of the interictal signature of TLE (ie, hypometabolism in the vicinity of a temporal lobe focus). Unexpectedly, however, the MB region, which likely encompasses the epileptogenic focus in our patients, displayed a significant elevation of metabolism relative to normal, both ipsilateral and contralateral to the seizure focus. Additionally, for lateral temporal cortex, metabolism ipsilateral to the focus was not significantly different from normal, while contralaterally, MI in this region was relatively elevated. Henry et al [21] also noted higher-than-normal mean metabolism for most cortical ROIs contralateral to the focus, as did Leiderman et al [32] for a contralateral midtemporal ROI. Our results provide further evidence that a unilateral MB focus imposes regionally differentiated interictal abnormalities bilaterally within the temporal lobes. The use of selective temporal lobe ROIs and a normal comparison group allowed us to define more clearly the spatial dispersion and the variety (depressions and elevations relative to normal) of these interictal metabolic derangements.
Asymmetry indices considered in isolation may conceal such effects. In our TLE group, temporal pole and lateral temporal cortex both manifested significantly abnormal asymmetry Figure 4, in a similar direction (ipsilateral less than contralateral) and of a magnitude resembling previous reports. [11,33,37,38] However, we found that similar asymmetry trends for mesial and lateral regions derived from rather different local abnormalities. For the temporal pole, the large negative asymmetry value reflected lower-than-normal metabolism ipsilateral to the focus (see above). For lateral temporal cortex, in contrast, the abnormal asymmetry index arose from an increase in MI contralateral to the focus. The ``normal'' asymmetry index for the MB region in TLE in fact concealed bilateral increases in regional metabolism relative to normal.
The discrepancies between our findings and prior PET studies of TLE most likely reflect differences in design and sampling strategies. For example, we used metabolic values from a normal reference group to anchor our designations of metabolic change in TLE as either elevations or reductions; some studies that report solely widespread hypometabolism ipsilateral to the focus have lacked such a comparison. [33,39] As discussed above, assessment of metabolic rate made only relative to the contralateral hemisphere (eg, asymmetry measures) in TLE may be problematic, as may a failure to address global metabolic shifts. The selection of ROIs undoubtedly is another source of contrast between our results and some prior studies. Sperling et al [40] reported metabolic data from the temporal lobe as a whole. Henry et al [9,21] sampled two extensive ROIs that together completely covered the lateral and medial aspects of the temporal lobe. Localized metabolic elevations in a relatively small region such as MB are likely to go undetected under such circumstances. To assess this possibility, we analyzed our images using similar large temporal ROIs. The use of large medial and large lateral ROIs did not produce evidence for interictal elevation of metabolism above normal, either ipsilateral or contralateral to the focus. Indeed, compared with normal, these large ROIs showed no mean regional metabolic changes ipsilateral to the focus, consistent with cancellation among subregions of hyper- and hypometabolism. With such large and undifferentiated ROIs, we found that asymmetry calculations provided the main indication of decreased metabolism ipsilateral to the focus (similar to the findings of Henry et al [9,21]), yet such evidence is confounded by the possibility of contralateral metabolic elevation.
When more selective ROIs have been used, PET studies of interictal metabolism have invariably revealed regional topography within the temporal lobe. The study of Hajek et al [38] most resembles ours in using similar temporal lobe ROIs and normalization by global metabolism. These authors report interictal hypometabolism within the lateral temporal cortex independent of whether the seizure focus was localized to the lateral or medial temporal lobe; the MB region appeared hypometabolic only if the seizure focus were electrographically localized to the MB region and associated with mesial gliosis by pathology. In contrast to our study, Hajek et al [38] found no evidence for elevated metabolism in any ipsilateral temporal ROIs; values for contralateral regions were not reported. However, these authors report atypically high metabolism for temporal ROIs in their control group (normalized values ranging from 105 to 121% of global brain value), while previously reported normal values for these regions, as in the present study, range from 70 to 95% of global. [23,35,36] Thus some systematic difference in normalization procedures may account for our differing conclusions about whether particular regions should be classified as hypo- or hypermetabolic.
In other regards, our study is consistent with prior PET reports on the interictal state in TLE. Sackellares et al [11] and Abou-Khalil et al [41] measured activity in single rows of pixels across both lateral and medial temporal areas in a sample similar to ours. These authors concluded that interictally the lateral temporal cortex was more likely than the medial temporal region to show rCMRglc depression ipsilateral to the focus, whether the focus resided in the medial or lateral temporal lobe. Metabolism in the temporal pole was not reported. Sadzot et al [12] made similar observations, including measurements of the temporal pole, and also reported a differential likelihood among temporal regions to manifest metabolic asymmetry in TLE (temporal pole more than lateral temporal more than MB). Sackellares et al [11] and Sadzot et al [12] did not find MB hypermetabolism; however, in both studies it appears that patients and normal subjects differed in global metabolism. Regional values were not accordingly normalized, hampering cross-study comparisons. However, apart from unsettled issues about global metabolism, these recent studies and our data agree in finding MB relatively resistant to manifesting interictal metabolic depression, while the lateral temporal region and especially the temporal pole are more susceptible.
Interictal metabolism in the MB temporal lobe.
Our evidence that interictal metabolism in the MB region may be elevated above normal is unusual among PET studies; only isolated similar findings have appeared. [42,43] More commonly, the temporal focus has been described in PET studies as a region of interictal FDG hypometabolism. However, a paradox arises from this view of the TLE focus, since electrophysiologic evidence indicates that the focus is a site of increased interictal neural activity. A variety of investigations suggest that within the TLE focus, hippocampal neurons are prone to fire hypersynchronously, but are held in check by enhanced active inhibitory processes; intermittent brief loss and restabilization of this balance, represented electrographically by the interictal spike-and-wave complex, may constitute an ``ictal fragment.'' [4] It has been suggested that these local electrophysiologic processes, which one would expect to cause a net increase in glucose metabolism, [44] do not have a metabolic manifestation in interictal FDG-PET images because only relatively few neurons per unit volume are so engaged. [4] Such an explanation is difficult to evaluate critically, since volume averaging is an inherent limitation in all FDG-PET studies, a problem exacerbated in many studies by the choice of relatively large ROIs. On the other hand, evidence indicating interictal elevation of metabolism in the MB region has the merit of potentially aligning FDG-PET findings with the substantial evidence for increased interictal neurophysiologic activity at the TLE focus.
Other findings indicative of interictal elevated metabolism at the seizure focus have appeared in single-photon emission computed tomography studies of regional cerebral blood flow (rCBF) in TLE. Rowe et al [45] and Newton et al [46] noted that relative hyperperfusion seen throughout the ipsilateral temporal lobe during a seizure rapidly gave way to widespread temporal hypoperfusion except in the ipsilateral MB region, where rCBF declined less dramatically from ictal levels, and relative hyperperfusion persisted into the postictal period. As flow was quantified primarily in terms of asymmetry in these studies (without accounting for possible flow shifts in the contralateral reference region) the prolongation of MB flow elevation even beyond the postictal period cannot be ruled out. If so, there is evidence for interictal hyperperfusion as well as hypermetabolism at the MB focus.
Studies of experimental epilepsy also suggest interictal hypermetabolism in the MB region. Van Landingham and Lothman [47] studied the interictal state in rats with 2-deoxyglucose autoradiography after recovery from transient (several hours) limbic status epilepticus induced by unilateral hippocampal stimulation. One week after status epilepticus had abated, behavioral and electrographic seizures were absent and the EEG was characterized by isolated, bilaterally synchronous interictal spikes. At this time, significant bilateral and symmetric hypermetabolism persisted in hippocampal and amygdalar regions. These effects largely resolved by 30 days following the end of status epilepticus, although even at this point the central nucleus of the amygdala remained hypermetabolic bilaterally. White and Price, [48] in a different experimental model, also found evidence for bilateral hypermetabolism in the amygdala/hippocampal complex during relatively quiescent phases of the ictal cycle. The degree of residual hypermetabolism in this region in experimental epilepsy may be related to the incidence of interictal spikes. [47,49]
Accepting the caveat that postictal and interictal states are somewhat arbitrarily defined in TLE, [32] these convergent observations indicate that not only during seizures, but also between ictal events, the MB region in TLE remains in a relatively hypermetabolic state, perhaps reflecting a dynamic balance between opposing, pathologic synaptic processes. Thus, in spite of the paucity of such reports in the past, there have been reasonable grounds to expect FDG-PET to reveal increased interictal metabolism at the TLE focus.
Bilateral metabolic equilibration between MB regions.
We found in TLE that metabolic activity in the MB region ipsilateral to the seizure focus was not significantly different from that in the contralateral homologous area, even while both appeared elevated relative to normal. As a result, the high degree of metabolic correlation between MB regions seen in the normal controls was preserved in the TLE sample. High metabolic correlations between homologous brain regions have generally been attributed to commissural connections. In the case of the hippocampal/amygdalar complex (sampled by our MB ROI), the presence of a functional commissural connection in the human has been debated, although evidence now favors its existence. The reports of Spencer et al [50] and Gloor et al [51] together suggest that the dorsal hippocampal commissure provides a ready avenue for ictal activity originating in one hippocampal formation to propagate to the contralateral homologous area. Such commissural transmission has been noted in experimental epilepsy, where epileptiform discharges and a variety of other excitatory and inhibitory influences pass between the two hippocampal regions, creating electrophysiologic synchrony during some phases of the ictal cycle. [52,53] MB glucose utilization similarly tends toward bilateral symmetry in experimental epilepsy (see above). Such findings indicate that in both the normal and epileptic brain, the two hippocampal regions equilibrate electrical and metabolic activity with one another, consistent with our observation of high metabolic correlations between the MB regions in both normal subjects and patients with TLE.
Studies of rCBF in the peri-ictal period also suggest functional equilibration or coupling between MB regions. Following a seizure, regional perfusion asymmetries resolve especially rapidly between the two MB regions, resetting the asymmetry index to a low value indistinguishable from normal. [46] This restoration of symmetry for MB occurs via reductions of ipsilateral MB perfusion in some patients and increases in perfusion in others, [46] perhaps reflective of a coupled system re-equilibrating around a metabolic set-point after ictal perturbation. Equilibration between hippocampal regions may also help explain the relationship between metabolic and neuropsychological findings in TLE. For example, Rausch et al [54] found that interictal metabolism and asymmetry calculations for the left hippocampal region (in left MB TLE) correlated surprisingly poorly with verbal cognitive deficits thought to reflect hippocampal cell loss, while lateral temporal metabolism did correlate well with the cognitive deficit. Such poor correlation might result if metabolism in the epileptogenic MB region were coupled to that of the opposite normal area to a degree that stabilized or masked the metabolic depression otherwise expected as a consequence of local cell loss (see also Henry et al [55]). Stabilization of this type could also underlie the relative resistance of the MB region to manifestation of interictal metabolic asymmetry. [11,12]
Other remote interictal effects in unilateral TLE.
In our correlation analysis, we observed a variety of remote interictal effects related to a unilateral focus, besides the apparent equilibration between the two MB regions. Correlations among the bilateral MB and lateral temporal areas were not significantly perturbed in the TLE sample, with a tendency toward greater positive magnitude. By contrast, the set of correlations relating metabolism in the MB/lateral temporal ensemble to metabolism in the temporal poles decreased in magnitude or took on negative values in the TLE sample, with half of these correlational differences achieving statistical significance. Concurrently there was a trend toward an abnormally positive correlation between the two temporal poles in TLE. No significant difference from normal was seen in correlations between thalamus and frontal cortex areas in the TLE sample. Sample size limitations precluded a separate examination of correlations in left- and right-sided TLE subgroups, and we used correlations averaged from both normal hemispheres for comparison; therefore, this correlation analysis should be considered preliminary.
The selective manner in which correlations appear to be perturbed in TLE may demarcate metabolically linked groups of regions. The significance of such ensembles has not been defined. However, similar observations have been made in various neuropsychiatric disorders [26,56] and may reveal functional systems involved in a disorder's clinical manifestation. Our evidence that bilateral MB and lateral temporal regions form a stable (or possibly more positively correlated) metabolic cluster in TLE is consistent with recent neuropsychological findings. For example, lateral temporal areas appear to interact with MB structures in the same hemisphere in short-term memory function. [57] Evidence for bilateral interplay among such areas comes from studies indicating temporal lobe dysfunction contralateral to a unilateral TLE focus; ``neural noise'' passed along pathways linking contralateral temporal areas to the focus is hypothesized to underlie some cognitive deficits. [6,7,58] From a different line of inquiry, Bancaud et al [59] postulated a bilateral network among mesial temporal and lateral temporal areas similar to the ensemble we have described, and provided evidence that some perceptual and experiential phenomena of TLE arise from abnormal function within this network. We also noted that this ensemble developed an abnormal (negatively correlated) metabolic relationship with the temporal poles in TLE, and that the poles themselves tended toward an abnormal positive relationship with one another. The poles are paralimbic regions involved in drive and affect regulation, and their metabolic dysregulation could plausibly contribute to interictal behavioral abnormalities. [60]
In conclusion, our study of interictal metabolic alterations in TLE employed (1) a normal control group as a stable reference for metabolic comparisons, (2) a TLE sample with unilateral seizure origin, allowing the discrimination of ipsilateral from contralateral effects, (3) a choice of ROIs that took account of evidence for functional differentiation within the temporal lobe, and (4) normalization of regional metabolism by each subject's global value, to control several potential sources of variability. We found both elevations and depressions of metabolism in the interictal state compared with normal, as well as altered metabolic correlations among regions. These abnormalities were widely distributed, extended to contralateral temporal regions, and were consistent with a variety of other neurobiologic findings in TLE.
Acknowledgments
We thank the members of the NSUH cyclotron staff for technical support and Drs. T. Pedley and S. Resor (Department of Neurology, Columbia University) for referral of clinical subjects. Dr. H.A. Sackeim provided helpful comments on an early version of this report.
- Copyright 1995 by Advanstar Communications Inc.
REFERENCES
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Cutaneous α-Synuclein Signatures in Patients With Multiple System Atrophy and Parkinson Disease
Dr. Rizwan S. Akhtar and Dr. Sarah Brooker
► Watch
Alert Me
Recommended articles
Seizure focus affects regional language networks assessed by fMRI
Prefrontal hypofunction in patients with intractable mesial temporal lobe epilepsy
11C-flumazenil PET in patients with refractory temporal lobe epilepsy and normal MRI
Limbic reductions of 5-HT1A receptor binding in human temporal lobe epilepsy