FDG-PET and volumetric MRI in the evaluation of patients with partial epilepsy

Positron emission tomography with Fluorine-18-fluorodeoxyglucose (FDG-PET) and MRI are the most useful of the many noninvasive radiologic techniques that identify the epileptogenic zone in patients with complex partial seizures (CPS). FDG-PET identifies a metabolic abnormality ipsilateral to the seizure focus in 60 to 85% of patients with temporal lobe onset ^[1-6]. Focal abnormalities on T_2-weighted MRI are present in a high proportion of subjects with CPS who are candidates for surgery, particularly when mesial temporal sclerosis (MTS) is present and a field strength of 1.5 tesla is used ^[7-12]. Several studies suggest interictal FDG-PET may be more sensitive in identifying epileptogenic zones than T₂ MRI ^[8,9,13-16]. MRI volumetry, however, may identify hippocampal formation (HF) atrophy associated with MTS with greater sensitivity and specificity than T₂ MRI ^[17-19].

The pathophysiologic basis for hypometabolism in temporal lobe epilepsy is uncertain. Many studies report relatively greater lateral than mesial reduction even though pathologic findings are more prominent mesially ^[4,5,20]. We compared the utility of volumetric MRI and FDG-PET in seizure focus localization in patients with complex partial epilepsy of temporal lobe origin and examined the relationship between HF volume and temporal lobe metabolism.

Methods. Patients. Eighteen consecutive patients from the NINDS Clinical Epilepsy Section Clinic who had a temporal lobe ictal focus demonstrated by scalp-sphenoidal EEG-telemetry (four patients had subdural EEG monitoring) were enrolled in this study after informed consent was obtained. Eleven patients were male and seven were female (table). The mean age at the time of study was 29 years (range, 15 to 41 years), and mean age at seizure onset was 9 years (1 to 18 years). All patients had CPS, 12 had simple partial seizures, and 16 had secondarily generalized seizures. Eight patients had a right temporal focus and 10 a left temporal focus. Seven patients were treated with a single antiepileptic drug (AED) at the time of PET, 12 were on two or more AEDs, and one was on either felbamate or placebo. All patients had 1.5-T MRI and FDG-PET.

PET. Interictal PET studies were performed on the Scanditronix 2048-15B scanner with a full-width half-maximum axial and in-plane resolution of 5.5 mm, scanning 15 simultaneous slices 6.6 mm apart. Scans were performed following a 4-hour fast in the awake resting state with eyes patched and ears plugged. Heads were immobilized by a thermoplastic mask and scans oriented along the canthomeatal plane. A measured attenuation scan was performed using a Germanium-68/Gallium-68 rotating pin source. Five mCi FDG was injected. Following a 30-minute uptake period, data were acquired over two 15-minute intervals. Continuous EEG was performed and patients observed by an investigator (W.D.G.) to exclude ictal activity. Fifteen patients had arterial sampling and absolute local cerebral metabolic rate of glucose (LCMRGlc; mg glc/min/100 g) measured ^[21]. Regional values normalized to whole brain counts were used in the three patients who did not have arterial sampling.

PET was analyzed using a standard template of 176 regions, 6-mm square, grouped into 13 paired anatomic areas placed on seven slices. Particular interest was directed to the temporal regions formed of 74 regions of interest arranged into four paired areas. An asymmetry index (AI) was calculated for each paired area: (right - left)/((right + left)/2). Normal volunteer data using this template and scanner demonstrated mean 2 SDs in AI to be 0.128; hence, an asymmetry greater than 12.8% was considered significant and indicating regional hypometabolism ^[22].

MRI. MRI was performed on the General Electric Medical Systems (Milwaukee, WI) Signa 1.5-T scanner in contiguous 2-mm-thick coronal slices with the following acquisition parameters: TR, 24 msec; TE, 5 msec; flip angle, 45 degrees; field of view, 24 x 24 cm; and matrix, 256 x 256. The voxel size was 0.9375 x 0.9375 x 2 cubic mm. Scans were obtained perpendicular to the orbitomeatal plane. The HF was outlined on each coronal image from the pes hippocampi anteriorly to the fasciolar gyrus at the level of the pulvinar. Voxels from each region were summed from each successive picture to calculate hippocampal volume as described previously ^[23].

HF volume data were compared to previously established normative data based on 29 normal volunteers ^[23]. HF atrophy was considered present if absolute HF volume was beyond 2 SDs for normal subjects and deemed strongly suggestive of relative atrophy if the HF ratio, L/R, was beyond 2 SDs ^[24].

Results. Nine patients had absolute HF atrophy, all ipsilateral to the ictal focus Table 1. Eight of these had abnormal HF ratios. Two other patients had abnormal ratios, one of which was falsely lateralizing compared with the ictal EEG. Both of these patients had properly lateralizing FDG-PET.

PET demonstrated focal temporal hypometabolism in 16 of 18 patients (89%), all ipsilateral to the ictal EEG focus. Most subjects had regional hypometabolism in the inferior lateral (11 patients) and mesial (nine patients) structures (table). Five subjects had regional hypometabolism extending beyond the temporal regions to adjacent frontal (four patients) or occipital (one patient) cortical areas. In our previous studies, using an older scanner and template, 2 SDs for normative AI was 0.150 ^[4,5]. All patients in this study with focal PET had at least one temporal region with an AI greater than 0.150. Using an AI of 0.128 indicated more extensive temporal hypometabolism in two patients and showed adjacent frontal hypometabolism in one other patient.

All patients with absolute HF atrophy had ipsilateral temporal hypometabolism. However, six patients with focal PET had normal HF volumes, two of whom had structural lesions (one a region of temporal encephalomalacia, the other a temporal tip vascular malformation). Eleven had abnormalities seen on T₂ MRI (in nine there was increased T₂ signal in HF and in two the structural abnormalities described above), which in two cases were not seen on either volumetry or PET. Four patients had abnormal volumetry (one falsely lateralizing) but normal T₂ MRI.

Nine patients had temporal lobectomy; eight are seizure free (mean follow-up, 28 months; range, 11 to 38 months). Pathology of resected tissue in these eight patients showed varying degrees of gliosis, neuron loss, and rare subcortical neurons; a pathologic diagnosis could not be made in two subjects. Pathologic studies were limited, since only portions of resected brain were available for review and the HFs were not removed en bloc for microscopic study. One patient (subject 8) has persistent seizures, although moderately decreased in frequency: he had left hemispheric onset on invasive monitoring thought to arise from the temporal region, left superior temporal and frontal hypometabolism, and an abnormal HF ratio suggesting relative contralateral atrophy. However, both hippocampi were more than 2 SDs larger than normal control values, and pathologic study of resected temporal tissue was normal. Either the temporal focus was not completely resected or the residual focus is frontal.

There were statistically significant correlations between HF volume ratio and AI for CMRGlc in the inferior lateral and inferior mesial temporal areas (r = 0.67, p < 0.01; r = 0.72, p < 0.01, excluding subject 8) Figure 1 and Figure 2. Also, there were significant correlations between absolute HF volume and regional CMRGlc in inferior lateral and inferior mesial, but not middle and superior, temporal areas (r = 0.412, p < 0.034; r = 0.539, p < 0.006; r = 0.308, p < 0.10; r = 0.119, p < 0.33).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Relationship between hippocampal formation volume ratio (L/R) and inferior lateral temporal (ILT) cortex CMRGlc percent asymmetry ((AI) x 100)

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Relationship between hippocampal formation volume ratio (L/R) and inferior mesial temporal (IMT) cortex CMRGlc percent asymmetry ((AI) x 100)

Discussion. We found that PET is superior to either HF volume measurements or T_2-weighted MRI in temporal lobe CPS focus localization. MRI volumetry and T₂ MRI were of equivalent use in detecting focal abnormalities. PET provided additional information to MRI in 22% of patients. Although PET proved more sensitive than volumetric MRI, when atrophy was present, PET did not provide additional information. Although PET hypometabolism and hippocampal atrophy predict successful temporal lobectomy, discordant studies or large hippocampal volumes may suggest higher risk for a poor surgical outcome and merit greater caution in surgical evaluation ^[5,17]. Regional temporal hypometabolism, in part, appears to reflect HF atrophy.

We used an AI, rather than absolute measures, to determine PET regional hypometabolism. Due to interindividual variability, absolute values may be a less reliable measure of relative hypometabolism. In addition, the effect of AEDs, which reduce global CMRGlc but not the degree of asymmetry, can be ignored ^[22,25-28]. Even when bilateral temporal hypometabolism was present in nine patients, significant asymmetry could be demonstrated in every subject; these nine patients also had global reductions in CMRGlc. Moreover, FDG AI predicts outcome after temporal lobectomy ^[5,29]. Both absolute HF volume measurements and L/R HF volume ratios can identify epileptic foci ^{[17-19,24,30]}. MRI volume measurements are probably not affected by AEDs.

The majority of patients who undergo temporal lobectomy have varying degrees of gliosis and neuronal loss ^[14,31-37]. HF atrophy may be proportional to gliosis and cell loss ^[18,19]. Although previous FDG-PET studies have not noted any correlation between extent or degree of metabolic abnormality and degree of pathologically identified gliosis, ^[14,38] our results suggest an association between temporal cortical metabolism and HF volume. In some patients the more prominent hypometabolism in lateral than mesial cortex may reflect decreased function in projection fields of gliotic hippocampus or technical factors such as partial volume averaging ^[4,5,20,39]. There may be other causes for temporal lobe hypometabolism, as when a structural lesion, such as a vascular malformation or tumor, or a neocortical focus is present. In these latter circumstances, normal HF volumes are found.

Six of the eight subjects whose LCMRGlc and hippocampal volumes both fell below 2 SDs had a history of febrile seizures. Several previous authors have found an association between prolonged febrile seizures and hippocampal atrophy ^[31,40,41]. A significant early insult may lead to sclerosis, atrophy, and an increased risk for seizures. Early recurrent seizures may also cause progressive neuronal loss ^[37,42]. Comparison of PET and MRI results may be facilitated by registration using digital image processing. Although this approach may improve analysis of the relationship between volume and metabolism analysis, partial volume effects will still occur in small structures such as the hippocampus.

Both FDG-PET and MRI are valuable tools for the identification of the epileptogenic zones in patients with CPS of temporal lobe origin. We suggest that MRI volumetry and T₂ MRI be performed first. MRI is more readily available than PET and, when abnormal, may demonstrate a cause for the seizures as well as assist in planning surgical resection. If these studies are not revealing, then FDG-PET should be performed. To reduce the cost as well as the morbidity associated with epilepsy surgery evaluation, it is important to use both techniques selectively and to avoid redundant studies.

Acknowledgments

We thank Drs. D. Katz, O. Devinsky, and E. Bromfield for providing pathology results.