Transaxial positron emission tomography with fludeoxyglucose F 18 image (A) and coronal T1-weighted magnetic resonance image (B) of patient 8, with left temporal hypometabolism (arrows) and normal hippocampal quantitative magnetic resonance imaging results.
Lamusuo S, Jutila L, Ylinen A, Kälviäinen R, Mervaala E, Haaparanta M, Jääskeläinen S, Partanen K, Vapalahti M, Rinne J. [18F]FDG-PET Reveals Temporal Hypometabolism in Patients With Temporal Lobe Epilepsy Even When Quantitative MRI and Histopathological Analysis Show Only Mild Hippocampal Damage. Arch Neurol. 2001;58(6):933-939. doi:10.1001/archneur.58.6.933
The relationship between reduced glucose metabolism in positron emission tomography with fludeoxyglucose F 18 ([18F]FDG-PET) and hippocampal damage (HD) in patients with temporal lobe epilepsy is still unclear.
To determine whether the presence and severity of HD verified by quantitative magnetic resonance imaging (QMRI) and histopathological analysis affect the degree of hypometabolism.
Patients and Methods
Sixteen patients with drug-resistant temporal lobe epilepsy underwent [18F]FDG-PET and QMRI (hippocampal volumetry and T2 relaxometry) before surgery. Histopathological analysis of the hippocampus included measurements of neuronal loss, proliferation of glial cells, and mossy fiber sprouting. The asymmetry in glucose metabolism described the degree of hypometabolism.
Temporal hypometabolism was not related to severity of HD as measured by QMRI or histopathological analysis. The degree of hypometabolism did not differ in patients with mild, moderate, or severe HD. In addition, [18F]FDG-PET revealed significant temporal hypometabolism even though hippocampal QMRI findings were normal or showed only mild HD. Thus, glucose consumption was reduced over and above the histopathological changes.
[18F]FDG-PET is sensitive for localizing the epileptogenic region in patients with temporal lobe epilepsy. However, it is insensitive to reflect the severity of HD.
POSITRON emission tomography (PET) with fludeoxyglucose F 18 ([18F]FDG) has shown the region of reduced interictal glucose metabolism in the epileptogenic temporal lobe,1- 4 and high-resolution quantitative magnetic resonance imaging (QMRI)5- 7 has reduced hippocampal volume (HCV) and prolonged T2-weighted signal (HCT2) in the epileptic hippocampus in patients with intractable temporal lobe epilepsy (TLE). Histopathologically, hippocampal damage (HD) found in 60% to 70% of patients with TLE contains neuronal loss in the hilus and CA1 subfield, gliosis, and synaptic reorganization.8- 10
Reduced HCV in QMRI has been shown to be related to histopathologically verified neuronal loss, whereas prolonged HCT2 is assumed to reflect the increased proliferation of glial cells in the hippocampus.11,12 However, the pathophysiological basis of the reduced glucose metabolism is still unclear. At least the severity of HCV loss in QMRI has been assumed to be unrelated to the temporal hypometabolism.13 Furthermore, a few studies have found no correlation between the histopathologically verified HD and hypometabolism,14- 17 although the results have been contradictory.13- 19 Therefore, we wanted to find out whether the relationship exists and whether [18F]FDG-PET could lateralize the epileptogenic region in patients with TLE even when hippocampal QMRI shows normal findings or only mild HD.
Sixteen patients (5 women and 11 men), all of whom underwent surgery because of intractable TLE, were preoperatively investigated with [18F]FDG-PET in the Turku PET Centre, Turku, Finland. Their mean (SD) age was 32.8 (10.9) years (median, 31.0 years); mean (SD) age at the onset of epilepsy was 15.9 (11.8) years (range, 0.8-43.0 years; median, 14.0 years), and mean (SD) duration of TLE was 17.0 (14.0) years (median, 13.0 years). Clinical data are presented in Table 1.
The presurgical evaluation was performed by the Kuopio University Hospital, Kuopio, Finland, epilepsy surgery team according to the method described previously.20,21 Fifteen of 16 operations were preoperatively evaluated as curative (unilateral TLE with a good outcome), and 1 was a palliative procedure (only a reduction in seizure frequency expected). The palliative operation was performed after careful patient counseling of the expected outcome and in agreement with the patient. The histopathological analyses of resected tissues were performed at the A.I. Virtanen Institute in Kuopio.
Written informed consent was obtained from all participants. The ethics committees of Turku University Central Hospital and Kuopio University Hospital approved the clinical and imaging parts of the protocol. The National Board of Medical Legal Affairs approved the use of human tissue for histopathological analysis.
The PET study was carried out with an 8-ring whole-body PET scanner (ECAT 931/08-I2; Siemens/CTI, Knoxville, Tenn), and the [18F]FDG was prepared according to methods described previously.20- 23 We used the values for the same healthy volunteers as in earlier studies.20,21 Elliptical regions of interest, with an average size of 0.5 × 2.0 cm, were placed individually in the lateral temporal neocortical gyri (superior, middle, and inferior temporal) and in the medial temporal lobe (covering the hippocampus and part of the amygdala) in addition to the regions shown in Table 2. The locations of regions of interest in different brain areas were defined using anatomical atlases and patients' MRI scans. Regional cerebral glucose metabolic rates (rCMRgluc in micromoles per milliliter per minute) were calculated according to the method of Patlak and Blasberg.24 The left-right asymmetry (asymmetry index [AI]) of rCMRgluc was measured for all pairs of homologous regions of the hemispheres using the widely used formula: AI = (L − R) × 100/([L + R]/2)%. The AI was used to indicate the region of hypometabolism and also to describe the degree of hypometabolism. Regions of interest with the greatest AIs showing the hypometabolic area ipsilaterally (ipsilateral referring to the side of the surgical lobe) in each of the 3 lateral temporal neocortical gyri and in the medial temporal lobe were selected when calculating comparisons between the results of hippocampal QMRI and histopathological analyses. Also the left-right ratio in glucose utilization in each temporal area was calculated, as the HCV ratio is usually determined in the same way. Data from controls were analyzed in the same manner by the investigator (S.L.) analyzing the patient data. When calculating the mean AIs, the absolute AI ( = |AI|) was used.
Results of histopathological analysis of resected hippocampus samples were available for 15 of 16 patients. Hippocampal tissue was histologically analyzed in patients and controls in the same manner as described previously.25- 28 In each case, the sum score of the density of neurons and the severity of astrogliosis can vary between 0 and 15 (the density of neurons: sum score 1-5 indicates mild damage; 6-10, moderate damage; and 11-15, severe damage; and the severity of astrogliosis: sum score 1-5, mild gliosis; 6-10, moderate gliosis; and 11-15, severe gliosis). The density of mossy fiber sprouting was scored from 0 to 5.29 If the score was lower than 2, the severity of mossy fiber sprouting was considered mild; 2 to 3, moderate, and 4 to 5, severe.
Patients were divided into 3 groups based on the histopathological findings, counting the sum scores of neuronal loss, astrogliosis, and mossy fiber sprouting together. In patients with mild HD, the maximum total sum score was 11 (5 + 5 + 1 = 11); in patients with moderate HD, the minimum total sum score was 14 (6 + 6 + 2) and the maximum was 23 (10 + 10 + 3); and in patients with severe HD, the minimum total sum score was 26 (11 + 11 + 4).
Preoperative HCV and HCT2 were available for all patients. The method used to measure HCV and HCT2 has been described previously.12,30,31 Briefly, the individuals were scanned with a 1.5-T Magnetom SP63 (Siemens, Erlangen, Germany) using a standard head coil and a tilted coronal 3-dimensional gradient-echo sequence (MP-RAGE: repetition time, 10 milliseconds; echo time, 4 milliseconds; inversion time, 250 milliseconds; flip angle, 12°; field of view, 250 mm; matrix, 256 × 192; and 1 acquisition). This gave 128 T1-weighted slices (thickness, 1.5-1.8 mm) oriented at right angles to the long axis of the hippocampus.
To evaluate the significance of the asymmetry, the confidence interval was calculated.20 Similar to earlier studies,3,20,32,33 [18F]FDG-PET results with metabolic asymmetries exceeding 15% in at least 2 adjacent image planes in any region of the brain were considered significant. In hippocampal QMRI, the reference range for all variables was defined as 2 SDs above and below the control mean (for normalized HCV [nHCV]: <2500 mm3 in the left side and <2717 mm3 in the right side; for HCT2: >109 milliseconds34). Statistical analysis was performed using Origin software (version 5.0; Microcal Software Inc, Northampton, Mass). The Pearson correlation coefficient (r) and the Mann-Whitney U test were used to determine the relationships between the results of [18F]FDG-PET, hippocampal QMRI, and histopathological analyses. The t test was used to determine when results of [18F]FDG-PET and hippocampal QMRI differed statistically from the values of the controls. The Bonferroni correction for multiple comparisons was used when calculating the comparisons shown in Table 3. The level of significance was set at P<.05.
Electroencephalography with sphenoidal electrodes lateralized seizure onsets unilaterally in the lateral temporal neocortical areas, the medial temporal lobe, or both in 5 patients and with foramen ovale electrodes in the medial temporal lobe in 1 patient (patient 5). In electroencephalography with subdural electrodes, epileptiform activity originated from the lateral temporal neocortical areas and the medial temporal lobe in 5 patients, only from the medial temporal lobe in patient 2, and bitemporally in patient 4.
Mild HD (range of the total sum scores of the histopathological analysis, 7-11) was found in 3 patients (patients 2, 10, and 13), moderate HD (range, 14-21) in 4 patients (patients 1, 8, 11, and 15), and severe HD (range, 27-34) in 8 patients. The differences in the mean sum scores of the histopathological analyses between these groups were significant except the difference in the mean sum score of the severity of mossy fiber sprouting between patients with mild or moderate HD (Table 3).
Table 2 summarizes separately the mean rCMRgluc and AIs in all determined brain regions in patients and controls. In controls, there were no significant differences in the mean left and right rCMRgluc values in any temporal areas or in other brain areas. In patients, the mean rCMRgluc was reduced by 4% to 33% from the control mean in all brain regions. However, the mean AIs in brain regions other than the temporal areas did not differ from those of the controls (Table 2).
[18F]FDG-PET lateralized the epileptogenic zone in agreement with the surgical lobe in 88% (14/16) of the patients. Patient 10 had visually reduced uptake of [18F]FDG in the ipsilateral temporal area but, quantitatively, the AI (12%) did not reach the cutoff threshold of 15%. In patient 16, [18F]FDG-PET revealed a wide hypometabolic area covering all the right temporal areas and the right parietal lobe (Table 4).
In 3 patient groups with differing severity of HD, the mean AIs were significantly different from the control mean in all temporal areas except in the inferior temporal gyrus in patients with mild HD (P = .09) and in the superior temporal gyrus in patients with moderate HD (P = .12) (Table 3).
In 8 of 16 patients, the ipsilateral, and in 2 of the 8 patients also the contralateral, nHCV was reduced significantly. Patient 15 had contralaterally reduced nHCV. In 10 (62%) of 16 patients, the ipsilateral, and in 2 of these 10 the contralateral, HCT2 was also prolonged significantly (Table 4).
In 7 of 16 patients, the ipsilateral nHCV and HCT2 were at least 2 SDs from the control mean, whereas in 4 patients, either ipsilateral nHCV or HCT2 was normal, indicating only a mild HD according to QMRI. Thus, hippocampal QMRI was congruent with the surgical lobe in 69% (11/16) of the patients. In patient 2, HCT2 was prolonged, and a tumor was found in the medial temporal lobe. In addition to bilateral HD, patient 7 had focal cortical dysplasia in the lateral temporal neocortex. In 5 (31%) of 16 patients, nHCV and HCT2 were referred normal. One of these 5 patients (patient 13) had focal cortical dysplasia in the lateral temporal cortex (Table 4).
Three patients with histopathologically verified mild HD also had significantly milder changes in the mean nHCV and HCT2 than the 2 other groups (Table 3). The mean HCT2 in patients with moderate or severe HD and the mean nHCV in patients with severe HD differed significantly from the control means (P<.001 for HCT2 and nHCV in patients with severe HD; P<.01 for HCT2 in patients with moderate HD). The mean nHCV has been compared with the left HCV of controls (Table 3).
All patients with abnormal and 3 of 5 patients with normal hippocampal QMRI results had ipsilateral temporal hypometabolism. Figure 1 shows patient 8 with normal findings of QMRI but abnormal findings of [18F]FDG-PET.
The left-right ratio and AI in [18F]FDG-PET correlated significantly with the left-right ratio in nHCVs in all temporal areas except the superior temporal gyrus (Table 5). However, the AI in any temporal area was not significantly related to the ipsilateral nHCV even though the relation was studied only in the 11 patients with preoperatively well-established HD in QMRI (data not shown).
The left-right ratio and AI in [18F]FDG-PET and the left-right ratio in HCT2 were significantly inversely related in all temporal areas (Table 5). However, the AI in any temporal area was not associated with ipsilateral HCT2 even when the relation in patients with abnormal hippocampal QMRI findings was studied separately (data not shown).
In no patients was the AI in either lateral temporal neocortical gyri or the medial temporal lobe related significantly to any aspects of the histopathologically verified HD: severity of neuronal loss, proliferation of glial cells, or sprouting of mossy fiber (data not shown). In addition, AIs between patients with differing severity of HD did not differ significantly in any of the temporal areas (Table 3).
The surgical outcome after at least a 1-year follow-up was evaluated according to Engel.35 In patients with a preoperatively assumed good outcome, 12 of 15 became seizure free (9 patients were totally seizure free, class Ia, and 3 patients experienced auras only after surgery, class Ib). One patient was initially seizure free but now has rare seizures (class IIa, <3 seizures per year). The seizure reduction was worthwhile (class IIIa, at least 80% seizure reduction) in 2 patients (patients 4 and 11). Patient 16 had no worthwhile seizure reduction (class IVa) (Table 4).
Because of the bilateral seizure onsets in patient 4, surgery was considered to be palliative. The patient still benefited from surgery (class IIIa) (Table 4).
In this study, temporal hypometabolism even in the medial temporal lobe was shown not to reflect the severity of HD verified by QMRI (volumetry and T2 relaxometry) or histopathological analysis. However, [18F]FDG-PET was lateralizing, even when hippocampal QMRI findings were normal or showed only mild HD in patients with drug-resistant TLE.
All the earlier studies13,18,19 have relied solely on the reduced HCV in MRI to discover the relationship between HD and temporal hypometabolism. However, reduced HCV is not the only feature of HD, and normal volumetry does not rule out the possibility of histopathological changes in the hippocampus,36 but the results of HCT2 must also be taken into account. In this study, the AI in the medial temporal lobe covering, for instance, the hippocampus or in the lateral neocortical temporal gyri were not related either to ipsilateral reduction in HCV or to prolongation in HCT2. In agreement with the results of a previous study,13 this suggested that HD would not be a determining factor for reduced glucose consumption. The significant correlations between the left-right ratio and AI in [18F]FDG-PET and the left-right ratios in HCV and HCT2 merely indicated that these methods were in agreement with lateralizing the side of the epileptogenic region in most patients.13
However, because hippocampal QMRI might be normal even though a patient had histopathological changes in the hippocampus, the relation between the results of [18F]FDG-PET and the aspects of histopathologically verified HD was also examined. The AI was unrelated to the severity of neuronal loss, proliferation of glial cells, or sprouting of mossy fiber, even in the medial temporal lobe. Moreover, the mean AIs in temporal areas did not differ significantly in patients with significant differences in HD. Thus, it seemed that even though the hippocampus was only mildly damaged, glucose consumption was reduced over and above histopathological changes. Therefore, [18F]FDG-PET was considered to be insensitive to reflect the severity of HD and, according to the general assumption, reflects only synaptic reorganization and especially decreased synaptic activity in the whole epileptogenic zone.13,16,37,38
The greatest utility of PET is in patients with normal or uncertain findings in QMRI. In the present study, 31% of patients had normally defined HCV or HCT2. However, [18F]FDG-PET lateralized the epileptogenic zone in 60% of these patients when the generally applied cutoff threshold (15%) in asymmetry was used, thus suggesting the usefulness of [18F]FDG-PET in these patients. In addition, 4 patients with only mild changes in hippocampal QMRI still showed clear reduction in glucose metabolism. An AI of more than 15% was chosen to avoid false-positive findings32 and to predict good surgical outcome.33 If based solely on the results of our controls, a cutoff threshold of 10% was used, 1 patient with normal hippocampal QMRI findings and visibly reduced uptake of [18F]FDG would also have had a quantitatively significant result.
There were some limitations to our study. The spatial resolution of our PET camera did not allow detection of the hippocampus as well as a scanner with better resolution would have done. The partial volume effect or spillover activity of surrounding tissues might have affected the values of glucose consumption, at least in the medial temporal lobe ipsilateral to the epileptic hippocampus with reduced volume. In addition, the PET studies were oriented in the orbitomeatal line, which might not be optimal for studying the medial temporal lobe. Patients also had diverse etiologies and electroencephalographic findings. However, the hippocampus was carefully analyzed after surgery. Because 2 patients with histopathologically moderate HD had normal hippocampal QMRI findings, the question of the ability of QMRI to detect mild to moderate damage is also raised.
Our study provides a detailed analysis of the presence and severity of HD and its relationship to the degree of hypometabolism in patients with TLE.
Accepted for publication December 4, 2000.
This study was financially supported by the Finnish Neurological Association, the Finnish Instrumentarium Foundation, the Turku University Foundation, the Research Foundation of Orion Corporation (Espoo, Finland), and the Päivikki and Sakari Sohlberg Foundation (Helsinki, Finland).
We are grateful to Prof Asla Pitkänen for data on the histopathological analysis of resected tissue and to the staff of the Turku PET Centre for their assistance.
Corresponding author and reprints: Juha Rinne, MD, PhD, Department of Neurology, University of Turku, PO Box 52, FIN-20521 Turku, Finland (e-mail: email@example.com).