Abstract
Objective: To compare the ability of diffusion-weighted MRI (DWI) and 1H MRS to lateralize to the temporal lobe of seizure onset and to predict postoperative seizure control in patients with temporal lobe epilepsy (TLE).
Methods: Forty TLE patients who subsequently underwent epilepsy surgery and 20 normal subjects were studied with 1H MRS and DWI. Medial parietal and temporal lobe N-acetylaspartate (NAA)/creatine (Cr) ratios and hippocampal and temporal stem apparent diffusion coefficients (ADC) were obtained. Lateralization to either temporal lobe with each MR measurement was based on the threshold values derived from ±1-SD right/left ratios of normal subjects.
Results: Temporal lobe NAA/Cr lateralized to the operated temporal lobe in 18 of 40 (45%), hippocampal ADC in 32 of 40 (80%), and temporal stem ADC in 26 of 40 (65%) patients. Almost all of the cases that lateralized to the surgical side with NAA/Cr ratios (94%) had an excellent postoperative seizure control (p = 0.01). Lateralization to the side of surgery was not associated with surgical outcome with hippocampal and temporal stem ADC (p > 0.05).
Conclusion:1H MRS and DWI complement each other in the clinical setting. DWI more frequently lateralized to the operated side, and 1HMRS was a better predictor of postoperative seizure control.
Seizure lateralization in temporal lobe epilepsy (TLE) is usually accomplished through clinical and electrophysiologic evaluation. MRI has an important diagnostic role, especially when a structural substrate is responsible for the seizures.1 Other MR techniques such as 1H MRS and diffusion-weighted MRI (DWI) provide lateralizing information that is independent from that obtained from structural MRI.
The metabolite N-acetylaspartate (NAA) is reduced in the epileptogenic temporal lobe and to a lesser extent in the contralateral temporal lobe.2-8⇓⇓⇓⇓⇓⇓ NAA is present only in neurons, and NAA levels decrease with neuron loss9. There is progressive loss of NAA in patients with TLE,10 and depressed NAA levels in the contralateral temporal lobe have been shown to normalize after cessation of seizures with surgical treatment.11,12⇓ Thus, a reduction of this metabolite may signify not only a decrease in neuronal number but also a physiologic response to epileptogenic activity or neuronal dysfunction in the temporal lobes of patients with TLE.10-13⇓⇓⇓
DWI is sensitive to the Brownian motion of water molecules in biologic tissue.14 Tissue microstructure determines the diffusivity of water in the brain.15,16⇓ Diffusivity of water, measured by the apparent diffusion coefficient (ADC), increases when there is damage to neurons and myelin that normally restrict the random motion of water molecules.16 ADC is higher in the sclerotic hippocampi of patients with mesial temporal sclerosis (MTS) than in normal hippocampi.17-19⇓⇓ Preoperative diagnostic tests that correlate with, and hence provide predictive information about, postoperative seizure control play a major role in selecting or rejecting individual patients for surgery. There have been no studies of which we are aware that assess the correlation between postoperative seizure control and preoperative measures of DWI. In addition, with a single exception,20 no published studies have correlated 1H MRS and DWI findings in a single cohort of patients with TLE.
The objectives of this study were threefold: 1) to compare the accuracy of temporal lobe ADC and NAA/creatine (Cr) measurements for identifying the epileptogenic temporal lobe in patients who subsequently underwent temporal lobectomy for medically intractable TLE, 2) to determine whether there is any association between these two MR measurements, and 3) to determine if seizure lateralization with these MR measurements is associated with postoperative seizure control.
Methods.
Patient recruitment and evaluation.
Over a 2-year period from July 1998 until July 2000, 40 consecutive patients with intractable TLE who subsequently underwent epilepsy surgery were recruited. The inclusion criteria were 1) agreement to participate in the study, 2) scheduled to have surgery, 3) having a pathologic substrate that did not involve the standardized temporal lobe region in which the 1H MRS voxel was placed, and 4) being in the age range of 17 to 50 years in order to avoid introducing possible age effects into our MR data. 1H MRS and DWI were performed while the patients were undergoing a comprehensive preoperative evaluation. This research study was approved by the institutional review board, and informed consent was obtained from every patient. The preoperative clinical evaluation for seizure onset localization included clinical assessment, in-patient video-EEG monitoring, and a clinically indicated structural MR study that was separate from the exam containing the DWI and 1H MRS sequences. Findings on DWI and 1H MRS exams were not used for preoperative evaluation. Time and date of the last seizure before the DWI and 1H MRS exam were recorded. Four of the patients underwent invasive EEG monitoring of seizures because noninvasive EEG evaluation revealed contradictory or bilateral abnormalities and was therefore inconclusive for seizure onset lateralization. All patients underwent anterior temporal lobectomy and amygdalohippocampectomy. The temporal lobe that was operated on based on the comprehensive preoperative evaluation will be referred to as the “ipsilateral” and the other as the “contralateral” temporal lobe throughout this article. Twenty-two of the patients underwent right-sided and 18 of them left-sided surgery. Epileptogenic substrates identified in the surgical specimens were MTS in 28 patients, tumors in 3, cavernous hemangiomas in 3, and developmental malformations in 2 patients. Four patients had no specific epileptogenic substrate identified in the surgical specimen. During the same period of time, we also studied 20 healthy volunteers at ages 21 to 51 to determine the normal ranges of the DWI and 1H MRS measurements.
1H MRS.
MR and single-voxel 1H MRS studies were performed on a 1.5 T scanner (Signa; General Electric Medical Systems, Milwaukee, WI). After an axial scout, T1-weighted images in coronal planes were obtained for localizing the 1H MRS voxels. Special attention was paid to the symmetric positioning of the patient’s head. 1H MRS studies were performed with the LX system automated single-voxel 1H MRS package: proton brain examination/single voxel (PROBE/SV)21 (General Electric Medical Systems). Point-resolved spectroscopy (PRESS) pulse sequence with repetition time (TR) of 2,000 milliseconds, 2,048 data points, and 128 excitations was used for the examinations. An echo time (TE) of 135 milliseconds was used to reduce the contribution from underlying broad resonances. The prescan algorithm of PROBE automatically adjusted the transmitter and receiver gains and center frequency. The local magnetic field homogeneity was optimized with the three-plane auto-shim procedure, and the flip angle of the third water suppression pulse was adjusted for chemical shift water suppression (CHESS) prior to PRESS acquisition.
Three different volumes of interest (VOI) were studied: right and left temporal lobe VOI and a medial parietal VOI. The 7.2-cm3 (2 × 2 × 1.8-cm) right and left temporal lobe VOI were prescribed on a coronal T1-weighted image. It would have been ideal to place the VOI over the anteromedial temporal lobes because seizures are felt to most commonly originate from this area in TLE patients. However, acquisition of reliable spectra from this area was not possible for two technical reasons: 1) difficulty in obtaining spectra from a voxel small enough to sample medial temporal lobe structures without significant partial voluming of the surrounding temporal lobe or extratemporal areas and 2) difficulty in achieving an adequately homogeneous magnetic field (the automated shimming algorithm failed) owing to the magnetic susceptibility effects at the tissue–air interface near the petrous bone. Based on preliminary feasibility studies, the most anterior and medial possible placement of the VOI that consistently produced high-quality spectra was positioned over the superior temporal lobes and inferior insular cortex (figure 1A). The 1H MRS voxels therefore did not include the pathologic substrate (the hippocampi) in patients with MTS. We excluded patients with pathologies extending into the superior temporal lobe who were operated over the time period of this study, so that all the 1H MRS studies would be performed in a consistent manner with anatomically visible structural pathology excluded from the 1H MRS VOI. With this placement, 1H MRS VOI covered only the normal-appearing temporal lobe regions on MRI.
Figure 1. Placement of the temporal lobe (A) and medial parietal (B) 1H MRS volumes of interest on coronal T1-weighted localizing MR images.
In addition to the two temporal lobe voxels, an 8-cm3 (2 × 2 × 2-cm) medial parietal VOI was prescribed on a coronal T1-weighted image and was placed behind the splenium of corpus callosum covering both of the medial parietal lobes (see figure 1B). We studied this region because it is distant from the temporal lobes where the seizures were thought to originate and thus served as a good reference for normality both in normal subjects and in patients. In a small number of subjects, we had to repeat one of the PROBE/SV acquisitions because of head motion, and we were able to obtain technically good-quality spectra from every patient with standard voxel placement.
DWI.
Single-shot echo planar fluid-attenuated inversion recovery (EPI-FLAIR) DWI was performed in the coronal plane with TR of 9,999 milliseconds, TE of 93 milliseconds, inversion time of 2,200 milliseconds, slice thickness of 5 mm, slice spacing of 2.5 mm, and field of view of 40 × 20 cm to cover whole head. A FLAIR image with b = 0 s/mm2 and DWI with b = 1,000 s/mm2 in three orthogonal directions were acquired from each slice. With the image analysis software FuncTool (General Electric Medical Systems), average ADC maps were computed pixel by pixel based on the Stejskal and Tanner equation.14
Elliptical regions of interest (ROI) 12 to 30 pixels (29.3 to 73.2 mm2) in size were drawn over the coronal EPI-FLAIR images that concurrently appeared on the ADC maps (figure 2). Two pairs of ROI were placed in each subject over the white matter of the right and left medial temporal lobes (temporal stem) and the hippocampi. Temporal stem is defined as the white matter connection between the temporal lobe and the frontal and parietal lobes.22 We placed the ROI for ADC measurements over the temporal stem because it partly covered the same region the 1H MRS VOI was covering, which therefore allowed us to obtain comparable data. No MR-visible structural lesions were included in the temporal stem ROI. The hippocampi were also studied using a second ROI because pathologic substrates of epilepsy most commonly involve the hippocampus. Owing to low spatial resolution of the EPI-FLAIR images, we also obtained coronal T1-weighted images that had identical slice thickness and location to be used as an anatomic reference for the placement and tracing of the ROI. The hippocampal ROI were manually traced over the hippocampal heads on the same slice in order to exclude the perihippocampal CSF spaces (see figure 3⇓). The ROI placements and tracings over the ADC maps in each subject were done in a uniform manner by the same investigator, who was blinded to the diagnoses of the subjects.
Figure 2. Coronal T1-weighted image (top) from the level of hippocampal heads is used to guide placement of the regions of interest (ROI). Hippocampal ROI are traced, and temporal stem ROI are placed over the fluid-attenuated inversion recovery (FLAIR) image (middle) that we used as the b = 0 image for calculating the apparent diffusion coefficient map (bottom).
Figure 3. Examples of 1H MRS from the right and left temporal lobes (TL) in a patient with right temporal lobe epilepsy. N-Acetylaspartate/creatine ratios (NAA/Cr) are lower in the right than the left temporal lobe.
Surgical outcome.
Information on surgical outcome was obtained from the medical records in 34 and phone interviews in 6 patients. Patients were followed by the neurologist investigators of this study who assess surgical outcome using the seizure frequency scoring system.23,24⇓ This is a standardized part of clinical practice on all surgical patients. Patients with seizure frequency scores of 0 to 4 (seizure-free or having only nondisabling simple partial seizures) were considered to have an excellent outcome, and those with scores of >4 were considered to have a nonexcellent outcome. Improvement in seizure frequency score by >2 points after surgery was considered as favorable improvement. Lateralizing findings with each MR measurement were compared among patients with excellent and nonexcellent outcome.
Statistical analyses.
Differences in the mean ages of the normal subjects and patients were assessed by t-tests, and differences in male/female ratios of these groups were assessed by the χ2 test. The two hemispheric measurements in control subjects were averaged to create a single value for every measurement in each subject. Differences in MR measurements between patients and normal subjects were tested by rank sum tests. Regression analysis was performed to determine if there was any association between age, temporal lobe NAA/Cr, temporal stem ADC, and hippocampal ADC. Regression analysis was also performed to determine if there was any association between the ipsilateral temporal lobe MR measurements (temporal lobe NAA/Cr, hippocampal and temporal stem ADC) and the time between the last seizure and the MR exam. The level of significance for all analyses was p < 0.05.
Both the 1H MRS and the DWI measurements are continuous variables. To create a single value from the right and left temporal lobe NAA/Cr and ADC measures in each subject, we divided the right-sided measurements by the left. To use such right/left ratios for lateralization, the continuous variables must be segregated into ranges that denote right- and left-sided seizure onset. Two threshold values were derived from the first ±1 SD of the right/left ratio in normal subjects. The range between the threshold values was considered as nonlateralizing. Values higher and lower than the first 1 SD of normal subjects were considered to lateralize to either the right or the left side. With use of these threshold values, the percentages of patients who were concordantly lateralized to the side of surgery, lateralized to the opposite side of surgery, or were nonlateralizing were calculated for each right/left MR measurement.
For purposes of correlating MR lateralization with surgical outcome, patients were classified as having concordant or nonconcordant lateralization to the side of surgery. The group of patients who were classified as nonconcordant were the ones who were nonlateralized or lateralized to the opposite side of surgery. The dichotomous variables (excellent or nonexcellent surgical outcome and concordant or nonconcordant lateralization to the side of surgery) were compared with χ2 tests for each MR measurement.
Results.
There was no difference between the mean ages of patients (36.97 ± 11.97 years) and normal subjects (34.25 ± 8.85 years) (p = 0.53). Male/female ratios of patients (18:22) were not different from those of control subjects (11:9) (p = 0.47). The ethnic background of all patients and control subjects was white, except for one African American in the patient group. The median (range) duration of TLE in patients was 20 (1 to 45) years. Medial parietal NAA/Cr ratios were not different from normal (median [range] 1.84 [1.64 to 2.19]) in the 40 patients with TLE (1.81 [1.12 to 2.18]) (p = 0.24), nor were the measures different from normal in the subset of patients with MTS (1.81 [1.12 to 2.18]) (p = 0.24) or other pathologic substrates (1.79 [1.68 to 2.14]) (p = 0.49). There was no association between age and temporal lobe NAA/Cr ratios, temporal stem, ADC, or hippocampal ADC within a range of 21 to 51 years in 20 normal subjects (p > 0.05).
NAA/Cr and ADC in patients and normal subjects.
The median (range) of temporal lobe NAA/Cr and temporal stem and hippocampal ADC obtained from the ipsilateral and contralateral temporal lobes of TLE patients and those of the 20 normal subjects are listed in table 1. We did not correct p values numerically for multiple comparisons, so the value of >0.01 in table 1 should perhaps be interpreted as a trend. In the ipsilateral temporal lobe of all TLE patients, NAA/Cr ratios were significantly lower and hippocampal and temporal stem ADC were significantly higher than normal. In the contralateral temporal lobe, NAA/Cr ratios were significantly lower than normal, but the hippocampal and temporal stem ADC were not different from normal (figure 3).
Comparison of temporal lobe NAA/Cr and temporal stem and hippocampal ADC in patients and normal subjects
Hippocampal ADC of TLE patients were positively correlated with temporal stem ADC (r2 = 0.131, p < 0.001). On the other hand, temporal lobe NAA/Cr ratios of TLE patients were not associated with temporal stem (p = 0.15) or hippocampal (p = 0.34) ADC.
Median (range) time between the last seizure and the MR exam was 3.7 days (2 hours 10 minutes to 62 days 11 hours). Five of the patients had complex partial seizures within the last 24 hours before the MR exam. There was no correlation between the ipsilateral temporal lobe MR measurements (p = 0.24 for NAA/Cr and p = 0.27 for hippocampal and p = 0.18 for temporal stem ADC) and the time between the last seizure and the MR exam.
Seizure onset lateralization with NAA/Cr and ADC.
For seizure onset lateralization, threshold values were derived from the first ±1 SD of the mean right/left ratios obtained from 20 normal subjects. For right/left NAA/Cr ratios, values of <0.94 were considered to lateralize to the right and >1.16 to the left temporal lobe. For hippocampal ADC, values of >1.06 were considered to lateralize to the right and <0.96 to the left temporal lobe. For temporal stem ADC, values of >1.05 were considered to lateralize to the right and <0.99 to the left temporal lobe. Table 2 lists the percentages of patients in whom side of lateralization was concordant with the side of surgery, who were within the nonlateralizing range, or who lateralized to the opposite side of surgery with each measurement.
Concordant, nonlateralizing, and nonconcordant lateralization with respect to side of surgery with each right/left MR measurement in 40 TLE patients
With right/left NAA/Cr ratios, lateralization to the side of surgery was possible in 18 of 40 (45%) patients (MTS 14/28 patients, other substrates 3/8 patients, no specific substrate 1/4 patients). Findings were nonlateralizing in 18 of 40 (45%) of the cases (MTS 12/28 cases, other substrates 3/8 cases, no specific substrate 3/4 cases). Findings lateralized to the opposite side of surgery in 4 of 40 (10%) of the cases (MTS 2/28 cases, other substrates 2/8 cases, no specific substrate 0/4 cases).
With right/left hippocampal ADC ratios, lateralization to the side of surgery was possible in 32 of 40 (80%) patients (MTS 25/28 patients, other substrates 4/8 patients, no specific substrate 3/4 patients). Findings were nonlateralizing in 6 of 40 (15%) of the cases (MTS 3/28 cases, other substrates 3/8 cases, no specific substrate 0/4 cases). Findings lateralized to the opposite side of surgery in 2 of 40 (5%) of the cases (MTS 0/28 cases, other substrates 1/8 cases, no specific substrate 1/4 cases).
With right/left temporal stem ADC ratios, lateralization to the side of surgery was possible in 26 of 40 (65%) patients (MTS 21/28 cases, other substrates 4/8 cases, no specific substrate 1/4 cases). Findings were nonlateralizing in 10 of 40 (25%) of the cases (MTS 5/28 cases, other substrates 2/8 cases, no substrate 3/4 cases). Findings lateralized to the opposite side of surgery in 4 of 40 (10%) of the cases (MTS 2/28 cases, other substrates 2/8 cases, no substrate 0/4 cases).
Surgical outcome.
The median (range) follow-up time after epilepsy surgery was 26 (12 to 36) months. Thirty of the 40 TLE patients had an excellent outcome (seizure frequency score 0 to 4). Except for one patient who had three nondisabling nocturnal seizures after decreasing the dose of the antiepileptic drug, all of the patients classified as having an excellent outcome were completely seizure-free. Ten of the patients continued to have seizures after surgery and were classified as having a nonexcellent outcome.26 None of the patients who were initially classified as having an excellent outcome during the first year were classified as having a nonexcellent outcome during the second year after surgery. Of the patients who experienced seizures after surgery, two of them had a decrease and one patient had an increase in seizure frequency during the second year after surgery.
Figure 4 shows the scatterplots of right/left MR measurements in patients by an excellent versus nonexcellent outcome. Table 3 compares the MR measurement lateralization to surgical outcome. The only MR measurement for which surgical outcome differed significantly between patients who had concordant versus nonconcordant lateralization was temporal lobe NAA/Cr ratios (p = 0.01). Almost all (17/18; 94%) of the patients whose right/left NAA/Cr ratios lateralized to the side of surgery had an excellent outcome. Only one patient whose NAA/Cr ratios lateralized to the side of surgery had nonexcellent outcome. This patient experienced four consciousness-impairing seizures within a 2-month period, starting at 2 months after surgery. The patient had been seizure-free for the last year after these episodes, with a decrease of three seizure frequency scores since before surgery, indicating a favorable improvement. Twenty-two patients had nonconcordant lateralization to the surgical side. Nine of 22 patients who were nonlateralizing (40%) had nonexcellent surgical outcome. The remaining 13 patients had excellent outcome. Nine of these 13 patients were nonlateralizing and 4 lateralized to the opposite side of surgery.
Figure 4. Scatterplot of right/left N-acetylaspartate/creatine ratio (NAA/Cr) (A) and hippocampal (B) and temporal stem (C) apparent diffusion coefficient (ADC) in patients who underwent right temporal lobe epilepsy surgery (RTLE) and left temporal lobe epilepsy surgery (LTLE). The patients who had excellent outcome are represented with filled circles and patients with nonexcellent outcome with open circles. Solid lines indicate the threshold values derived from the ±1 SD of normal subjects for determining the side of lateralization in patients.
Seizure onset lateralization with respect to side of surgery in patients with excellent and nonexcellent outcome
Four of the patients had a normal MRI, and in all of these no epileptogenic substrate was identified in the surgical specimen. Lateralization with NAA/Cr was concordant with the side of surgery in only one of these patients, and this patient had an excellent postoperative outcome. In the remaining three patients, side of NAA/Cr lateralization was not concordant with the side of surgery. One of those patients had an excellent and the other two had a nonexcellent outcome.
Discussion.
This study evaluated single-voxel 1H MRS and DWI for seizure onset lateralization and association with postoperative seizure control. The side of lateralization with right/left measurement ratios was based on two cutoff values that were determined from +1 SD and −1 SD of the MR values in normal subjects. Values within this range were considered as nonlateralizing.25 Our rationale in selecting this thresholding method, which demarcates a range of values as nonlateralizing, was based on the fact that some degree of right/left asymmetry is present in normal subjects. This is presumably a function of measurement noise, biologic variability, or both. It seemed inappropriate to use a method, for example, a single threshold value with no nonlateralizing range, that would label one of the temporal lobes as “epileptogenic” in subjects. We believe that in a clinical setting, the thresholds selected ought to appropriately label patients with extratemporal lobe epilepsy and pseudoseizures as nonlateralizing in addition to lateralizing patients with TLE as accurately as possible.
Temporal lobe NAA/Cr ratios were lower in the epileptogenic and to a lesser extent in the contralateral temporal lobe in TLE patients. In this study, temporal lobe 1H MRS voxels were not positioned over pathologic substrates that were apparent on MRI, including the hippocampi in MTS. Therefore, our findings indicate that in TLE, the decrease in NAA/Cr involves a substantial portion of the epileptogenic (and the contralateral) temporal lobe. A similar decrease in NAA/Cr levels has been shown in the medial temporal lobes of patients with extrahippocampal and extratemporal pathologies,26 which further supports this conclusion. The fact that we did not find any difference between the medial parietal NAA/Cr in patients and normal subjects indicates that there is a significant decrease in NAA/Cr at the temporal lobes, where the seizures originate, but not in the medial parietal lobe, a region remote from the site of seizure onset in TLE.
We did not find any association between ipsilateral NAA/Cr and the time between the last seizure and the 1H MRS exam. Our findings are in agreement with a previous study, which showed that NAA ratios remain stable during the interictal and postictal state.27
In other published series, lateralization with 1H MRS findings was concordant with EEG in 60 to 97% of cases.5,7,28-30⇓⇓⇓⇓ Reasons for the lower accuracy of NAA/Cr ratios for identifying the side of surgery in our data (45%) than other studies may be explained by the heterogeneity of the pathologic substrates in the TLE patients we studied. In addition the high variation in the right/left NAA/Cr of normal subjects increased the number of patients classified as nonlateralizing using criteria derived from these normal subjects. Had we placed the voxel over the anteromedial temporal lobe, the sensitivity of 1H MRS for lateralizing to the side of surgery would logically have been higher in patients with MTS.28 However, such voxel placement would have biased the data in MTS patients relative to patients with other substrates in whom the pathologic substrate would be outside the 1H MRS voxel boundaries. An additional consideration was based on our experience that the quality of the spectra is unacceptably poor in some subjects if the voxel is placed too close to the skull base. We therefore elected to position the voxel slightly off the floor of the middle fossa so that good-quality spectra could be obtained uniformly in all subjects. Methodologic differences between previous studies and this study include patient characteristics, spectroscopy techniques, temporal lobe regions that were studied, magnetic field strength, and the criteria used for lateralization.5,7,28-30⇓⇓⇓⇓
Lateralizing findings that are concordant with EEG do not always guarantee a satisfactory surgical outcome.29,31⇓ Thus, the gold standard for evaluation of an imaging modality for surgical localization is its ability to predict surgical outcome. The literature is discordant on the issue of predicting outcome with MRS. One study in TLE patients (n = 24) showed that findings with hippocampal 1H MRS were not associated with surgical outcome.29 On the other hand, in TLE patients with a normal MRI or bilateral hippocampal atrophy (n = 21), concordant lateralization to the side of surgery with 1H MRS predicted a good surgical outcome.32 In a third study of 40 TLE patients at 4.1 T, higher Cr/NAA in the nonoperated temporal lobe was associated with surgical failure.33 Our data revealed that concordant lateralization to the side of surgery by 1H MRS was almost always associated with an excellent outcome. With the one exception of a patient who had favorable improvement after surgery, all of the patients with concordant lateralization of NAA/Cr ratios to the side of surgery were completely seizure-free at least 12 months after surgery. Although concordant lateralization to the side of surgery was possible in only 45% of the patients, temporal lobe NAA/Cr ratios were able to predict excellent surgical outcome with an accuracy of 94% in this group. Conversely, nonconcordant lateralization to the side of surgery did not rule out an excellent outcome. Only 40% of the patients who had nonconcordant lateralization had nonexcellent outcome. Certainly, larger numbers of patients need to be studied to definitively prove or disprove the ability of MRS to predict outcome. However, our data indicate that a relationship does exist.
Diffusivity of water molecules was higher in the hippocampus and the temporal stem that was operated on compared with the contralateral and the normal temporal lobes. Hippocampal ADC has been shown to be higher than normal in the ipsilateral but not the contralateral hippocampus in patients with MTS.17-19⇓⇓ Besides confirming this finding, our study also revealed that temporal stem ADC were higher in the ipsilateral but not in the contralateral temporal lobe of patients with MTS.
In the brain, diffusivity of water increases with disruption of structural barriers such as cell membranes, axons, or myelin that normally restrict the random movement of water molecules. Neuron loss has been documented in the hippocampi of patients with MTS and to a lesser extent in TLE patients with extrahippocampal pathology.34 Therefore, increased hippocampal diffusivity in the ipsilateral temporal lobe of patients with MTS may be explained by a decrease in hippocampal neuronal density, gliosis, or both. The temporal stem serves as a pathway for white matter tracts that connect the temporal lobe to the parietal and frontal lobes.22 An increase in temporal stem ADC may be due to Wallerian degeneration secondary to neuron loss in the anteromedial temporal lobe similar to forniceal atrophy in MTS.35,36⇓ A positive correlation between the hippocampal and temporal stem ADC in our data supports this notion. Another assumption for increased temporal stem ADC would be expansion of extracellular space due to gliosis that is commonly present in the temporal lobes of patients with TLE.
ADC decreases in the epileptogenic region of the brain during status epilepticus and normalizes after recovery.37-39⇓⇓ Although the temporal course of this ADC depression is not so clear in humans, a maximum decrease in ADC occurs 24 to 72 hours after experimental status epilepticus in rats.40-42⇓⇓ The decrease in ADC normalizes within a week and increases further in the long term similar to our findings.40,41⇓ In the patients with TLE that we studied, there was no association between ipsilateral ADC measurements and the time period between the last seizure and the MR exam. The reason may be that none of our patients experienced status epilepticus during their evaluation and postictal ADC changes in shorter seizures may not be similar to status epilepticus.43
Of the DWI measurements, hippocampal ADC lateralized to the side of surgery in 80% of the patients, higher than the temporal stem ADC (64%). Surgical outcome was not different in patients with either concordant or nonconcordant lateralization to the surgical side with hippocampal and temporal stem ADC. In fact, with hippocampal ADC, the percentage of patients who had an excellent outcome was the same (75%) for both concordant and nonconcordant lateralization groups. Although lateralization with hippocampal ADC was highly concordant with electrophysiologic findings, it was not associated with surgical outcome.
Accuracy of hippocampal ADC for lateralizing to the side of surgery was higher than temporal lobe NAA/Cr ratios and temporal stem ADC. This is undoubtedly because the majority of the patients in this study had MTS, and hippocampal ADC is a direct hippocampal measurement, whereas the other two measurements were obtained from other regions in the temporal lobe. However, temporal stem ADC more frequently lateralized to the operated temporal lobe than NAA/Cr ratios. Both the temporal stem ADC and the 1H MRS measurements were obtained from regions outside the anteromedial temporal lobe, where most seizures originate. A possible explanation for the greater lateralizing sensitivity of temporal stem ADC versus 1H MRS is that ADC measurements from the contralateral temporal lobe were normal, but contralateral NAA/Cr ratios were lower than normal. This decreased the asymmetry of the NAA/Cr ratio measured from the two temporal lobes and may be the reason for the lower lateralizing sensitivity of 1H MRS than DWI in this study.
We found no association between NAA/Cr and ADC measurements, which suggests that they are sensitive to different phenomena in the pathology of TLE. This is further evident in the lateralizing findings of patients with excellent surgical outcome. Whereas concordant lateralization of NAA/Cr ratios to the side of surgery was associated with excellent outcome, hippocampal and temporal stem ADC were not.
Thirty-six of the 40 patients had lesions at surgical pathology, and in every case the preoperative structural MR exam identified the nature and location of the pathologic substrate. Because the MR results were used to select patients for surgery, we did not consider the structural MR data appropriate for research analysis. Only four patients had no epileptogenic substrate identified. We therefore have insufficient data to rigorously analyze the clinical utility of 1H MRS and DWI in TLE patients with no pathologic substrate, that is, negative MRI. Our results suggest that 1H MRS measurements may provide predictive information also in patients with no pathologic substrates. However, a larger prospective study is required to arrive at a definitive conclusion.
It is unfortunate that patients with ambiguous lateralization by structural MRI with electroclinical criteria, which stand to benefit most from newer MR techniques, are precisely those in whom validation is most difficult to obtain. Like most studies aimed at validating the lateralizing utility of an imaging modality, we studied only patients who underwent epilepsy surgery and in whom seizure onset localization was therefore well enough defined to warrant surgical resection. The greatest clinical usefulness of 1H MRS and DWI would be in TLE patients with ambiguous lateralization. Our findings do not address this question but provide a basis for further studies in such patients.
Acknowledgments
Supported by NIH-NS28374.
Acknowledgment
The authors thank Kathleen M. Cicora for her help in recruitment of patients.
- Received September 6, 2001.
- Accepted in final form March 12, 2002.
References
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.