MR SPECTROSCOPY IN EPILEPSY

Epilepsy, the condition of recurrent seizures, is a relatively common neurological disorde. A multitude of etiologies cause epilepsy, including tumors, developmental abnormalities, febrile illness, trauma, or infection. However, not infrequently, the cause is unknown. Many patients with epilepsy can be successfully treated pharmacologically, but when medical management fails to adequately control seizure activity, surgical resection of the epileptogenic tissue may be considered. For surgery to be successful, seizures must be of focal onset from a well-defined location. It has been estimated that up to 10% of patients with epilepsy are medically intractable, of whom approximately 20% may be candidates for surgical treatment.
Traditionally, scalp electroencephalography (EEG) and often invasive (subdural grid or depth electrode) EEG are used to identify the epileptogenic regions of the brain, but increasingly magnetic resonance imaging (MRI), positron emission tomography (PET), ictal single photon emission computed tomography (SPECT), and, more recently, magnetoencephalography (MEG) are also used.
MRI is the modality of choice for identifying brain tumors, cortical malformations, infectious and other causes of epilepsy. In mesial temporal sclerosis (MTS), the most common abnormality in patients with temporal lobe epilepsy, MRI typically shows hippocampus volume loss, with abnormal signal intensity on T2-weighted images, which corresponds histologically to neuronal loss and gliosis. Sensitivity of MTS detection may be increased by performing careful, quantitative T2 measurements from multiple echo data acquisitions, or by using the CSF-suppressed FLAIR sequence. Quantitative volume measurements more reliably detect small changes in hippocampal volume and are generally preferable, particularly when atrophy may be subtle. Lateralization of seizure focus in patients with temporal lobe epilepsy has been reported to be over 90% efficient with volumetric analysis of hippocampal and amygdaloid formations using high-resolution 3D scans. While these studies show that MRI is a sensitive tool for the detection of MTS, the clinical significance of these findings should be carefully considered. First, many published studies have been performed in retrospectively selected patients who were already candidates for epilepsy surgery by other criteria, such as EEG. This may increase sensitivity and specificity by excluding patients who might have negative MRI findings, or who are “complicated” cases. Second, a significant number of patients will have symmetric hippocampi (either no atrophy or bilateral atrophy), and yet still have successful seizure control after surgery, indicating that bilateral sclerosis is not necessarily a contra-indication for surgery. Third, longer-term follow-up post-surgery is often not reported; one study found seizure-free outcome in 70–80% of patients 1 year after surgery, but by 5 years this number had fallen to 50–60%. Interestingly, relapse only occurred in the patients who originally presented with hippocampal atrophy. Collectively, these studies demonstrate that MRI is a valuable tool for the evaluation of patients with epilepsy, but that it also has limitations. For this reason, other imaging studies are often considered for the evaluation of epilepsy patients, particularly “functional” techniques that measure blood flow and metabolism, such as SPECT and PET.
Epilepsy has been extensively studied by PET since the early 1980s, the majority of the studies using 18F-fluorodeoxyglucose (FDG) to measure glucose metabolism. Interictally, glucose uptake is reduced compared to normal brain, while ictally increases in uptake may be observed. The hypometabolic region is usually larger on interictal PET scans than the electrically defined volume of pathology. Seizure foci also have been found to be associated with changes in cerebral perfusion, which can be monitored with oxygen-15 PET, SPECT, or perfusion MRI. In one study, PET and MRI were determined to be comparable in terms of ability to detect abnormalities in patients with temporal lobe epilepsy, but PET had better concordance than MRI with the EEG localization.
Worse surgical outcome was also associated with hypometabolism which extended beyond the region of the temporal lobe. MRI and PET imaging are usually done interictally for logistical reasons.
The main advantage of SPECT, despite having lower resolution and being less quantitative than PET, is that SPECT can be performed ictally, i.e. the SPECT tracer is injected while (or just after) the patient experiences a seizure. Ictal SPECT may provide unique information about the location of the epileptic focus.
All together, these studies demonstrate the additional value of flow and metabolic based imaging studies, in addition to structural anatomic scans, for the evaluation of patients with non-lesional epilepsy.
 

MR spectroscopy in epilepsy
31P MRS
The use of MR spectroscopy for the metabolic evaluation of animal models of epilepsy was first investigated by the Yale group in the early 1980s. At that time, 31P was the most widely used nucleus for in vivo MR studies of the brain, although some proton MRS studies were also performed. Generally, status epilepticus is associated with reductions of high-energy phosphates (nucleotide triphosphates (NTP) and phosphocreatine (PCr)), increases in low-energy phosphates (inorganic phosphate (Pi)), and cerebral acidosis as determined by the chemical shift of the Pi peak. However, not all studies report these findings; in an interictal cortical spike focus in the rat, no significant 31P MRS or MRI changes were reported. After these initial results in animal models, the use of 31P MRS in humans with seizure disorders was reported in the late 1980s and early 1990s. Interictally, seizure foci were found by 31P MRS to be alkaline, and this was proposed as a means of lateralization of the seizure foci, although this finding was not reproduced subsequently. 31P MR spectroscopy of infants experiencing status epilepticus showed a decrease in the PCr/Pi ratio, and in general, in most forms of adult epilepsy, the most common finding (ictally or interictally) appears to be bioenergetic impairment (i.e. reduced ratios of PCr/Pi and/or PCr/ATP). In one study using 31P MRS at high field (4.1 T) in a group of 30 patients, 31P MRS successfully lateralized temporal lobe epilepsy in 70–73% using either PCr/Pi or ATP/Pi ratios, a rate that was actually higher than that achieved with MRI in the same study. An example of a 31P MRS of a 2-year-old child with Lennox–Gastaut syndrome, before and after initiation of the “ketogenic diet” for seizure control, a small but noticeable increase in PCr can be determined.
However, the relatively coarse spatial resolution and low sensitivity of MRS (≈30 cm³ voxel size for human brain studies at 1.5 T) limit the application of this technique to rather large focal abnormalities or diffuse brain pathologies. The technique does not appear to be able to map the extent of epileptogenic tissue because of its low spatial resolution. Finally, it is not particularly widely available since appreciable hardware modifications are required on most MRI scanners as a result of the lower resonant frequency
of the 31P nucleus. For all of these reasons, there have been many more proton MRS studies of epilepsy than 31P, although with high-field (i.e. 3 T and above) MRI systems becoming increasingly available, there is still some interest in using 31P to investigate the biochemical processes occurring in patients with epilepsy.
High-field 31P MRS offers higher sensitivity which results in better spatial resolution (~ 6–12 cm³ in 40–50 min scan time) compared to lower fields.
Proton MRS
Proton spectroscopy has a considerable sensitivity advantage compared to 31P which allows significantly better spatial resolution, and can be used on most MRI scanners without hardware modifications. Over the last several years, therefore, most spectroscopy studies of human epilepsy have utilized the proton nucleus. The first published study involved two patients with Rasmussen’s syndrome (both of whom had abnormal MRI scans). Both patients showed decreased N-acetyl aspartate (NAA), and the single patient who had seizures during spectral acquisition showed increased lactate. Since the NAA signal is believed to originate from neuronal cells, the reduction in NAA has been attributed to neuronal loss within the seizure focus, which is also a common histological finding. Increased lactate in patients who are experiencing active seizures is consistent with the hypermetabolism observed in ictal FDG-PET scans, indicating that the increased glucose uptake is at least partly metabolized anaerobically to lactate, as opposed to the normal path through pyruvate to the TCA cycle. Ictal, or early post-ictal (up to about 6 h) elevations of lactate have therefore been found to be useful in the identification of seizure foci. However, most spectroscopy studies of epilepsy are performed interictally (where lactate is not normally observed), and the most universal finding associated with seizure foci is a decrease in NAA, either measured quantitatively or as a ratio to creatine, ratio to choline, or both. In some cases, increases in choline may also be observed, perhaps as the result of gliosis or neoplastic proliferation, since glial cells are believed to have high choline levels. In short echo time spectroscopy, lower levels of glutamate and glutamine (Glx) than in control subjects have also been reported in patients with hippocampal sclerosis. This is consistent with lower glutamate (the major component of the “Glx” peak) in association with neuronal loss – in an in vitro study of temporal lobe specimens from patients undergoing epilepsy surgery, it was found that Glu gave an excellent correlation with NAA, with both Glu and NAA showing trends for negative correlations with hippocampal neuronal counts. Consistent with this, an MRSI study in temporal lobe epilepsy also found lower levels of Glu in patients with temporal lobe epilepsy: lower in the ipsilateral temporal lobe, but lower than healthy controls in the contralateral temporal lobe as well. Finally, in vitro NMR spectroscopy studies of perchloric acid extracts of gliotic hippocampal tissue have also shown increased myo-inositol and decreased glutamate, consistent with gliosis and neuronal loss.
Temporal lobe epilepsy
In early published studies of patients with temporal lobe epilepsy (TLE), reductions of NAA in the affected hippocampus were found in 100%- 88% (note 40% of the cases had bilateral reductions, so that lateralization was obtained in 60% of cases, 90%, 100%, and 100% of cases studied. An example of proton MRS in a patient with unilateral mesial temporal sclerosis, showing decreased NAA, is depicted in Figure 1. Generally, sample sizes in these studies were in the range between 10 and 25 subjects, and they most likely contained carefully pre-selected cases with clear-cut abnormalities on other modalities. Larger and more recent studies have reported more variable success rates in terms of lateralizing abnormalities in patients with temporal lobe epilepsy; in one MRSI study in 50 patients, success of localization (based on neuroradiological interpretation of spectra) varied from 62% to 76%, while in another the NAA/Cr ratio only successfully lateralized 18 of 40 cases (45%); however, cases which were lateralized by MRS had excellent surgical outcomes. In a study of 100 cases (using MRSI and volumetric MRI), MRSI was found to correctly localize 86% of cases, very similar to volumetric MRI and EEG localization rates.
Despite published studies such as these with high success rates, MRS and/or MRSI have had relatively little clinical impact over the last few years for presurgical evaluation of epilepsy patients. There are probably several reasons for this.
(1) Research studies typically pre-select well characterized patients to study, who are somewhat different from the general epileptic population, and not typical of the “difficult” cases that may be referred for special MRS studies.
(2) The spectroscopic changes are subtle, so while group studies may show statistically significant differences, decision-making confidence in individual patients may be low.
(3) The site of abnormality in many patients, the anterior mesial temporal lobe, is often in a region of poor field homogeneity because of magnetic susceptibility effects from adjacent paranasal sinuses and mastoids, leading to poor quality spectra that are difficult to interpret.
(4) Metabolic abnormalities may be bilateral, even in patients with unilateral MTS or who have good surgical outcome following unilateral temporal lobectomy (i.e. seizure free after 1 year – class 1 on the Engel surgical outcome scale).
(5) In many patients, MRI and MRS may be concordant, which, while improving confidence in the diagnosis, may not warrant the performance of the MRS study in addition to conventional MRI.
Points (4) and (5) above warrant some extra discussion below.
Diffuse metabolic abnormalities in temporal lobe epilepsy
As already indicated, MRS may frequently show bilateral hippocampal abnormalities (i.e. NAA is lower than normal control values in both left and right hippocampi in patients with seizures). This could reflect bilateral sclerosis, or it could represent the effect of seizure propagation from the epileptogenic hippocampus to the other side, causing metabolic impairment. In this regard, it is interesting to note that the metabolism of the contralateral hippocampus typically “improves” after ipsilateral surgery and seizure control, suggesting neuronal dysfunction rather than irreversible neuronal loss. This effect seems to occur over a time period of months following surgery. Conversely, untreated patients may show progressive worsening of NAA/Cr ratios overtime.
More recent studies using MRSI with 1 ml nominal spatial resolution have suggested that the network of brain regions affected by seizures originating in the mesial temporal lobe can be mapped by looking for correlations between metabolite levels in different regions of the brain. Figure 2 shows an example of network connection derived from MRSI; in the TLE patients in this study, in addition to low NAA/Cr in ipsi- and contra-lateral hippocampi, NAA/Cr was also lower than controls in both ipsi- and contralateral thalami. Furthermore, ipsi-lateral hippocampal NAA/Cr values were correlated with contralateral hippocampi, and ipsi- and contra-lateral thalami and putamina, suggesting these structures are all functionally linked and metabolically affected by seizure activity.
Seizure activity may also result in more widespread metabolic abnormalities; for instance, it has been found that frontal lobe NAA levels are lower in TLE patients than in controls in both gray and white matter regions, as well as in other lobes. Conversely, in patients with epilepsy in the neocortex, hippocampal NAA reductions have also been reported. These factors should be kept in mind when using MRSI to evaluate whether seizures are of temporal, neocortical, or extratemporal origin; however, in making this distinction, the largest metabolic abnormality is generally reported to be in the site of seizure onset.
Finally, it is interesting to note that many published MRS studies are apparently successful in identifying the epileptogenic temporal lobe, despite the fact that large regions-of-interest are often used for spectroscopic
analysis (e.g. 8 cm³ for single-voxel studies). Voxels are even larger for 31P studies (e.g. ≈ 30 cm³ or larger). Since the hippocampus occupied only a small fraction of these localized volumes, these results might indicate that there are diffuse spectroscopic abnormalities in the temporal lobe, even when the only MRI finding is that of hippocampal atrophy. These results are therefore consistent with the common observation by PET of extensive hypometabolism throughout the temporal lobe. Since seizure control is often obtained by selective amygdalohippocampectomy, clearly not all of the metabolically abnormal tissue is epileptogenic.
MRS in TLE cases where MRI is normal, or symmetrically abnormal
Arguably, the most useful scenario for MRS is when results from other modalities (particularly MRI) are either normal, ambiguous, or bilaterally abnormal. In these TLE patients, MRS has the potential of lateralizing the epileptic focus. Presence of bilateral MRI abnormalities does not necessarily indicate a poor surgical outcome, but does increase the difficulty in correctly lateralizing the epileptogenic source. In an MRSI study of 21 patients with bilateral hippocampal atrophy who were operated on the side of greatest EEG abnormality, it was found that factors in favor of good surgical outcome were: (1) concordant MRSI EEG localization; (2) greater asymmetry of NAA/Cr between hippocampi; and (3) an absence of contralateral posterior NAA/Cr abnormalities.
MRS may also play a role when MRI is normal – in a study of 7 patients with intractable epilepsy but completely normal MRI findings, it was found that 5 of 7 cases had abnormal NAA/(Cr+Cho) ratios, 2 of which were bilateral. Although this study did not report detailed EEG correlation or surgical outcome, it did suggest that MRS may provide additional information when MRI is normal. NAA has also been found to be lower than normal control values in the ipsilateral hippocampus (to EEG) in another study of MRI negative patients. However, another study found that metabolic abnormalities (in particular, well localized NAA asymmetry) surprisingly did not predict seizure-free outcome after surgery, although the presence of contralateral abnormality did predict poor outcome in this group.
Cortical malformations
There have been a number of reports of MRS in patients with malformations of cortical development (MCD). Despite their frequent epileptogenic nature, MCDs typically show only subtle (or sometimes no) metabolic abnormalities; when metabolic abnormalities are observed, most commonly NAA is reduced and Cho increased, particularly for focal cortical dysplasias. As with other types of epilepsy, metabolic changes remote from the presumed seizure focus (e.g. in the contralateral hemisphere) may be different (typically lower NAA/Cr) from healthy controls, and in fact were not significantly different from the ipsilateral side.
Frontal lobe epilepsy
There have been fewer reports of MR spectroscopy in extratemporal epilepsy than in TLE. Garcia et al. have studied frontal lobe epilepsy using both 31P and 1H MR spectroscopy. In the proton study, all eight cases exhibited a reduced NAA/Cr ratio in the epileptogenic tissue compared to an anatomical similar contralateral location. Stanley et al. also reported the results of proton spectroscopy imaging in 20 cases with frontal lobe epilepsy. As in TLE, it was found that the ratio of NAA/(Cho+Cr) successfully lateralized the epileptogenic tissue as defined by EEG. Widespread NAA reductions were also noted (i.e. the contralateral NAA was also lower than control values), indicating extensive neuronal loss not confined to just the side of seizure onset, as commonly observed in TLE.
Childhood epilepsies, Rasmussen’s encephalitis
A number of epilepsies of childhood have been studied by MRS. MRS has also been used to investigate cerebral metabolism in the “ketogenic diet”, which is becoming increasingly popular as an alternative to pharmacological means of seizure control – using 31P MRS, improvements in bioenergetic status have been reported while proton MRS has shown that the ketone bodies such as β-hydroxy-butyrate and acetone may be detected. Rasmussen’s encephalitis (RE) is a rare, chronic, and progressive epilepsy of childhood involving one hemisphere of the brain. While the cause is largely unknown, currently the only treatment that can reliably provide effective seizure relief is hemispherectomy.
Definitive diagnosis is usually made on the basis of clinical, electroencephalographic, and neuroimaging findings; however, in the early stages diagnosis may not be straightforward. Breiter et al. found hemispheric NAA reductions in 5 cases of Rasmussen’s syndrome, MRSI the whole hemisphere shows low NAA, consistent with neuronal loss. It can also be seen that choline is elevated in the affected hemisphere, particularly the white matter, consistent with the microglial proliferation typically seen on pathology. The hemispheric nature of the metabolic involvement confirms the clinical observation that complete hemispherectomy is necessary for effective seizure control in most cases.
An alternative pattern of involvement shows a more regional distribution, with high Cho and low NAA primarily occurring in the insular cortex, putamen, and frontal lobe.
Neurotransmitters: brain GABA levels
While the majority of MRS studies of epilepsy have studied the most readily observed metabolites Cho, Cr, and NAA, by the use of spectral editing methods it is also possible to measure the inhibitory neurotransmitter, γ-aminobutyric acid (GABA). Typically, macromolecules and the dipeptide homocarnosine also co-edit with GABA, so that the peak observed by MRS is sometimes labeled GABA+ to distinguish it from “pure” GABA. It has been reported that GABA may be (globally) decreased in the brains of epilepsy patients with poor seizure control, and that GABA levels can be increased (and seizure control obtained) using the antiepileptic drugs vigabatrin, topiramate,] and gabapentin. Techniques of this type are promising for monitoring the effects of therapy and establishing optimal drug dosages, and will be more widely used as the MEGA-PRESS editing technique becomes commercially available. It should be noted that other compounds of potential clinical significance, such as glutamate and N-acetyl-aspartyl-glutamate (NAAG) can also be measured using MEGA-PRESS.
Some recommendations for MRS protocols for patients with epilepsy
Since metabolic changes in interictal epilepsy patients are often subtle, high quality MRS with good SNR is essential. Ideally, high-field field scanners (e.g. 3 T) with multiple phased-array receiver coils should be used. For temporal lobe epilepsy, the simplest protocol is to compare the body of the left and right hippocampi at intermediate TE (typically 140 msec) using single-voxel PRESS, or PRESS-MRSI, angulated along the long axis of the hippocampus. While this may seem relatively simple, in fact considerable care has to be taken with these protocols. Voxels should be positioned carefully (both because of the small structures to be observed, and also because of metabolic and field homogeneity changes along the hippocampus), preferably using full 3-view localizers (axial T1-weighted 3D “MP-RAGE” scan reconstructed in both sagittal and coronal views works well. High bandwidth slice-selective RF pulses are preferable to minimize left–right asymmetries due to chemical shift displacement effects. Second-order shimming is also important (especially at higher fields such as 3 T), particularly for MRSI protocols, since significant non-linear field inhomogeneities occur in the temporal lobes which cannot be corrected using linear shims alone. At the anterior tip of the temporal lobe (pes or head of the hippocampus), field homogeneity is usually particularly poor because of susceptibility effects from the nearby paranasal sinuses.
Poor field homogeneity results in insufficient quality spectra for analysis in most adult subjects, even with high order shimming corrections. This is unfortunate, since this is often the target of surgical resection when anterior mesial temporal lobectomies are performed, and is presumably the primary site of pathology. Because of the poor field homogeneity in anterior regions, as well as potential lipid contamination from retro-orbital fat and other skull base structures, it is helpful to apply saturation bands in these anterior regions.
For extratemporal lobe epilepsy, the site of the epileptogenic focus may be unknown, or even when known (e.g. MCDs) the spectroscopic findings may be heterogeneous, therefore in these cases the best approach is to use MRSI with high spatial coverage (e.g. multi-slice or 3D). MRSI is also important since the observation of abnormal metabolism remote from (e.g. surrounding, connected via fiber pathways, or contralateral to) the primary focus is also commonly reported, and may be of clinical significance.
Since MRS changes in epilepsy are generally subtle, demands are placed on the accuracy of both the acquisition technique and spectral analysis software methods. For single-voxel epilepsy MRS, the LC model software is particularly recommended, since it can provide metabolite concentrations and yield uncertainty estimates. In addition, it is advisable to have matched control data using the same scanner, brain region, and MRS technique for comparison; more advanced studies use statistical tests (such as a z-score) to estimate how abnormal any particular spectrum may be. Precision will also be improved by applying corrections to metabolite concentrations according to the gray matter, white matter, and CSF composition within the localized voxel.
Conclusions
In summary, MRS of epilepsy is now a relatively mature field, with the reduction of NAA in abnormal tissue the most common finding. Despite this observation, and the relatively high reported sensitivity and specificity of MRS for seizure focus lateralization in TLE in most research studies, the technique has not found widespread application in clinical practice. This is due to the reasons that are listed in the introduction, of which the most likely is the relatively subtle metabolic changes that are found in most patients. Such subtle changes, along with data that are susceptible to minor instrumental imperfections, make interpretation of individual studies challenging. The “added-value” of MRS compared to other diagnostic techniques remains questionable, although it is apparent that it may be helpful in at least some of the cases that are MRI negative or symmetrically abnormal. As high field scanners and proton MRS techniques improve (i.e. with improved SNR and accuracy of quantitation), and as the use of editing techniques for neurotransmitters such as GABA or glutamate increases, it is hoped that the clinical utilization of MRS in the evaluation of patients with epilepsy will increase. In the long run, MRS may help obviate the need for invasive EEG procedures and expensive alternative imaging procedures such as PET.

Case of PGE with bitemporal affection with no morphological data supporting mesial sclerosis, but the NAA is decreased in both mediobasal temporal lobes with no noticeable asymmetry.( Fig.1,2,3,4,5)

Fig-1

Fig-2

Fig-3

Fig-4

Fig-5