Research ArticleEpilepsy

Glutamate imaging (GluCEST) lateralizes epileptic foci in nonlesional temporal lobe epilepsy

See allHide authors and affiliations

Science Translational Medicine  14 Oct 2015:
Vol. 7, Issue 309, pp. 309ra161
DOI: 10.1126/scitranslmed.aaa7095

Toward visualizing the focus

Many seizures, especially those that originate in the brain’s temporal lobe, start at a single spot in the brain. If drugs fail, excision of this region can often provide relief from seizures. A new imaging method that harnesses the power of a 7-T magnet shows promise in locating hard-to-find epileptic foci by visualizing the neurotransmitter glutamate.

In a pilot study, the authors used glutamate chemical exchange saturation transfer (GluCEST), a very high resolution magnetic resonance imaging contrast method, to measure how much glutamate was in the hippocampi of four patients with epilepsy. Glutamate is elevated in epileptic foci. The amount of glutamate was clearly higher in one of the hippocampi in all four patients, and confirmatory methods (electroencephalography or magnetic resonance spectra) verified independently that the hippocampus with the elevated glutamate was located in the same hemisphere as the epileptic focus.

Although the authors have only taken a first step toward noninvasively finding epileptic foci, their demonstration that GluCEST can localize small brain hot spots of high glutamate is promising. This approach can potentially allow a higher rate of successful surgeries in this difficult disease.


When neuroimaging reveals a brain lesion, drug-resistant epilepsy patients show better outcomes after resective surgery than do the one-third of drug-resistant epilepsy patients who have normal brain magnetic resonance imaging (MRI). We applied a glutamate imaging method, GluCEST (glutamate chemical exchange saturation transfer), to patients with nonlesional temporal lobe epilepsy based on conventional MRI. GluCEST correctly lateralized the temporal lobe seizure focus on visual and quantitative analyses in all patients. MR spectra, available for a subset of patients and controls, corroborated the GluCEST findings. Hippocampal volumes were not significantly different between hemispheres. GluCEST allowed high-resolution functional imaging of brain glutamate and has potential to identify the epileptic focus in patients previously deemed nonlesional. This method may lead to improved clinical outcomes for temporal lobe epilepsy as well as other localization-related epilepsies.


Epilepsy affects ~65 million people worldwide and is a significant source of neurological morbidity. About one-third of epilepsy patients have seizures that are not controlled by medications (1). Ongoing seizures degrade patients’ quality of life by limiting driving and employment, and by causing social isolation and psychological harm (2). In addition, uncontrolled seizures are associated with 11 times more mortality than expected on the basis of age (3). The estimated cost of epilepsy in the United States is about $10 billion, including medical expenditures and informal care (4).

Localization-related epilepsy (LRE), also termed partial onset epilepsy, is the most common type of epilepsy and is present in 80% of drug-resistant patients (5). In adults, temporal lobe epilepsy (TLE) accounts for 65% of LRE (6, 7). Mesial temporal sclerosis can be identified on structural magnetic resonance imaging (MRI) in about two-thirds of patients with TLE and is associated with the most favorable outcomes from resective epilepsy surgery, with 70 to 80% of patients seizure-free after temporal lobectomy (812). There is a two to three times greater chance of a good postsurgical outcome if an MRI or histopathological lesion is identified (13, 14). The distribution of drug-resistant (refractory) epilepsy patients is illustrated in Fig. 1.

Fig. 1. The distribution of epilepsy patients worldwide, with details for those with drug-resistant epilepsy.

Currently, patients with drug-resistant epilepsy undergo multimodal structural and functional imaging for surgical planning. In addition to conventional 3-T MRI with fine cuts through the mesial temporal structures, this may include 18-fluoro-deoxyglucose positron emission tomography (FDG-PET), single-photon emission computed tomography (SPECT), and magnetoencephalography (MEG). Unfortunately, these methods, even in combination with scalp electroencephalography (EEG), still do not adequately localize the seizure focus in a large percentage of patients. About one-third of patients with TLE show no lesion by conventional MRI (1, 510, 1214). Nevertheless, in the patients in this group who undergo resective surgery, histopathology is abnormal in 87% (15). This statistic suggests that lesions are present, but that current imaging technology lacks the sensitivity to detect them.

These findings highlight the need for improved tools to map epileptic networks. It is widely postulated that many conventional MRI-negative patients have abnormalities that might be identified by advanced imaging techniques (16). Because it is also well established that patients with lesional epilepsy have better surgical outcomes than those with nonlesional epilepsy (14), new neuroimaging techniques capable of detecting subtle lesions could potentially improve patient care and increase the chance of seizure freedom after surgery.

Both human and animal studies suggest that glutamate can serve as a marker of epileptic networks and support the hypothesis that glutamate is elevated in epileptogenic foci (1720). It has been hypothesized that elevated glutamate within the glial-neuronal unit is a key manifestation of both the mitochondrial and metabolic injuries that induce the hyperexcitable state characterizing seizures (21). Microdialysis and pathological studies in human epilepsy reveal increased glutamate ictally (during seizures), interictally (between seizures), and post-mortem in the epileptic focus (18, 22). Results with other imaging markers are concordant with these findings; decreased hippocampal volume on MRI has been associated with increases in extracellular glutamate in drug-resistant TLE patients during intracranial EEG evaluation (23).

Magnetic resonance spectroscopy (MRS) studies in patients designed to measure glutamate have been performed mainly with lower-field MRI magnets, and the results have not been clear. Unlike the prominent resonances of N-acetylaspartate (NAA) and creatine (Cr), which are singlets, glutamate’s resonances are triplets or higher-order multiplets, resulting in smaller spectral peaks distributed over a broader range of frequencies. Glutamate also shows spectral overlap with glutamine, which complicates spectral interpretation at low field strength. In data from a 1.5-T magnet, the combined resonance of glutamate and glutamine (Glx) has been reported to be decreased in EEG-defined neocortical epileptogenic regions (24). Also with data at 1.5 T, it has been reported that the Glx resonance is decreased in sclerotic hippocampi but increased in the epileptic hippocampus in patients with MRI-negative disease (25, 26). In a small study of five patients with hippocampal sclerosis on clinical MRI, MRS at 4 T showed that glutamate levels were decreased in the sclerotic hippocampi (27). Volume loss in sclerotic hippocampi may have confounded this investigation, as MRS generally obtains data from only one relatively large voxel.

There is evidence for dysfunctional glutamate cycling in TLE with both down-regulation of glutamine synthetase (GS) in astrocytes (resulting in slowed glutamate clearance that is most marked in CA1 and CA3 of the hippocampus) and elevated levels of the glutamate-synthesizing enzyme phosphate-activated glutaminase (PAG) in the epileptogenic hippocampus (2831). In another study, GS mRNA was increased by ~50% in CA3 in MRI-negative TLE patients compared to patients with hippocampal sclerosis on MRI and controls. In these patients, PAG mRNA was also increased in CA1, CA2, CA3, and the dentate gyrus in MRI-negative TLE patients compared to TLE patients with sclerosis on MRI (29).

The chemical exchange saturation transfer (CEST) technique measures proton exchange between the exchangeable protons of the solute with the much larger pool of bulk water protons (32, 33). When the magnetization from the exchangeable protons of the solute is saturated with a frequency-selective radio frequency pulse, there is a proportional decrease of the water signal as a result of accumulation of saturated protons in the bulk water pool. The difference between signals obtained with and without saturation of the solute pool is measured as the CEST effect. For glutamate (Glu), the amine proton resonates at 3 ppm (parts per million) down field from water, and the exchange rate is within a slow to intermediate exchange regime, making glutamate an ideal neurotransmitter for CEST imaging with MRI scanners at 7 T and higher fields (34). GluCEST has at least two orders of magnitude higher sensitivity than traditional 1H MRS method for measuring glutamate (34). Indeed, this method images glutamate in vivo at much higher spatial resolution than can be achieved with MRS or spectroscopic imaging (34). GluCEST has been used to study the brain and spinal cord of healthy subjects and Alzheimer’s disease mouse models (3538). In an Alzheimer’s mouse model, glutamate alone contributes >90% of the GluCEST signal with <10% contribution from other metabolites (37, 38).

Here, we applied the MRI method GluCEST (glutamate chemical exchange saturation transfer) at 7 T to patients with nonlesional TLE. We hypothesized that the GluCEST method would be able to localize the hemisphere containing the epileptic network by visualizing increased glutamate in the hippocampus of patients with nonlesional TLE.


Four nonlesional drug-resistant epilepsy patients and 11 healthy controls were included in the analysis (tables S1 and S2). One of the four epilepsy patients subsequently underwent intracranial EEG evaluation and right temporal lobectomy, with pathology consistent with mesial temporal sclerosis. The patient and control hippocampal volume data, GluCEST signal data including clinical data (table S1), both quantitative values and GluCEST maps, and MRS data (tables S2 and S3 and figs. S1 and S2) are available in the Supplementary Materials (tables S1 to S5).

In all four epilepsy patients, concentrations of glutamate measured by GluCEST were higher in the epileptogenic (ipsilateral) hippocampus than in the contralateral hippocampus, both qualitatively and quantitatively. Independent visual analysis of bilateral hippocampal images by two epileptologists (B.L. and J.P.), blinded to patient information, accurately lateralized seizure onset in four of four patients. Figure 2 illustrates the lateralized GluCEST signal in two right-sided temporal epilepsy patients and two left-sided temporal epilepsy patients. Seizure-onset side was determined by the Penn Epilepsy Center surgical conference consensus blinded to GluCEST results. A t test comparing the ipsilateral to contralateral hippocampal GluCEST signal in patients was statistically significant, with the higher GluCEST signal in the hippocampus ipsilateral to the location of seizure onset (one-tail, P = 0.011) (Fig. 3 and Table 1). We also performed a separate analysis on the head and tail subregions of the hippocampus. In the brain slice measured in our experiments, the head of the hippocampus was composed primarily of the CA1 region. In patients, the GluCEST signal in the ipsilateral (to seizure onset) head was significantly different from that of the contralateral head (one-tail, P = 0.03). Similar comparisons of the hippocampal tail, whole hemisphere, and hemisphere not including occipital lobe (consisting largely of temporal lobe and mesial temporal structures) showed no significant differences.

Fig. 2. Axial sections from four patients with drug-resistant TLE, showing the GluCEST signal.

(A) A 40-year-old female with nonlesional right TLE, with a visible increase in the GluCEST signal in the right hippocampus. (B) A 47-year-old female with nonlesional right TLE, with a visible increase in the GluCEST signal in the right hippocampus. (C) A 25-year-old female with nonlesional left TLE, with a visible increase in the GluCEST signal in the left hippocampus. (D) A 47-year-old male with nonlesional left TLE, with a visible increase in the GluCEST signal in the left hippocampus.

Fig. 3. Increased glutamate in the hippocampus ipsilateral to seizure onset, as measured by GluCEST.

GluCEST contrast in the hippocampi ipsilateral and contralateral to seizure onset, as measured by percentage displaced water protons. Green and gray portions of each box represent the second and third quartile values, respectively; the upper and lower range of values is indicated by whiskers. P = 0.011, one-tailed t test.

Table 1. Summary of GluCEST findings in nonlesional patients with TLE.

Comparisons are made between regions ipsilateral and contralateral to seizure onset.

View this table:

We compared the mean absolute asymmetry in the hippocampus of 11 controls to that of the 4 patients. The confidence interval (GluCEST contrast, %) was −0.42 to 0.73 for the controls, and the mean absolute asymmetry in the patients was 0.16 higher than in the controls. A t test comparing the right to left hippocampal GluCEST signal in controls did not reach statistical significance (two-tail, P = 0.268).

There was no significant difference in hippocampal volume between the ipsilateral and contralateral hippocampal slices in patients (P = 0.478, assuming equal variances for two-tailed test), or between the volume of the right and left hippocampi in the control group (P = 0.302, assuming equal variances for two-tailed test). Data are in the Supplementary Materials (table S3).

We also performed MRS of the bilateral hippocampi in all patients and controls. However, because motion and susceptibility artifacts significantly degrade MRS measurement in this region, interpretable MRS results were only obtainable in the bilateral hippocampi of one patient (patient 4, who underwent intracranial EEG monitoring and right temporal lobectomy) (table S3). Results were unreliable in one of the hippocampi of the other three patients. In controls, bilateral MRS was available in five subjects. In the patient with bilateral MRS, glutamate was increased in the hippocampus ipsilateral to seizure onset (Fig. 4) (15.77 mM ipsilateral versus 12.26 mM contralateral). NAA/Cr ratios were also increased in the hippocampus ipsilateral to seizure onset (NAA/Cr, 1.39) versus the contralateral hippocampus (NAA/Cr, 1.28). In the control subjects, four of the five with available bilateral MRS showed subtle glutamate asymmetries that corresponded to the lateralization of the GluCEST signal. Data are included in the Supplementary Materials (tables S2 and S3 and figs. S1 and S2).

Fig. 4. Increased glutamate in the hippocampus ipsilateral to seizure onset in a patient with right TLE.

Single-voxel proton magnetic resonance spectroscopy (1H MRS) was performed individually on the left and right hippocampi of the nonlesional right TLE subject (Cho, choline; mI, myo-inositol). Heights of the peaks were measured in arbitrary units (A.U.).

One subject included in this study underwent intracranial EEG evaluation to lateralize seizure onset. Symmetric bilateral hippocampal and amygdalae depth electrodes (4) and lateral temporal and subtemporal subdural strip (12) electrodes were placed. A total of four clinical electrographic seizures were captured, all arising from the right hippocampus. On the basis of these intracranial EEG findings, the patient underwent right temporal lobectomy. An example of a seizure arising from the right hippocampal depth electrode is shown in Fig. 5. Coregistration of images was performed as published previously (Fig. 5, A to G) (39). Intracranial EEG recordings of seizure onset in the right hippocampus were congruent with GluCEST findings, with increased GluCEST in the right hippocampus in this subject. Hippocampal pathology as reported by the University of Pennsylvania pathology analysis of subject 4 showed dispersion of granule cells, endplate gliosis, and mild loss of pyramidal neurons consistent with mesial temporal sclerosis.

Fig. 5. Intracranial EEG data from patient 4, with nonlesional TLE arising from right hippocampus.

Background traces show EEG data from the individual contacts of subdural and depth electrodes in patient 4, indicating the localization of seizure onset to the CA1 region of the hippocampus (red arrow). (A, C, and E) Three MRI sections through the hippocampus and amygdala showing the multicontact right mesial temporal depth electrode [each contact of the electrode is shown as a yellow square; seizure onset electrode is the electrode most distal from scalp (first electrode from left)]. (B, D, and F) Duplicate of the sections shown in (A), (C), and (E) with segmentation of hippocampal subfields superimposed on the image. Red, CA1; green, dentate gyrus; dark blue, Brodmann area 36; tan, CA3; light blue, Brodmann area 35; violet, subiculum. Segmentation was performed as described (47). (G) Right lateral view of the reconstructed MRI brain image of patient 4 (39). The positions of the right hemisphere electrodes are shown in green, coregistered with the MRI image. The multicontact electrodes used for seizure localization are labeled [Depths, depth electrodes (right hippocampal and amygdalar); RPT, right posterior temporal electrode; RMT, right mid-temporal electrode; RAT, right anterior temporal electrode; RAF, right anterior frontal electrode]. The unlabeled electrodes were not used in the analysis described in this article.


We present a technique that has the potential to identify seizure foci in epilepsy patients who were previously determined “nonlesional” on the basis of currently available imaging methods. Although our findings were observed in only a small group of four patients, we present them now for three reasons: (i) our findings are compelling despite being from an unselected group of patients; (ii) the ability to detect seizure foci with GluCEST imaging could potentially improve patient care and quality of life in epilepsy patients, and sharing these results may expedite validation of this technique; and (iii) the results are consistent with the presence of interictal increases of glutamate in seizure foci. We detected by visual inspection a larger amount of hippocampal glutamate by GluCEST in the hippocampus from the hemisphere in which seizures initiated in all patients, with 100% interrater reliability. In one patient from whom we could obtain data, glutamate concentration as measured by hippocampal MRS correlated with GluCEST quantification of glutamate in patients and controls, lending further confidence in the GluCEST technique.

Histopathological lesions are present in 87% of nonlesional TLE patients by examination of resected tissue (15). Our work indicates that GluCEST can detect asymmetric hippocampal glutamate levels in these patients. If validated in a larger population of epilepsy patients, GluCEST imaging could reduce the need for invasive intracranial monitoring, which is associated with morbidity, mortality, and expense. In addition, GluCEST measures may yield prognostic information that can help physicians and patients determine the best form of treatment. A number of new surgical and medical options are now available to epilepsy patients including laser ablation therapy, NeuroPace Responsive Neurostimulator System, and closed-loop vagal nerve stimulation (40).

The GluCEST technique has multiple advantages over MRS, the only other imaging modality available to measure brain glutamate noninvasively in humans. GluCEST has higher spatial resolution, potentially allowing for more precise visualization of the functional excitatory network, and thus shows promise to further elucidate the etiology and progression of the epilepsy disease state. MRS acquisition is time-intensive, and only a single voxel at a time is typically acquired (15 to 20 min per voxel in this study with time for shimming). In addition, the rectangular voxels used to measure the hippocampal region in MRS likely capture regions outside the hippocampus and vary significantly from subject to subject. The limited MRS data presented here represent the challenges with MRS. MRS required extensive shimming at 7 T and was exquisitely sensitive to movement artifacts. Finally, the single-voxel acquisition limited spatial resolution and resulted in partial volume effects with contamination of surrounding structures. Furthermore, GluCEST has higher spatial resolution than PET, which has been used to measure glutamate receptors only in healthy controls (41).

These findings are consistent with results from animal models that indicate that glutamate synthetase dysfunction leads to accumulation of excessive glutamate in epilepsy (2831). In addition, our findings that glutamate is elevated in the hippocampal head, largely composed of the CA1 region of the hippocampus, but not the tail region is supported by the literature (2831).

There are multiple limitations to this study, most notably the small sample size, the lack of postsurgical validation of seizure lateralization in all but one patient, and the use of single slice imaging. Substantial variations in GluCEST values observed in control subjects, which appeared to randomly lateralize, may have resulted from differences in slice location in each hemisphere. The single-slice method also prohibited measurement of the entire epileptic network. In addition, 7-T MRI is not uniformly available, although most epilepsy surgery is performed at academic centers, many of which now have 7-T MRI. This should facilitate multisite validation of GluCEST as an imaging biomarker of seizure foci. The ability to accurately localize seizure foci would also provide a strong clinical motivation for the deployment of additional 7-T MRI systems or for patients to travel to 7-T sites for the procedure. Our study demonstrated the feasibility of lateralizing seizure foci in TLE, but GluCEST could also potentially localize drug-resistant neocortical epilepsy, where current epilepsy imaging techniques are even more limited.



All studies were conducted under an approved Institutional Review Board protocol of the University of Pennsylvania. GluCEST MRI was acquired on 7.0-T whole-body MRI scanner (Siemens Medical Systems) with a 32-channel phased-array head coil (Nova Medical Inc.) in 11 healthy control subjects (3 male, aged 23 to 54 years; 8 female, aged 24 to 56 years; mean age, 35 years) and 4 TLE subjects (1 male, aged 47 years; 3 female, aged 25 to 47 years; mean age, 40 years).

The four TLE subjects were recruited from the Penn Epilepsy Center and had undergone presurgical evaluation including prior scalp EEG and, in one case, intracranial EEG capturing seizures. Because this was initially a pilot study, no predetermined sample size was calculated, and data collection is ongoing. Inclusion criteria included epilepsy classification of mesial temporal epilepsy as determined by the Penn Epilepsy Center conference, at least 18 years of age, and unremarkable clinical MRI as interpreted by a clinical radiologist at the University of Pennsylvania. Exclusion criteria were contraindications to 7-T MRI scanning (for example, metallic implant), prior intracranial surgical intervention, claustrophobia prohibiting scanning at 7-T MRI without sedative medications, and pregnancy. Imaging for GluCEST was done for each subject at one time point. The investigator (R.P.R.N.) analyzing the GluCEST data was blinded to the lateralization of seizure onset in each subject. Table 2 summarizes the clinical data for the four epilepsy patients recruited to participate. None of the patients had an identifiable epilepsy risk factor (for example, no prior history of febrile seizures, no family history of seizures, and no history of head trauma). All patients were resistant to multiple antiepileptic medications, having tried and failed between two and seven prior antiepileptic medications (average of five prior drug trials). Patient 3 declined evaluation for surgery. Patient 4 underwent a right temporal lobectomy after intracranial EEG monitoring confirmed that seizures arose from the right temporal region. Patients 1 and 2 are undergoing presurgical evaluation.

Table 2. Epilepsy patient clinical information.

SPS, simple partial seizure; CPS, complex partial seizures; GTC, generalized tonic clonic seizure.

View this table:

MRI scans and data analysis

The MRI protocol consisted of the following steps: a localizer, T1-weighted anatomical three-dimensional (3D) magnetization-prepared rapid gradient echo (MPRAGE) images of whole brain [176 axial slices, repetition time/echo time/inversion time (TR/TE/TI) = 2800/4.4/1500 ms, α = 7°, 0.8 × 0.8 × 0.8 mm3 resolution, iPAT = 2, scan time 4:43], T2-weighted 3D using the variable flip angle turbo spin echo (TSE_VFL) sequence (224 coronal slices, TR/TE = 3000/388 ms, 0.4 × 0.4 × 1.0 mm3 resolution, iPAT = 2, scan time 7:42) followed by the acquisition of the GluCEST sequences (34, 37, 38, 42). The T1 images were used to identify the location of the hippocampi and then to manually position the T2 scan such that the direction of the imaged slice was parallel to the long axes of the hippocampi. The TR for the T2 scan was shortened from its optimal setting for T2 contrast to bring the scan time to a manageable duration. For GluCEST, anatomical images acquired from T2-weighted imaging were used to select the axial hippocampal slice. The GluCEST imaging parameters were as follows: slice thickness = 5 mm, field of view (readout) = 200 mm, field of view (phase encoding) = 162.5 mm, matrix size = 256 × 208, gradient recalled echo (GRE) readout TR = 6.2 ms, TE = 3 ms, number of averages = 2, shot TR = 10,000 ms, shots per slice = 2, with an 800-ms-long saturation pulse train consisting of a series of 96-ms Hanning windowed saturation pulses with a 4-ms interpulse delay (100-ms pulse train) at a B1rms of 3.06 μT. Raw CEST images were acquired at varying saturation offset frequencies from ±1.8 to ±4.2 ppm (relative to water resonance) with a step size of ±0.3 ppm. GRE images at two echo times (TE1 = 4.24 ms; TE2 = 5.26 ms) were collected to compute B0 map. B1 map was generated from the two images obtained using square preparation pulses with flip angles 30 and 60°. All of the image processing and data analysis were performed with in-house written programs in MATLAB (MathWorks, version 7.5, R2009b). Overall, acquisition time of CEST images, B1 and B0 field maps, was about 20 min. CEST images obtained from ±1.8 to ±4.2 ppm were interpolated using the cubic spline method to generate images with a fine step size of 0.01 ppm. B0-corrected CEST images at ±3 ppm were generated from the interpolated CEST images by picking signals according to the frequency shift in the B0 map. The B0-corrected ±3.0 ppm images were then used for computing the percentage GluCEST contrast, which is equal to 100 × (M−3ppm − M+3ppm)/M−3ppm, where M−3ppm and M+3ppm are B0-corrected images saturated at −3 ppm and +3 ppm, respectively, with respect to water (34). B1 inhomogeneity artifacts in GluCEST maps were removed using B1 calibration curves as reported (42). The B0- and B1-corrected GluCEST contrasts were then averaged within expertly drawn (performed by R.P.R.N. and confirmed by K.A.D.) regions of interest (ROIs) in the bilateral hippocampi, hippocampal head (largely composed of CA1), hippocampal tail, right hemisphere, left hemisphere, right hemisphere excluding occipital lobe (largely composed of right temporal lobe and mesial temporal structures), and left hemisphere excluding occipital lobe (largely composed of left temporal lobe and mesial temporal structures).

Hippocampal volumes were measured from a slice corresponding to the 2D CEST slice. Hippocampal volume of each ROI = [TP − (NE/2)] × VP [TP, total number of pixels within the ROI; NE, number of edge pixels within the ROI; VP, volume of each pixel in the CEST 2D slice = 3.2 mm3 (0.8 mm × 0.8 mm × 5 mm)].

For single-voxel proton magnetic resonance spectroscopy (1H MRS), the voxel of interest [AP (anterior-posterior), 20 mm; LR (left-right), 10 mm; HF (head-foot), 5 mm; right/left hippocampus] was positioned on the GluCEST axial hippocampal slice. Automated first- and second-order shimming of the B0 field was performed on voxel of interest to obtain a localized water linewidth of ~24 Hz or less using FASTMAP shim method (43, 44) provided by Siemens as work-in-progress package. Single-voxel spectra for glutamate were obtained using PRESS (point resolved spectroscopy) sequence with the following parameters: number of points, 2048; averages, 8 (water reference spectrum)/128 (water suppressed spectrum); TR, 3000 ms; and TE, 20 ms. The total acquisition time to obtain each spectrum was ~7 min 12 s. For post-processing of spectroscopy data, we used the raw multichannel time domain data from the scanner. From the water reference data, channel-wise time-dependent phase shifts due to eddy current and amplitude scale factors were obtained and saved. Both spectra were obtained after channel-wise eddy current correction and adaptive combination (45). Metabolite peaks from water-suppressed spectrum were fitted as Lorentzian functions with nonlinear least-squares fitting (MATLAB “nlinfit” routine) by taking into account the prior knowledge of the 8 macromolecular peaks and 14 metabolite peaks over the frequency range of 0.5 to 4.3 ppm (46) followed by integration, and then normalized by water reference signal for absolute quantification of glutamate.


Paired two-sample t test for means was performed on the control and epilepsy subjects (two-tailed and one-tailed, respectively). GluCEST measurements were calculated for bilateral hippocampi, bilateral hippocampal heads, bilateral hippocampal tails, bilateral hemispheres, and bilateral hemispheres excluding the occipital lobe (n = 4 epilepsy patients, n = 11 control subjects). Covariates such as age, sex, seizure frequency, and concurrent antiepileptic drugs were not included because of low sample size. Hippocampal volume was calculated as described in Materials and Methods, and asymmetry was assessed using a two-sample t test assuming equal variances.


Table S1. Epileptic subjects’ demographics and clinical information.

Table S2. Control subjects’ demographics and summary of data.

Table S3. Hippocampal volume, GluCEST, and MRS data in epilepsy patients.

Fig. S1. GluCEST and MRS data in control subjects.

Fig. S2. GluCEST and MRS data in epilepsy subjects.

Data S1. Fig. S1 High-resolution controls (Excel file).

Data S2. Fig. S2 High-resolution patients (Excel file).


  1. Acknowledgments: We thank J. Stein for assistance with the electrode coregistration (Fig. 5). Funding: P41 EB015893 [NIH NIBIB (National Institute of Biomedical Imaging and Bioengineering)], P20 NS12006 [NIH NINDS (National Institute of Neurological Disorders and Stroke)], McCabe Pilot Award (University of Pennsylvania), and Center for Biomedical Image Computing and Analytics Seed Award (University of Pennsylvania). Author contributions: K.A.D. and R.P.R.N. contributed equally to this work. K.A.D. developed the clinical protocol, recruited human subjects, and wrote the paper with R.P.R.N. R.P.R.N. assisted with data acquisition and analysis of GluCEST and MRS, in addition to writing the paper with K.A.D. S.D. assisted with hippocampal segmentation. S.H.C. and P.N.H. assisted with subject recruitment and clinical data collection. J.R.P. and B.L. reviewed the GluCEST images in addition to contributing to conceptual development. T.H.L. provided neurosurgical expertise. R.T.S. provided assistance with statistical analysis. H.H. contributed to development of GluCEST technique. M.A.E. assisted with image sequence development. J.A.D. and R.R. contributed to development of GluCEST and conceptual development. Competing interests: R.R. and H.H. hold the patent (U.S. 20120019245 A1) on CEST MRI methods for imaging metabolites and the use of these as biomarkers. Data and materials availability: All data are available in the Supplementary Materials.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article