Hippocampal Region Asymmetry Assessed by 1H-MRS in Rolandic Epilepsy


Address correspondence and reprint requests to Dr. S. Lundberg at University Children's Hospital, S-751 85 Uppsala, Sweden. E-mail: staffan.lundberg@kbh.uu.se


Summary:  Purpose: In a previous study, we reported hippocampal abnormalities on magnetic resonance imaging (MRI) in six of 18 children with rolandic epilepsy (RE). In this study, metabolic changes were analyzed in the hippocampal region with proton magnetic resonance spectroscopy (1H-MRS).

Methods: In 13 children with electroclinically typical RE and 15 healthy controls, 1H-MRS results of both hippocampal regions were analyzed. The voxels, 2 × 2 × 4-cm each, were placed to include the head and body of the hippocampus. A PRESS sequence with TR 2,000 ms and TE 32 ms was used. Total N-acetylaspartate (tNAA), glutamine and glutamate (Glx), and choline compounds (tCho) were related to total creatine (tCr), and asymmetry indices (AIs) were calculated. MRI was performed in all 13 patients and in 13 controls.

Results: The tNAA/tCr AI of the hippocampal region was significantly higher in children with RE than in control children (z = 4.49; p < 0.001). The AIs of Glx/tCr and tCho/tCr did not show a significant difference between the groups. Lateralization of the interictal epileptiform activity corresponded with the lower tNAA/tCr ratio in 10 of 13 patients. MRI revealed a hippocampal asymmetry in four of 13 in the RE group, three of them showed concordance between the lateralization of the lower tNAA/tCr ratio and the smaller hippocampus. In the control group, a subtle asymmetry in four of 13 children was found.

Conclusions: A significant asymmetry of the hippocampal regions, measured by tNAA/tCr ratios, indicates an abnormal neuronal function in children with RE.

Rolandic epilepsy (RE) is one of the most common epilepsy syndromes in childhood. According to the International League Against Epilepsy (ILAE) classification of 1989 (1), it is an idiopathic, localization- and age-related syndrome with typical semiology and characteristic EEG features occurring in otherwise neurologically normal children, with an onset between ages 3 and 13 years. The syndrome has been subjected to many different synonyms (2) including benign childhood epilepsy with centrotemporal spikes, which was the name proposed in the ILAE classification. The syndrome is apparently benign but only with respect to outcome of seizures (i.e., the child will recover before puberty). In other aspects, a generally favorable prognosis has been questioned, based mainly on a risk for neuropsychological deficits (3–8). With this background, the term benign should be avoided; RE is preferred.

Over the last years, proton magnetic resonance spectroscopy (1H-MRS) has become an important noninvasive technique for exploring the human brain in vivo. One of several clinical applications for 1H-MRS is epilepsy, and 1H-MRS is considered as complementary tool to other neuroimaging techniques, electrographic methods, and clinical observations, to localize or lateralize the epileptogenic zone. Most studies including the assessment of the hippocampus concern adults with temporal lobe epilepsy (TLE). Some of these reports also include a small number of children (9–12). A few studies consist of only children with focal epilepsy, mainly TLE, and control children (13,14), or just children with epilepsy (15,16). Children with TLE also have been compared with adult controls (17–19). 1H-MRS studies have, to our knowledge, not been performed in idiopathic focal epilepsy.

To elucidate whether there are morphologic CNS abnormalities in RE, we earlier investigated 18 children with MRI (20). By using visual evaluation, hippocampal abnormalities were found in six children with asymmetries in size, and/or signal intensities on T2-weighted images.

The aim of this study was to investigate the hippocampal region in children with RE by using 1H-MRS to ascertain whether there are any metabolic changes in comparison with healthy controls.


Patient group

Seventeen children with electroclinically typical RE were recruited from our epilepsy clinic. There were 11 girls and six boys with a median age of 10.9 years (mean age, 11.3 years; range, 8.2–13.6 years). All children were neurologically normal, and all were right-handed. One girl (case 9) had had one typical febrile seizure episode in infancy, but otherwise neither febrile seizures nor status epilepticus had been reported.

EEG was performed by using the standard 10-20 International System placement of scalp electrodes. The interval between the 1H-MRS and EEG varied from 4 days to 3 years (median, 12 months). In all subjects, magnetic resonance imaging (MRI) was performed, and the interval between this and the 1H-MRS was from 2 weeks to 3 years 10 months (median, 5 months). Eleven children had been included in the previous MRI study (20). Only those with high spectral quality were evaluated (see Results).

Control group

The control group consisted of 22 children, 11 girls and 11 boys, with a median age of 11.6 years (mean age, 11.5 years; range, 8.1–14.8 years). Twelve of them had participated as controls in a neuropsychological study on children with RE (3), in which study, they had been matched for age, sex, and school performance. The additional 10 children had been matched for age and sex. All were healthy and with normal psychomotor development, and there was no history of febrile or any other seizures. MRI was performed in 19 children, as three were reluctant to participate in two examinations. The interval between the 1H-MRS and MRI varied from 2 weeks to 10 months (median, 5 months).

Only those with high spectral quality were evaluated (see Results). Written informed consent for participation was obtained from all control children and their parents.

Magnetic resonance spectroscopy

The study was performed with a 1.5-T clinical MR system (Gyroscan NT; Philips, Best, The Netherlands) by using the standard quadrature transmitter–receiver birdcage head coil. No individual tuning and matching was performed. T1-weighted transverse, sagittal, and coronal spin-echo images (repetition time/echo time = TR/TE = 500/20 ms) were used to position the volume of interest (VOI). We used a single-voxel technique. Typical position is shown in Fig. 1. The VOIs were individually adjusted to cover most of the hippocampus and to minimize partial volume effects with other tissues. The size and shape of the VOI was a rectangular box of dimensions 40 × 20 × 20-mm3 (16 ml). Special care was taken to avoid susceptibility effects from the skull base. The resulting VOI consisted of the hippocampal head, body, a small part of the tail, some of the perihippocampal brain tissue, and cerebrospinal fluid (CSF). The content of the VOI is henceforth called the hippocampal region. Proton spectra of both hippocampal areas were recorded with use of a PRESS sequence (TR/TE = 2,000/32 ms; 256 accumulations; 1,024 points; spectral bandwidth, 1,000 Hz), which also included unsuppressed water reference scans for neurochemical quantitations. The acquisition time for each hippocampus was 8 min 40 s. The total time required for the 1H-MRS measurements was ∼45 min. No sedation or general anesthesia was used.

Figure 1.

T1-weighted magnetic resonance images showing a placement of a volume of interest (40 × 20 × 20 mm3 = 16 ml) selected for 1H-magnetic resonance spectroscopy of the hippocampus (spin echo, TR/TE = 500/20 ms).

Data processing

Automated quantification of absolute metabolite concentrations was accomplished by LCModel (linear combination of model spectra of metabolite solutions in vitro) (21), a user-independent frequency domain-fitting program. In vivo spectra were corrected for residual eddy-current effects and analyzed by a superposition of a set of in vitro basis spectra by means of a constrained regularization algorithm finding the best compromise between lineshape and baseline consistent with the data. The method uses the experimentally determined spectral pattern of each metabolite without further analysis. Concentrations are calculated as millimolars (i.e., mmol/L VOI), and are not corrected for residual T1 or T2 relaxation effects and CSF contributions. With respect to the concentration correction, we note that the TE and TR influence the LCModels results. However, the ratio tNAA/tCr remains constant, reflecting the similar relaxation times of creatine (Cr) and N-acetyl aspartate (NAA) according to Pfeuffer et al. (22) and Choi and Frahm (23). Because of limited spectral resolution, N-acetyl compounds are given as total N-acetylaspartate (tNAA) (i.e., the sum of N-acetylaspartylglutamate and NAA). Accordingly, total creatine (tCr) represents both Cr and phosphocreatine. The concentration ratio tNAA/tCr (R) was calculated for the left (Rl) and right (Rr) hippocampal region. The glutamine/glutamate complex (Glx) and choline-containing compounds (tCho) in relation to tCr were calculated in the same way. For the assessment of asymmetries in the metabolism of tNAA, Glx, and tCho, respectively, we used an asymmetry index (AI):


The physicist performing the analyses did not know whether the children were patients or controls.

Magnetic resonance imaging

A 1.5-Tesla imager was used. MRI included sagittal T1-weighted spin-echo images, and axial T2-weighted images in the plane of the long axis of the hippocampus. Oblique coronal images perpendicular to the plane of the long axis of the hippocampus with T2-weighted spin-echo and T1-weighted inversion recovery images were obtained with slice thickness of 2.5 mm and with no gaps between the slices.

Two blinded observers evaluated the images visually, and in case of deviating interpretations, a consensus was used in a final analysis.

Statistical analysis

A Mann–Whitney U test was applied to analyze statistical differences between controls and patients, with a significance level of 0.05.

The Ethics Committee of the Medical Faculty of Uppsala University approved the study.


Only the subjects with a high spectral quality on both sides were selected for the final assessment. The major reason for exclusion of a subject was movement artifacts due to inability to keep still during the time required for the examination.

In the control group, 15 of 22 subjects showed a high spectral quality bilaterally. There were eight girls and seven boys with a median age of 11.1 years (mean age, 11.1 years; range, 8.1–14.1 years).

In the RE group, 13 of 17 subjects showed a high spectral quality bilaterally. There were eight girls and five boys with a median age of 12.4 years (mean age, 11.7 years; range, 8.2–13.6 years). Six of the children were treated with antiepileptic drugs [AEDs; valproate (VPA), carbamazepine (CBZ), or sulthiame (STM)]; four children had been without treatment for ≥1 year, and three children had never been treated. Three children had inactive epilepsy (i.e., they had been seizure free for a little more than 5 years (cases 6, 10, and 12). In the other patients, the interval between the 1H-MRS and the latest seizure episode was ≥1 month.

Table 1 summarizes clinical data and the results of EEG, MRI, and 1H-MRS in the 13 children with RE as well as the 1H-MRS results of the 15 controls. As AIs were not age dependent in the controls, they are presented as one entity. The 1H-MRS results of the tNAA/tCr ratio AI also are demonstrated in a box-plot and whisker diagram (Fig. 2).

Table 1. 1H-MRS of the hippocampal region in 13 children with RE
Age at
predom. side
  • F, female; M, male; R, right; L, left; ?, uncertain lateralization; HC, hippocampal region; AI, asymmetry index; tNAA, N-acetylasparate and N-acetylaspartyl glutamate; tCr, creatine and phosphocreatine; Glx, glutamine/glutamate; tCho, choline compounds.

  • The AIs >95th percentile of those in controls are indicated in bold.

  • a

     In children with inactive RE, the EEG findings refer to the active phase.

1/F8.26LRL < R1.622.010.222.922.730.070.290.310.056
3/F9.88?R > LL < R1.551.390.111.554.330.940.230.260.14
4/F10.27RL > Rno1.261.620.251.411.500.
6/F10.95RL > RaL < R1.251.460.161.471.490.
7/M12.411RL > Rno1.571.270.211.911.580.190.190.340.58
9/F13.15LR > Lno1.881.180.462.481.050.810.330.160.67
12/M13.69LL > Rano1.341.600.181.740.700.850.280.280.011
13/F13.67RL > RL < R1.661.990.181.772.
Median12.47   1.571.500.221.641.580.
Controls(n = 15)             
Median11.1    1.431.440.0451.381.
95th percentile       0.086  0.71  0.39
Figure 2.

Asymmetry index of tNAA/tCr in the hippocampal region in children with rolandic epilepsy and normal control subjects. The ends of the boxes fall at the upper and lower quartiles; the solid line in the middle of the boxes is the median; and the whiskers represent the range of the data.


No age dependence of the AIs was found in the RE group, so they are handled as one entity in the statistics calculations. The AI of the tNAA/tCr ratios was significantly higher in the patients than in the controls (z = 4.49; p < 0.001). In seven patients, the tNAA/tCr ratio was lower in the left, and in six, in the right hippocampal region. The AI of the Glx/tCr ratios in the patients was not significantly different compared with controls (z = 1.54; p = 0.123). In five patients, the AIs of Glx/tCr were above the 95th percentile of those of the controls. There was no significant correlation between the lateralization of the tNAA/tCr and Glx/tCr ratios. The AI of tCho/tCr ratios was similar in both groups.

Of the three children with inactive epilepsy at the time of the 1H-MRS (cases 6, 10, and 12), two showed a high AI of Glx/tCr (cases 10 and 12), whereas the AIs of tNAA/tCr were close to the average of the RE group.

Representative spectra of a normal and an abnormal hippocampal region are shown in Fig. 3.

Figure 3.

LCModel analysis of a 1H-MRS spectra of a child with rolandic epilepsy (case 5). The dotted curves display the spectra as measured. The thick solid curves show the fitted spectra. LCModel spline baseline was subtracted from all spectra. The left hippocampal region (a) shows a normal spectrum (tNAA/tCr, 2.06) and the right (b) an abnormal spectrum (tNAA/tCr, 1.40). NAA = N-acetylaspartate and; Cr = creatine; Cho = choline; Ins = myo-inositol; ppm = parts per million.

MRI versus 1H-MRS

MRI was performed in 13 of 15 controls, and a subtle hippocampal asymmetry was revealed in four. In the RE group, a distinct asymmetry was observed in four of 13 (cases 1, 3, 6, and 13), the left hippocampus being smaller than the right. There was concordance between the lateralization of the smaller hippocampus and a lower tNAA/tCr ratio in three of them (cases 1, 6, and 13). There were no high signal intensities in hippocampus or elsewhere in these four examinations.

EEG versus 1H-MRS

Lateralization of interictal epileptiform activity seen on EEG corresponded to the lateralization of the lower tNAA/tCr ratio in 10 (77%) of the 13 examinations. In three of the five patients with AI of Glx/tCr above the 95th percentile of those of the controls (cases 8, 9, and 10), lateralization of the low Glx/tCr ratio corresponded to EEG lateralization. The remaining two children had bilateral epileptiform activity but predominance on the contralateral side (cases 3 and 12).


This study demonstrates a significantly higher asymmetry of the tNAA/tCr ratios of the hippocampal region in children with RE compared with controls, whereas the Glx/tCr ratios did not show any significant difference between the groups. The NAA/Cr or NAA/(Cr + Cho) ratios are used mainly as indicators of neuronal function in 1H-MRS studies, and a few studies have used Glx/Cr ratios in addition for the same purpose (15,24,25).

NAA is confined to neurons and neuronal processes both in gray and white matter, and a decreased amount indicates loss of neurons or a neuronal dysfunction (26,27). Kuzniecky et al. (28), in a study including MRI volumetry, 1H-MRS, and histopathology in adults with TLE, found that the neuronal/glial ratio in the resected hippocampus correlated with hippocampal volumetry but did not correlate with the metabolic measurement obtained with 1H-MRS. The conclusion was that the metabolic events measured by 1H-MRS reflect neuronal and glial dysfunction rather than neuronal cell loss. In our group of children with RE, four had some degree of asymmetry of the hippocampus size. However, 1H-MRS showed a significant asymmetry in all 13 patients, which may indicate an abnormal neuronal function in the hippocampal region.

Glutamate is the main excitatory neurotransmitter in the brain. Ictal studies have shown increased Glx concentrations ipsilateral to the epileptic focus (15). Interictally, both increased (25) and decreased Glx concentrations (15) have been reported. In our study, divergent results were found. The usefulness of analyzing absolute or relative concentrations of Glx to establish the epileptogenic zone in individuals with epilepsy is still to be evaluated.

Several factors on 1H-MRS such as technique, acquisition parameters and positioning of the VOI, and qualitative versus quantitative measurements may vary. We used a rectangular voxel with a constant volume of 16 ml, adjusted for covering the whole hippocampus. The large voxel size gives a better signal-to-noise ratio, but conversely, a large voxel can overlook minor local abnormalities. Some perihippocampal tissues and CSF will be encompassed and may influence the spectral profile. However, by using ratios of the metabolites and calculating AIs, their influence on the results can be regarded as minimal. One of the advantages with single-voxel spectroscopy in contrast to multivoxel spectroscopy (chemical-shift imaging) is that it allows shorter echo times and thereby possibilities to detect more metabolites (i.e., Glx). The long acquisition time raises high demands of immobilization of the subjects to acquire high-quality spectra. In our study, 11 (28%) of 39 children examined were excluded mainly because of movement artifacts.

These varying factors emphasize the importance of having a proper control material in 1H-MRS studies, but this goal is not always fulfilled. Aging also causes metabolic alterations and must be taken into consideration (29–32). Van der Knaap et al. (29) performed 1H-MRS on 41 healthy children, aged 1 month to 16 years, with examination of predominantly cerebral white matter. They found the most rapid change of NAA/Cr during the first 3 years of life, but minor or no metabolic changes within the same age group as in our study (i.e., 8–14 years). We have not found any 1H-MRS studies of the hippocampal region related to normal aging during childhood. In our study, we could not demonstrate any age dependency. Concerning the three children with inactive epilepsy, their AIs of tNAA/tCr did not differ significantly from those of the other children. They were still quite young (10.9–13.6 years), again emphasizing that there is no age dependency within our group. A follow-up study after some years would elucidate this issue.

The typical epileptiform EEG focus in RE is centrally localized (i.e., in the centrotemporal and centroparietal but also midtemporal leads). The discharges can be unilateral or bilateral. In the present study, the lateralization of the interictal epileptiform activity was concordant to the lower tNAA/tCr ratio of the hippocampal region in 10 (77%) of 13 subjects. This finding may indicate projections between a unilateral focus and the ipsilateral hippocampal region. However, the organization of potential pathways between the rolandic area and this region is still not mapped. We know that hippocampus is an essential structure for processing and organizing learning and memory functions; intricate and complex neuronal pathways lie within the hippocampus and the parahippocampal region, which are interconnected with cortical structures (33,34). As 1H-MRS reflects metabolic function, our finding of an asymmetry of tNAA/tCr ratios of the hippocampal region may indicate some metabolic, probably genetic, imbalance. This might explain the occurrence of at least some cognitive deficits in children with RE. With respect to cognition, all patients in the present study underwent neuropsychological evaluation during the active phase of their epilepsy (3; unpublished data), and seven (54%) had some degree of verbal shortcoming. The current explanation of this cognitive deficit is that the centrotemporal interictal epileptiform activity in RE exerts a direct, mainly transient, influence (6,35). However, a hypothesis is that the hippocampal imbalance may explain both a cognitive dysfunction and an effect on cortical structures. As normalization of cognition in RE usually occurs, even if some dysfunction may remain, maturational factors could have an impact on the pathophysiologic course. Further longitudinal studies including MRI, 1H-MRS, neurophysiology, and measurements of cognitive functions are necessary to understand the development of RE and the intracerebral projections regarding this syndrome.

Acknowledgment: This study was supported by grants from Erik Lysholm Memorial Foundation, Föreningen Margarethahemmet, Family Norin Epilepsy Research Foundation, Glaxo-Smith-Kline, and Swedish Medical Research Council K2001-73X-13158-03A.