Anterior to posterior hippocampal MRS metabolite difference is mainly a partial volume effect

Transcription

1 Acta Radiologica ISSN: (Print) (Online) Journal homepage: Anterior to posterior hippocampal MRS metabolite difference is mainly a partial volume effect Göran Starck, Barbro Vkhoff-Baaz, Maria Ljungberg, Kristina Malmgren, Eva Forssell-Aronsson, Lars Jönsson & Sven Ekholm To cite this article: Göran Starck, Barbro Vkhoff-Baaz, Maria Ljungberg, Kristina Malmgren, Eva Forssell-Aronsson, Lars Jönsson & Sven Ekholm (2010) Anterior to posterior hippocampal MRS metabolite difference is mainly a partial volume effect, Acta Radiologica, 51:3, To link to this article: Published online: 04 Mar Submit your article to this journal Article views: 89 View related articles Full Terms & Conditions of access and use can be found at

2 ORIGINAL ARTICLE Acta Radiologica Anterior to posterior hippocampal MRS metabolite difference is mainly a partial volume effect Gö r a n St a r c k 1,2, Ba r b r o Vi k h o f f-ba a z 1,2, Ma r i a Lj u n g b e r g 1,2, Kristina Ma l m g r e n 4, Ev a For s se l l-ar o n s s o n 1,2, La r s Jö n s s o n 5 & Sv e n Ek h o l m 3 1 Division of Medical Physics and Biomedical Engineering, Sahlgrenska University Hospital, Göteborg, Sweden, 2 Department of Radiation Physics, University of Gothenburg, Göteborg, Sweden, 3 Department of Radiology, University of Rochester Medical Center, Rochester, NY, USA, 4 Institute of Neuroscience and Physiology, Epilepsy Research Group, Sahlgrenska Academy at the University of Gothenburg, Göteborg, and 5 Department of Radiology, Sahlgrenska University Hospital, Göteborg, Sweden Background: The concentration of N-acetylaspartate (NAA) in hippocampus, as measured with magnetic resonance spectroscopy (MRS), and the ratio of NAA/(choline (Cho) creatine (Cr)) are valuable tools in the lateralization of temporal lobe epilepsy (TLE). MRS of hippocampus is also increasingly used to study certain psychiatric and degenerative diseases. However, the reliability of such measurements of hippocampus has been questioned. Purpose: To re-evaluate MRS imaging data from prior control subjects with regard to variation of metabolite concentrations in hippocampus from anterior to posterior and the partial volume contribution to the measurements from adjacent tissue. Material and Methods: Twelve healthy subjects, mean age 33 years, were studied with MRS imaging. The measurement volume was angled along the temporal horns and metabolite concentration images were reconstructed at the MR system. Regions of interest (ROIs) in the anterior, medial, and posterior parts of both hippocampi were evaluated. Signal normalization to the total MRS signal from all ROIs permitted pooling of individual data with different and unknown signal scaling. One subject was re-examined with a high resolution three-dimensional (3D) volume of the brain for evaluation of partial volumes in the MRS examination. Results: Overall, there were significantly lower concentrations of NAA in the anterior parts, and of (Cho Cr) in the posterior parts, while the NAA/(Cho Cr) ratio in the posterior parts of the mesial temporal lobes was significantly higher. Hippocampus accounted for one-half, one-third, and one-quarter of the anterior, middle, and posterior ROIs, respectively. The NAA/ (Cho Cr) ratio thus showed a reverse relationship to the relative volume of hippocampal tissue within the ROI. Conclusion: Metabolite concentrations in the mesial temporal lobe obtained with MRS imaging represent the mean value of hippocampus and a considerable amount of adjacent tissue. To assess the hippocampus alone, an actual voxel well below 1 cm 3 and a sub-centimeter slice thickness are required. Key words: NAA; 1-H MRS; MRSI; hippocampus; temporal lobe epilepsy Göran Starck, Sahlgrenska University Hospital, Magnetkamera, Bruna Stråket 13, S Göteborg, Sweden (tel , fax , . Goran.Starck@VGregion.se) Submitted May 28, 2009; accepted for publication November 27, 2009 Magnetic resonance spectroscopy (MRS) is considered promising for clinical use in the presurgical evaluation of patients with intractable epilepsy. Although several studies have demonstrated the value of this technique 1 3, questions have been raised as to the trustworthiness and accuracy of these measurements. In 2000 Ve r m a t h e n et al. (4) published an article about anterior to posterior (AP) N-acetylaspartate (NAA) differences in epilepsy patients and subjects. They suggested that this AP difference could either be the result of fewer neurons in the anterior part compared with the posterior hippocampus, or a result of the normal variation in the thickness of hippocampus with subsequently more contributions from adjacent tissue posteriorly. This is an important question since the concentration of NAA and in particular the ratio of NAA/(choline (Cho) creatine (Cr)) has become an important tool in the lateralization of temporal lobe epilepsy (TLE). MRS of hippocampus is also increasingly used to study certain psychiatric and degenerative diseases. DOI / Informa UK Ltd. (Informa Healthcare, Taylor & Francis AS)

3 352 G. Starck et al. There are many methodological problems with MRS imaging of the mesial temporal lobe region; one of the most important is the relatively large slice thickness of the MRS imaging volume that so far has been used when compared with the thickness of the hippocam pus, in particular with regard to the difference in hippocampal volume from anterior to posterior. In the past our group has published a couple of articles within this field using MRS imaging to establish the origin of seizure focus and the results have overall seemed very promising (1, 2). Since this question of potential differences in me tabolite concentration has such importance with regard to all studies of disease processes in this region we decided to re-evaluate the data for our control subjects that were examined in the past and compare these results with those of other groups. Specifically, our aim was to evaluate the degree, if any, of anterior-posterior variation in NAA, (Cho Cr), and NAA/(Cho Cr) in the hippo campus region, and to examine if such an anterior-posterior variation may be related to partial volume contributions from adjacent brain tissue included in the MRS imaging data. To permit efficient pooling of the arbitrarily scaled individual MRS imaging data, a normalization factor had to be established. Material and Methods Metabolite data from neurologically healthy control subjects who had undergone MRS imaging in a previous study (2) were re-evaluated. Subjects younger than 18 years were excluded to avoid metabolite levels in the developing adolescent brain. The present study included data from 12 healthy subjects: 9 females and 3 males, mean age 33 years (range 23 48, median 29). All measurements and evaluations were performed on either of two 1.5T whole body MR systems (Gyroscan S15/ACS II and Gyroscan Intera, Philips Medical Sys tems, Best, The Netherlands). The standard 1 H transmit/receive quadrature mirror head coil was used for both imaging and spectroscopy. The MRS imaging procedure is briefly described below. The full details of the examination protocol were published earlier (1, 2). The MRS volume was planned on T2-weighted turbo-spin-echo (TSE) images and susceptibility sensitive images. A point-resolved spectroscopy (PRESS) sequence defined the volume of interest (VOI) and was angled along the temporal horns in the axial and sagittal images and symmetrically over the temporal lobes in the coronal images (2). The slice thickness was mm, and the left-to-right (LR) and AP dimensions varied between 60 and 100 mm depending on the size of the anatomical structures and the degree of susceptibility interference. Shimming reduced the full-width-at-halfmaximum (FWHM) of the water peak to 4 7 Hz. MRS quality from this VOI was assessed with a single voxel scan before proceeding with the MRS imaging (Fig. 1A). The size and position of the MRS imaging field of view (FOV) (200 mm) were the same as for the initial imag ing. The FOV was spatially encoded by sampling a circular region with a 32-point diameter, on a rectilinear grid in k-space. The quality of the MRS imaging was verified by inspection of MRS imaging voxel spectra (Fig. 1B). All MRS imaging measurements were performed with repetition time (TR) 2000 ms, echo time (TE) 272 ms, sampling frequency 2000 Hz with 1024 samples. A frequency-selective double inversion recovery technique with a band-width of 100 Hz was used for water suppression. Metabolite concentration images of NAA and (Cho Cr) were reconstructed at the MR system. All MRS imaging signals were B 0 -corrected, i.e. corrected for variations in the resonance frequency, time-dependent periodic magnetic field fluctuations synchronous with TR and zero th order of phase (5). Lorentz-Gauss filtering, and digital shift accumulation (DSA) filtering to suppress the remaining water signal, were applied in the spectral dimension. Cosine filtering was applied in the two spatial dimensions. The images were interpolated to (ACS II) or (Intera) pixels using Fourier interpolation (FoI), i.e. zero-filling in the two spatial dimensions before Fourier transformation (FT). The pixel values in the metabolite concentration images were obtained as the corresponding peak area(s). The spectral intervals for peak integration were chosen from a B 0 -corrected spectrum. No spatial correction of signal intensity was performed in the images. Six elliptic, nonoverlapping regions of interest (ROIs) were drawn in an oblique axial TSE image Fig. 1. Spectral quality was checked in all MRS imaging examinations. Modulus spectra from: A. the whole PRESS volume, B. a single MRS imaging voxel in the right anterior hippocampus/brainstem.

4 Anterior to posterior hippocampal MRS metabolite difference is mainly a partial volume effect 353 in the anterior part (anterior left (AL), anterior right (AR)), medial part (medial left (ML), medial right (MR)), and posterior part (posterior left (PL), posterior right (PR)) of left and right hippocampi, respectively. The ROIs were made as large as possible within each hippocampus and, hence, varied slightly between individuals (Fig. 2). These ROIs were evaluated for ROI area and mean signal intensity in the metabolite images. The obtained signal values were relative and varied from examination to examination. To pool the data from all subjects the signal values were normalized to the mean signal intensity of all ROIs. The normalization factor (Nf) was calculated, for each MRS imaging scan, as the mean signal sum Cho Cr NAA in the union of the six ROIs. Nf [( Cho+Cr ) + NAA ]. area n ( AL,AR,ML,MR,PL,PR) area n n ( AL, AR,ML,MR,PL,PR) n n n (1) The normalized metabolite signal was calculated as the signal intensity in the metabolite image divided by the normalizing factor. To check the efficiency in terms of reduction of scan to scan variation, another set of normalization factors was calculated for comparison. This set of factors was obtained through an iterative search, minimizing the sum of inter-examination signal variances of all ROIs. n [ Var[( Cho+Cr) ] +Var( NAA )] n n ( AL,AR,ML,MR,PL,PR) (2) One of the male subjects (aged 48 years) was reexamined (at the age of 53 years) on the Gyroscan Intera system with high resolution MRI for evaluation of partial volumes in the MRS examination. A threedimensional (3D) volume of the brain was acquired using a T 1 -weighted turbo field echo (T1TFE) (magnetization prepared) scan with 1.25 mm isotropic resolution, TR 6.84 ms, and TE 3.15 ms. The PRESS volume and the ROIs were carefully reproduced (Figs 3 and 4) and visually evaluated for partial volumes by two radiologists (S.E., L.J.). The spatial resolution of the MRS imaging can be described by the point spread function (PSF) or the spatial response function (SRF). These two are mathematically identical, their difference being purely conceptual. The PSF describes the image of an object consisting of one single point, whereas the SRF describes the regional contributions of the object to one single point (i.e. pixel) in the image. FoI calculates pixel values on a denser grid but does not change the SRF (6, 7). The spatial response of each ROI was, therefore, calculated as the sum of the SRFs of the interpolated MRS imaging pixels of the ROI. Contributions to the ROI signal were calculated, from the ROI contour and from the regions enclosed by the isocontours for the 10%, 50%, and 90% levels of the ROIs spatial response. Non-parametric statistics were used, i.e. the Wilcoxon signed rank sum test, when comparing paired (within-subject) observations and median and 95% confidence interval for the median when plotting data. The study was approved by the local ethics committee and informed consent was obtained from all subjects. Results Fig. 2. Position of the PRESS VOI for scanning and the ROIs for evaluation (black lines). The spatial response of each ROI is shown at 10% and 90% of peak level (white dashed lines). The spatial responses at 50% of peak level were congruent with the ROIs and were therefore omitted from the plot. The SRF is shown at the 10%, 50%, and 90% levels as reference for the spatial resolution of the MRS imaging. Centimeter scale at the bottom of the figure. The inter-subject variation in metabolite MR signal values in this study was several times larger than both the variation associated with accuracy of the measurements and the expected inter- and the observed intra-subject variations in metabolite concentration. Normalization to the mean signal intensity of all ROIs

5 354 G. Starck et al. Fig. 3. Axial and sagittal sections of hippocampus. Black lines indicate the PRESS volume, white solid lines the ROIs, and white dashed lines indicate the positions of the corresponding cross-sections in each pair of axial (upper panel) and sagittal (lower panel) views. (Eq. 1) reduced the coefficient of variation from 0.7 to 0.1 for NAA and for (Cho Cr) in all three regions of the hippocampi (Fig. 5). Also, the ratios of metabolite signal maximum over minimum decreased 10-fold after normalization. The ratio NAA/(Cho Cr) was not affected by the normalization. The normalized (Eq. 1) signal values were close to identical with those obtained when the sum of inter-examination variances (Eq. 2) was minimized. Both normalizations gave the same results in the comparisons of metabolite signals. The metabolite signals in the ROIs varied significantly from anterior to posterior. The NAA signal in the anterior ROI was lower than in the middle (P 0.001) and posterior (P 0.02) ROIs (Fig. 5). The (Cho Cr) signal in the posterior ROI was lower than in the middle ROI (P 0.05). The NAA/(Cho Cr) ratio showed a tendency to increase from anterior to posterior and there was a significant difference between anterior and posterior (P 0.05) and between middle and posterior (P 0.05) ROIs. No difference was found in the NAA or (Cho Cr) signals when comparing the left and right ROIs, neither when comparing the sum of the anterior, middle, and posterior ROIs, or the ROIs separately. The partial volume of hippocampus in the ROIs decreased from anterior to posterior (Figs 3 and 4). In total, an estimated three-quarters of the hippocampus volume was included in the anterior, middle, and posterior ROIs combined, but only about one-third of the total volume of the ROIs represented hippocampal tissue. As regards the anterior ROI: about 90% is gray matter (GM) and the remaining part is white matter (WM). The hippocampus occupies one-half of the anterior ROI volume and the remaining GM volume belongs to

6 Anterior to posterior hippocampal MRS metabolite difference is mainly a partial volume effect 355 Fig. 4. Coronal sections of hippocampus. Black lines indicate the PRESS volume, white solid lines the ROIs, and white dashed lines indicate the corresponding positions of the axial and the coronal cross-sections. the amygdala. As regards the middle ROI: about 40% is GM and the remainder is mainly WM with a small amount of cerebrospinal fluid (CSF). Hippocampus occupies approximately one-third of the middle ROI volume and the remaining GM volume belongs to the inferior part of putamen (basal ganglia). As regards the posterior ROI: approximately 65% is GM and the remaining part is mostly WM with a minor amount of CSF. The hippocampus occupies approximately one-quarter of the posterior ROI volume and the remaining GM volume belongs to thalamus. The spatial resolution of the MRS imaging in terms of the SRF with 2 cm 3 effective volume, and spatial responses of the ROIs with 3 cm 3 effective volumes is shown in Fig. 2. Approximately 100%, 60%, and 50% of the total ROI signal originated from inside the isocontours at 10% and 50% of the ROIs spatial response and the ROI contour, respectively. Discussion A few studies have previously been published regarding the metabolite distribution in the hippocampus. Hsu et al. (8), comparing single-voxel versus chemical shift imaging MR spectroscopy, obtained similar results with both techniques with regard to NAA/(Cho Cr) ratios (n 12). They found no side-by-side difference or significant difference between the anterior and posterior parts of the hippocampal regions. The latter is in contrast to the results of the present study. However, in a later paper Hsu et al. (9) reported significantly lower NAA/(Cho Cr) ratios in the anterior part (n 30). Ve r m a t h e n et al. (4) found, in agreement with this study, an increase in both NAA (n 10) and NAA/(Cho Cr) (n 14), as well as a decrease in Cho (n 10) and Cr (n 10) from anterior to posterior. Also, McLe a n et al. (10) found a trend towards lower NAA/(Cho Cr)

7 356 G. Starck et al. Fig. 5. MR signal intensity obtained from the pairs of ROIs in the anterior, middle, and posterior parts of both sides of hippocampi. The plots show median and 95% confidence interval for the median of the MR signal intensities of N-acetylaspartate and the sum of choline and creatine, respectively, A. with out normalization and B. normalized to the total MR signal. C. The quotient NAA/(Cho Cr). (n 10) in the anterior part of the hippocampus. However, their interpretation of these results was that this was more likely to be due to an increase in Cho anteriorly than to changes in NAA or Cr, which partly is in contrast to the results of the present study. The normal hippocampus has a size of 2 3 cm 3, but the size differs between men and women and between left and right (11 13). The curved elongated shape of the hippocampus has a (coronal) cross-section that decreases from anterior to posterior. In the re-examined subject, the hippocampus accounted for about onehalf, one-third, and one-quarter of the ROI volumes within the anterior, middle, and posterior part, respectively. Hence, extra hippocampal tissue dominated in the middle and posterior ROIs. In an attempt to examine the effect of extra hippocampal contributions to metabolite measurements with MR, Sh u f f et al. (14) segmented the hippocampus manually in MRI and compared metabolite concentrations and hippocampal content in MRS imaging voxels and found NAA to be inversely correlated with the hippocampal content. Their conclusion was that the NAA concentration in the hippocampus is lower than in the surrounding tissues, which is supported by our study. Ch u et al. (15), using a 4.1T magnet and smaller voxels, came to the same conclusion that the increase posteriorly in the NAA/Cr more likely resulted from an increased contribution from extra hippocampal tissues. In two other studies, Po u w e l s and Fr a h m (16) and Ch o i and Fr a h m (17) using the same single voxel technique, found lower NAA and higher Cho in the hippocampus when compared with cortical gray matter and white matter, and also when compared with thalamus. The Cr concentrations were similar in hippocampus, cortical gray matter, and thalamus, but lower in white matter. Ch o i and Fr a h m (17) also offered an explanation for their findings based on differences in cytoarchitecture of the neocortex and archicortexes of the hippocampus. Our findings of lower NAA anteriorly, as compared with the middle and posterior ROIs, and lower (Cho Cr) posteriorly as compared with the middle ROI, and the increasing trend from anterior to posterior in NAA/ (Cho Cr), support these data. Ve r m a t h e n et al., Sc h u f f et al., Hsu et al., and McLe a n et al. at 1.5T (4, 8 10, 14), and Ch u et al. at 4.1T (15), analyzed single MRS imaging voxels but the k-space sampling, spatial filtering, and slice thickness in their studies differed. Although the actual voxel in terms of the SRF or the necessary detail to calculate it is seldom reported, spatial in-plane resolution can be estimated and compared, in terms of the effective volume of the MRS imaging voxel divided by the slice thickness. The voxels in the present study had an effective in-plane size of 1 cm 2. This is the same in-plane size as in the 4.1T study, while the other cited studies indicated larger in-plane sizes, up to 2 cm 2. The ROIs analyzed in this study were of 1.5 cm 2 effective in-plane size, comparable with the MRS imaging voxels in the cited 1.5T studies but more adapted to the elongated shape of the hippocampus. Regardless of the approach, the low spatial resolution of the MRS imaging is the fundamental limitation that makes the analysis extend into structures both medial and lateral to the hippocampus. A problem when comparing studies performed with different MR systems is that the numbers for slice thickness are only approximately comparable. Slice profiles as well as definition of slice thickness can differ. For example, Ve r m a t h e n (4) noted that the 90 slice profile had a trapezoid shape while the 180 slices had imperfect nontrapezoid profiles, one of which is defining the MRS imaging slice in the standard PRESS sequence of their MR system. This is consistent with published data (18). However, the two MR systems used in this study used RF pulses with trapezoid profiles for the MRS imaging slice as well as for the two remaining dimensions of the PRESS selection (19). Slice thickness was defined by this manufacturer as the width containing 99% of the area under the slice profile. With realistic

8 Anterior to posterior hippocampal MRS metabolite difference is mainly a partial volume effect 357 slopes at the edges of the slice profile, as used in this study, an 18 or 20 mm thick slice will have a smaller FWHM. This is important since the FWHM is a more common definition of slice thickness used by other vendors. A basic assumption, when evaluating metabolite signal in different positions within an MRS imaging map, is that the signal sensitivity is independent of the ROI position. Imperfect slice profiles in the AP and LR directions can cause systematic errors in the measures of NAA and (Cho Cr) that depend on the position of the ROI. However, errors of this kind were not expected since the MR systems used in this work had trapezoidal slice profiles (19). Also, the main result showed different variations in NAA and (Cho Cr) depending on ROI position (Fig. 5), contradicting the proposition that the results could be caused by imperfect slice profiles. The different chemical shifts of the metabolites cause the slice excitations to be shifted correspondingly. The selected volume for (Cho Cr) will consequently be slightly shifted in the spatial dimensions with respect to the volume for NAA. However, the spatial encoding of voxels is based on phase encoding of spins, and is not affected by chemical shift differences. Therefore, the ROIs for NAA and for (Cho Cr) correspond to identical anatomical regions whereas the excitation volumes of NAA and (Cho Cr) can differ. This can cause systematic errors that act differently on the measures of NAA and (Cho Cr), especially at the edges of the selected volume (20). The chemical shift difference of approximately 1 ppm between (Cho Cr) and NAA caused an LR shift of approximately 3 mm and an AP shift of less than 1 mm of the selected volume in this study. The shift was calculated as chemical shift difference times VOI size in the shift direction divided by excitation bandwidth. It is evident from Fig. 2 that not even the ROIs closest to the edge of the VOI, i.e. the anterior ones, were influenced by this shift. Due to fundamental experimental conditions, for example, variations in relaxation rates and coil loading, the MR signal is obtained on a relative scale that varies between examinations and between subjects. Different metabolite ratios have therefore been used in the literature to make comparisons possible (3, 8, 15). However, NAA, Cho, and Cr have different biochemical roles, making their ratios difficult to interpret. In addition to the relative nature of MR signal, in this study, an arbitrary scaling of image signal intensity on a scan basis in the two MR systems caused a variation between examinations, much larger than would be explained by any reasonable variations in metabolite concentration and experimental conditions together. Therefore, normalization to a stable reference was necessary to make pooling of individual measurements possible without the use of metabolite ratios. With only ROI statistics on metabolite signal available, a possible and most stable reference was the mean signal intensity of all ROIs (Eq. 1). Normalization to this reference efficiently removed the inter-examination variation in MR signal, as well as between-subject global variation in metabolite signal. This result was confirmed by the normalization that minimized the sum of signal variances of all ROIs. However, within-subject signal contrast remained, allowing for comparisons between regions and metabolites. Still, caution is recommended when comparing single metabolites between groups, especially when pathology is present, since the normalization factor itself may also be affected by the pathology. For example, necrotic tumor in an ROI would contribute much less to the mean signal intensity of all ROIs, than would normal tissue. This would result in overall increased, biased, normalized signal values in this subject as compared with normalized signal values in a healthy subject. 1 H MRS and MRS imaging hold great clinical potential in the presurgical evaluation of TLE. Willman et al. (3) reported a meta-analysis of 22 studies focused on intractable TLE during comprising a total of 596 patients. The individual studies differed in methodological aspects, technical aspects, evaluation, and interpretation. Still, the meta-analysis gave an 82% positive predictive value for good outcome, for all patients with ipsilateral MRS abnormality. It should be emphasized that this result is based on MRS measurements that do not exclusively comprise hippocampus but also include portions of gray and white matter from adjacent brain structures within the measurement volume. The diagnostic importance of maximizing the partial volume of hippocampus was demonstrated by Im b e s i (21), who compared lateralization based on 1 H MRS in hippocampus using 8 cm 3 single voxels of cm 3 and cm 3, respectively. Lateralization improved with the rectangular voxel, which included more of hippocampus than the cubic voxel. Although this order of size is common for single voxel 1H MRS of hippocampus, smaller voxels have been used. Du c et al. (22) assessed hippocampal metabolite concentrations in TLE using cm 3 voxels and Ha m m e n et al. (23) correlated metabolite alterations in hippocampus with neuropathology in TLE, using the same voxel size. Despite the inclusion of substantial proportions of extrahippocampal tissues in relatively large voxels and thick slices, 1 H MRS and MRS imaging have been proven valuable in the lateralization of TLE. This is likely to be explained by a spread of neuronal and metabolic impairment in TLE into the surrounding tissue of the hippocampus, as has been reported for widespread areas of the brain (24 26). This surrounding

9 358 G. Starck et al. tissue is not necessarily structurally abnormal but may be affected metabolically, from the spread of seizure activity. However, to truly assess the hippocampus alone, especially the posterior part, a sub-centimeter slice thickness as well as an actual voxel size well below 1 cm 3 is required. This is not likely to be accomplished at 1.5T or even 3T but might be achievable with yet higher fields. This study had several limitations. The number of subjects was small, two separate MR systems were used, and only one subject was re-examined for evaluation of partial volumes. In conclusion, normalization to the mean signal intensity of all ROIs together effectively removed the confounding inter-examination variance in the MRS imaging data. Significantly lower NAA concentration was found in the anterior part of the mesial temporal lobe as compared with the middle and posterior parts. Also, significantly lower (Cho Cr) concentration was found in the posterior part of the mesial temporal lobe as compared with the middle part. Consequently, a significantly higher NAA/(Cho Cr) ratio was found in the posterior part of the mesial temporal lobe as compared with the middle and anterior parts. Hippocampus accounted for approximately one-half, one-third, and one-quarter of the ROIs in the anterior, middle, and posterior parts, and the fractions of gray matter were approximately 90%, 40%, and 65%, respectively. Hence, the NAA concentration co-varied anterior to posterior with the volume fractions of white matter. Acknowledgment This work was supported by grants from the Swedish Medical Research Council (grant no ) and the Lundberg Foundation, Göteborg, Sweden. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. References 1. Vikhoff-Baaz B, Ljungberg M, Starck G, Forssell-Aronsson E, Jönsson L, Alpsten M, et al. Performance of 2D 1H spectroscopic imaging of the brain: some practical considerations regarding the measurement procedure. Magn Reson Imaging 1999;17: Vikhoff-Baaz B, Malmgren K, Jönsson L, Starck G, Ljungberg M, Forssell-Aronsson E, et al. Lateralisation with magnetic resonance spectroscopic imaging in temporal lobe epilepsy: an evaluation of visual and region-of-interest analysis of metabolite concentration images. Neuroradiology 2001;43: Willmann O, Wennberg R, May T, Woermann FG, Pohlmann-Eden B. The role of 1H magnetic resonance spectroscopy in pre-operative evaluation for epilepsy surgery. A meta-analysis. Epilepsy Res 2006;71: Vermathen P, Laxer KD, Matson GB, Weiner MW. Hippocampal structures: anteroposterior N-acetylaspartate differences in patients with epilepsy and control subjects as shown with proton MR spectroscopic imaging. Radiology 2000;214: Klose U. In vivo proton spectroscopy in presence of eddy currents. Magn Reson Med 1990;14: Bracewell RN. The Fourier transform and its applications, chapter 18: The discrete Fourier transform. Second edition. Singapore: McGraw-Hill Book Co; p Vikhoff-Baaz B, Starck G, Ljungberg M, Lagerstrand K, Forssell-Aronsson E, Ekholm S. Effects of k-space filtering and image interpolation on image fidelity in 1 H MRSI. Magn Reson Imaging 2001;19: Hsu YY, Chang C, Chang CN, Chu NS, Lim KE, Hsu JC. Proton MR spectroscopy in patients with complex partial seizures: single-voxel spectroscopy versus chemical-shift imaging. Am J Neuroradiol 1999;20: Hsu YY, Chen MC, Lim KE, Chang C. Reproducibility of hippocampal single-voxel proton MR spectroscopy and chemical shift imaging. Am J Roentgenol 2001;176: McLean MA, Woermann FG, Simister RJ, Barker GJ, Duncan JS. In vivo short echo time 1H-magnetic resonance spectroscopic imaging (MRSI) of the temporal lobes. Neuroimage 2001;14: Jack CR Jr, Twomey CK, Zinsmeister AR, Sharbrough FW, Petersen RC, Cascino GD. Anterior temporal lobes and hippocampal formations: normative volumetric measurements from MR images in young adults. Radiology 1989;172: Maller JJ, Réglade-Meslin C, Anstey KJ, Sachdev P. Sex and symmetry differences in hippocampal volumetrics: before and beyond the opening of the crus of the fornix. Hippocampus 2006;16: Eckerström C, Olsson E, Borga M, Ekholm S, Ribbelin S, Rolstad S, et al. Small baseline volume of left hippocampus is associated with subsequent conversion of MCI into dementia: the Göteborg MCI study. J Neurol Sci 2008;272: Schuff N, Ezekiel F, Du AT, Weiner MW. Automated voxel selection and assessment of partial volume effects in 1H MRSI studies of hippocampus [abstract #425]. Proc 8th Int Soc Magn Reson Med, Chu WJ, Kuzniecky RI, Hugg JW, Abou-Khalil B, Gilliam F, Faught E, et al. Statistically driven identification of focal metabolic abnormalities in temporal lobe epilepsy with corrections for tissue heterogeneity using 1H spectroscopic imaging. Magn Reson Med 2000;43: Pouwels PJ, Frahm J. Regional metabolite concentrations in human brain as determined by quantitative localized proton MRS. Magn Reson Med 1998;39: Choi CG, Frahm J. Localized proton MRS of the human hippocampus: metabolite concentrations and relaxation times. Magn Reson Med 1999;41: Burtscher IM, Johansson E, Holtås S, Ståhlberg F. Quality assessment of localization technique performance in small volume in vivo 1H MR spectroscopy. Magn Reson Imaging 1999;17: Vikhoff-Baaz B, Starck G, Ljungberg M, Forssell- Aronsson E, Ekholm S. Signal profile measurements for experimental evaluation of the PRESS sequence [abstract #374]. ESMRMB Annual Meeting, 2000.

10 Anterior to posterior hippocampal MRS metabolite difference is mainly a partial volume effect Starck G, Vikhoff-Baaz B, Ekholm S, Forssell-Aronsson E. Intensity correction may introduce left/right and anterior/ posterior differences in volume selective CSI [abstract #213]. ESMRMB Annual Meeting, Imbesi SG. Proton magnetic resonance spectroscopy of mesial temporal sclerosis: Analysis of voxel shape and position to improve diagnostic accuracy. J Comput Assist Tomogr 2006;30: Duc CO, Trabesinger AH, Weber OM, Meier D, Walder M, Wieser HG, et al. Quantitative 1H MRS in the evaluation of mesial temporal lobe epilepsy in vivo. Magn Reson Imaging 1998;16: Hammen T, Hildebrandt M, Stadlbauer A, Doelken M, Engelhorn T, Kerling F, et al. Non-invasive detection of hippocampal sclerosis: correlation between metabolite alterations detected by (1)H-MRS and neuropathology. NMR Biomed 2008;21: Capizzano AA, Vermathen P, Laxer KD, Matson GB, Maudsley AA, Soher BJ, et al. Multisection proton MR spectroscopy for mesial temporal lobe epilepsy. Am J Neuroradiol 2002;23: Vermathen P, Laxer KD, Schuff N, Matson GB, Weiner MW. Evidence of neuronal injury outside the medial temporal lobe in temporal lobe epilepsy: N-acetylaspartate concentration reductions detected with multisection proton MR spectroscopic imaging initial experience. Radiology 2003;226: Hetherington HP, Kuzniecky RI, Vives K, Devinsky O, Pacia S, Luciano D, et al. A subcortical network of dysfunction in TLE measured by magnetic resonance spectroscopy. Neurology 2007;69: