Proton magnetic resonance spectroscopy yields metabolic information and has proved to be a useful addition to structural imaging in neurological diseases. We applied short-echo time Spectroscopic Imaging in a cohort of 42 patients with sec- ondary progressive multiple sclerosis (SPMS). Linear modelling with respect to brain tissue type yielded metabolite levels that were significantly different in white matter lesions compared with normal-appearing white matter, suggestive of higher myelin turnover (higher choline), higher metabolic rate (higher creatine) and increased glial activity (higher myo-inositol) within the lesions. These findings suggest that the lesions have ongoing cellular activity that is not consistent with the usual assumption of ‘chronic’ lesions in SPMS, and may represent a target for repair therapies. Keywords Multiple sclerosis · Magnetic resonance spectroscopy · Brain metabolites · White matter lesions · Normal- appearing white matter * Ian Marshall Department of Brain and Behavioural Sciences, University Ian.Marshall@ed.ac.uk of Pavia, Pavia, Italy Brain MRI 3T Research Centre, IRCCS Mondino Centre for Clinical Brain Sciences, University of Edinburgh, Foundation, Pavia, Italy Edinburgh, UK Edinburgh Clinical Trials Unit, Usher Institute of Population Institute of Cardiovascular and Medical Sciences, University Health Sciences and Informatics, University of Edinburgh, of Glasgow, Glasgow, UK Edinburgh, UK Centre for Cardiovascular Sciences, University of Edinburgh, Division of Health Sciences, University of Warwick, Edinburgh, UK Warwick, UK Edinburgh Imaging QMRI Facility, University of Edinburgh, Institute for Science and Technology in Medicine, Keele Edinburgh, UK University, Newcastle, UK Dental Translational and Clinical Research Unit, School Academic Department of Neuroscience, The Sheffield NIHR of Dentistry, Faculty of Medicine and Health, University Translational Neuroscience Biomedical Research Centre, of Leeds, Leeds, UK University of Sheffield, Sheffield, UK Department of Neurology, Barts and the London NHS Trust, Queen Square Multiple Sclerosis Centre, Department London, UK of Neuroinflammation, UCL Institute of Neurology, UCL Institute of Neurology, Queen Square MS Centre, University College London, London, UK University College London, London, UK Vol.:(0123456789) 1 3 1796 Journal of Neurology (2018) 265:1795–1802 Introduction Methods Multiple sclerosis (MS) is a disabling neurological disease 111 patients with SPMS who had had no relapses within affecting some 2.5 million people worldwide (http://www. the previous 3 months gave informed consent to be mstru st.org.uk). There are currently no effective treatments recruited in our centre as part of the study. At baseline, 43 for the progressive phases of MS when disability accu- patients (mean age 55 (standard deviation, SD 8) years; mulates irreversibly. In clinical trials of potential disease- 30 female and mean EDSS score 6.0 (SD 0.7)) underwent modifying therapies, the standard outcome measure is the MRSI as part of an MRI examination at 3T (Siemens Expanded Disability Status Scale (EDSS) . In imaging Verio, Siemens Healthcare, Erlangen, Germany) using a studies, lesion load and brain atrophy are widely used bio- standard 12-channel matrix head coil. PD-T2-weighted and markers. White matter lesions (WML) are visible on routine FLAIR structural scans were acquired parallel to the ante- magnetic resonance imaging (MRI) sequences, and represent rior commissure–posterior commissure (ACPC) line with current or previous inflammatory activity. More advanced matrix 256 × 256, field of view (FOV) 250 × 250 mm and MRI methods enable estimation of myelin status through 60 contiguous slices 3 mm thick. The sequence timings white matter connectivity and magnetisation transfer imag- were TR/TE1/TE2 = 3050/31/82 ms (turbo factor 7) and ing . Complementary information on neuronal integrity TR/TI/TE = 9500/2400/124 ms (turbo factor 28) respec- is available from proton MR spectroscopy and spectroscopic tively. A 3D inversion recovery prepared T1-weighted imaging (MRSI) studies of brain metabolites. gradient echo scan (MPRAGE) was also acquired with A consistent finding of spectroscopic studies is that the matrix 256 × 256, field of view (FOV) 250 × 250 mm, level of the neuronal marker N-acetyl aspartate (NAA) is 160 sagittal slices 1 mm thick, flip angle 8°, TR/TI/ reduced in the normal-appearing white matter (NAWM) of TE = 2400/1000/3 ms and parallel imaging acceleration patients compared with controls [3–5], and further reduced factor 2. in WML [4, 6], although this latter finding may apply spe- Proton MRSI was carried out using a manufacturer- cifically to chronic rather than acute lesions [ 7]. In a lon- supplied prototype with semi-LASER excitation  and gitudinal study, Obert et al.  found that NAA decreased TR/TE of 2000/43 ms. Data were acquired from a sin- in the NAWM of patients with secondary progressive MS gle 10-mm-thick slice of spectroscopic voxels prescribed (SPMS) and in the WML of patients with relapsing–remit- immediately above the lateral ventricles in the plane of ting MS (RRMS). the T2-weighted images (Fig. 1). Thus, the MRSI slice Other findings from these studies are that choline, a thickness extended through three adjacent T2-weighted potential marker of myelin turnover, is increased in acute images. The excitation volume was 120 mm × 120 mm. lesions relative to NAWM , and in acute WML and The spatial encoding matrix was 24 × 24 with elliptical NAWM relative to WM in control subjects . Raised lev- k-space sampling, interpolated to 32 × 32 by the scanner, els of myo-inositol, associated with glial activity, have been resulting in 1 mL spectroscopic voxels. Free induction found in acute WML, chronic WML and NAWM relative decays consisted of 1024 samples with a dwell time of to WM in controls , in grey matter (GM) and NAWM 500 µs. Weak water suppression was applied, together with relative to controls , and in WML relative to NAWM . four 30-mm-thick saturation bands positioned to suppress Srinivasan et al.  found increased levels of glutamate, scalp lipid signals. but not glutamine, in acute WML and NAWM relative to Spectroscopic data were analysed in LCModel  control WM. using a spectral basis set matching the semi-LASER The non-invasive nature of MRSI lends itself to use in sequence, generated using simulation routines within the longitudinal studies, both to monitor the natural progres- FID-A toolkit . The LCModel results for NAA, cho- sion of disease and in clinical trials of potential disease- line, creatine, myo-inositol and Glx (the sum of glutamate modifying treatments. One such trial is MS-SMART (http:// and glutamine) were multiplied by the scanner transmitter w w w . m s - s m a r t . o r g), a 2-year, multicentre trial of three reference voltage to adjust for subject-specific loading of repurposed drugs for treatment of SPMS. Here, we report the head coil . Finally, metabolite measurements were on MRSI measurements of brain metabolites made at the corrected for relaxation effects using literature values for baseline time point of this trial. Using image segmentation T1, derived from NAWM in a group of RRMS and SPMS and linear modelling we were able to investigate metabolite patients  to yield concentrations in Institutional Units. differences between tissue types. Voxels were excluded from further analysis if they were judged to be not completely inside the brain or if the LCModel Cramer–Rao bounds (i.e., % standard devia- tions) of the fitted spectra exceeded 20% for any metabolite 1 3 Journal of Neurology (2018) 265:1795–1802 1797 Fig. 1 Prescription of the MRSI scan showing (left) the semi-LASER excitation region with red outline and outer volume suppression regions with blue outlines; (right) angulation of the MRSI slice parallel to the AC–PC line immediately above the lateral ventricles or 40% for myo-inositol . Additionally, voxels judged by C = Intercept + (NAWM + GM) + (NAWM − GM) met 1 2 an experienced spectroscopist to have poor quality spectra + WML + (subject eﬀect). (e.g., with grossly distorted baselines) were discarded. The 3D T1-weighted and 2D FLAIR structural volumes where the metabolite concentrations C and the tissue met were co-registered to the T2-weighted images and seg- probabilities (NAWM, GM and WML) have been meas- mented to generate tissue probability maps for cerebrospi- ured for each voxel. The β coefficients, Intercept and sub- nal fluid (CSF), GM, NAWM and WML using tools freely ject effects are estimated by the model. The model fit was available in FSL (https://f sl.fmrib.o x.ac.uk) and ANTS . assessed by examining plots of residuals. We also ran an WML probability masks were automatically created based extended model that included the EDSS score as a vari- on a FLAIR signal intensity threshold chosen empirically to able. 19(17) Patients had EDSS scores of 6.0(6.5), with the provide WML masks that optimised the correspondence to remaining six patients having scores in the range 4.0 to 5.5. masks manually segmented by an experienced neuroradiolo- Given that the data were so sparse for these latter values, we gist. The tissue segmentation maps were resampled to MRSI combined them with the patients having EDSS of 6.0, i.e., resolution and summed over three adjacent slices to match dichotomising the patients into two groups with EDSS ≤ 6.0 the MRSI slice thickness, thereby determining the overall and EDSS = 6.5. fractional tissue content corresponding to each spectrum. Tissue probabilities were used to predict metabolite concentrations (and ratios with respect to creatine) in a lin- Results ear mixed model in SAS Studio 3.5 (http://www.sas.com). Subjects were regarded as random effects (i.e. a random Image registration failed for one patient, who was removed intercept model) to allow for within-subject correlation of from the analysis. In the remaining 42 patients, a total of voxels and so that we could look specifically at differences 5349 spectroscopic voxels were located wholly within brain in metabolite levels between tissue types within subject. To tissue, of which 4558 (85%) passed the Cramer–Rao tests. avoid problems with the collinearity of NAWM and GM, 916 (20%) of these voxels failed the visual test of spectral we used the variables (NAWM + GM), (NAWM-GM) and quality, leaving 3642 voxels (87 ± 17 per patient: range WML in the regression. Thus, the model was 1 3 1798 Journal of Neurology (2018) 265:1795–1802 33–116) in the final analysis. Tissue probabilities averaged Significant differences were found between all three tissue across all accepted voxels were GM 31%, NAWM 56% and types for all the metabolite ratios with respect to creatine. WML 3%, with the balance being CSF. An example of tis- Specifically, despite the higher creatine in WML compared sue segmentation is shown in Fig. 2. The overall mean (SD) with NAWM (Table 1), the ratios choline/creatine and myo- linewidth of all accepted spectra was 7.0 (2.3) Hz. Repre- inositol/creatine are also higher in WML than NAWM. On sentative fitted spectra are shown in Figs. 3 and 4. the other hand, the ratios NAA/creatine and Glx/creatine are Tissue-specific metabolite concentrations determined lower in WML than NAWM, whereas NAA and Glx were from the linear mixed model (disregarding EDSS score) are individually not different between WML and NAWM. given in Table 1 and Fig. 5. Several previous studies [3, 6, In the ancillary mixed model with dichotomised EDSS 14, 15] have reported ratios with respect to creatine and we scores, we found no significant differences with respect have included these in Table 2 and Fig. 6. All metabolite to EDSS for any metabolite or metabolite ratio (all p concentrations and ratios are reported as mean (standard values > 0.3). error). We found that the level of choline was higher in NAWM than GM (196 (5) vs 147 (5) Institutional Units) whereas creatine was lower (809 (14) vs 858 (15)) as was Glx (1164 Discussion (21) vs 1392 (24)) (p < 0.001 for all these comparisons). There were no significant differences in myo-inositol level Short-TE MRSI coupled with tissue segmentation and lin- between GM and NAWM, nor in NAA levels between any ear modelling enabled estimation of brain metabolite levels of GM, NAWM and lesions. in GM, NAWM and WML despite very few voxels having Choline, creatine and myo-inositol levels were all signifi- ‘pure’ tissue content. Including subjects as random effects cantly (p < 0.001) higher in WML compared with NAWM enabled us to focus on metabolite differences between tis- and GM. There were no significant differences in NAA or sue types. Our findings for NAWM compared with GM are Glx levels between WML and NAWM, but Glx was lower consistent with the literature except that we found no signifi- in WML than in GM (p = 0.002). cant difference in myo-inositol levels. Of particular interest is our finding that WML had higher choline, creatine and myo-inositol levels than did NAWM, suggesting ongoing abnormal metabolism in these lesions. NAWM–GM differences Our findings for the concentrations of choline, creatine and Glx in NAWM compared with GM are consistent with previ- ous studies in healthy controls [12, 16, 17]. We found no significant difference in NAA level between NAWM and GM. Studies in control subjects have found NAA to be either lower or higher in GM than in WM, with a range from 15% lower to 46% higher being reported by Schuff et al. . They speculated that these mixed find- ings might in part be due to regional variations, the age of the participants, relaxation values or technical differences in acquisition. Our finding of no significant difference between NAWM and GM levels of myo-inositol differs from Llufriu et al.  who found levels 50% higher in GM in a cohort of 59 patients with MS of undisclosed subtype. Similarly, in an early single-voxel MRS study of healthy young controls, Michaelis et al.  found myo-inositol levels 30% higher in GM than WM. The different excitation scheme (STEAM Fig. 2 MRSI voxel grid and tissue segmentations overlaid on the rather than the usual PRESS or semi-LASER), shorter echo T2-weighted image for a representative patient. The spectroscopic time (20 vs 43 ms) and mixed tissue content in that study excitation volume is shown with a red outline. Green shading indi- cates NAWM and pale blue indicates WML may partly explain the disparate findings. 1 3 Journal of Neurology (2018) 265:1795–1802 1799 Fig. 3 Spectra for the same patient as in Fig. 2 fitted by LCModel. Those shown in red passed the Cramer–Rao Bounds tests, whilst those shown in black did not. Spectra passing the CRB tests but judged to be of poor quality are shown in blue in WML. As did Obert et al. , we found increased myo- WML–NAWM differences inositol in WML relative to NAWM. Srinivasan et al.  used a custom single-voxel spectro- Numerous studies have compared NAWM in MS patients scopic sequence designed to separate the heavily overlapped with NAWM in control subjects, but relatively few have glutamate and glutamine components of the composite ‘Glx’ compared NAWM with WML. This is likely due to the signal. In a mixed cohort of RRMS, primary and secondary difficulties of making MRS measurements in lesions of progressive MS patients, they compared 12 chronic WML generally small size, irregular shape and low overall load. voxels with 17 NAWM voxels. They found that NAA, Cucurella et al.  used a single 8 mL MRS voxel in choline and glutamate were lower in chronic WML than each of 18 SPMS patients, placed either in predominately NAWM. When including the signal from glutamine, there NAWM or predominately WML, finding that NAA was was no difference for the combined Glx, as we found in the lower in WML than NAWM. Kapeller et al.  found present study. reduced NAA and increased myo-inositol in WML relative Our finding of significantly higher levels of choline, to NAWM in a group of 32 patients with MS of undis- creatine and myo-inositol in WML relative to NAWM sug- closed subtype. gest higher membrane turnover, higher metabolic rate and Using a mixed linear model we found no difference in increased glial activity, respectively. Inflammatory dis- NAA or Glx between NAWM and WML although the ratios ease activity was low in this cohort, as evidenced by the NAA/cre and Glx/cre were respectively lower and higher 1 3 1800 Journal of Neurology (2018) 265:1795–1802 Fig. 4 Representative examples of (left column, in red) spectra that passed the Cramer–Rao tests but which were rejected at visual that passed the Cramer–Rao tests; (middle column, in black) spec- inspection. LCModel baselines are shown with dashed lines, and tra that failed the Cramer–Rao tests; (right column, in blue) spectra spectral fits with heavy lines. All spectra have the same scaling Table 1 Principal metabolite concentrations in institutional units (mean ± std error) with respect to brain tissue type in 42 patients with SPMS NAWM WML GM Choline * 263 ± 9 * 196 ± 5 * 147 ± 5 Creatine * 958 ± 32 * 809 ± 14 * 858 ± 15 NAA 0.74 1469 ± 62 0.71 1449 ± 23 0.96 1447 ± 27 Myo-inositol * 677 ± 24 * 463 ± 10 0.50 469 ± 13 Glx 0.28 1220 ± 53 * 1164 ± 21 * 1390 ± 24 Significance of pairwise comparisons is shown as actual p values or as * (p < 0.01) Fig. 5 Principal metabolite concentrations (mean ± std error) in insti- NAWM normal-appearing white matter, WML white matter lesions, tutional units with respect to brain tissue type. Columns are shaded GM grey matter grey, white and black to indicate GM, NAWM and WML, respec- tively. NS: not significantly different. Further details in Table 1 eligibility criteria and the structural MR brain features. The cohort can, therefore, reasonably be expected to have WML that were overwhelmingly non-recent. However, it would General points be simplistic to assume that the lesions were “chronic” (a neuropathological descriptor defined by apparent biologi- A strength of our study is that the study group comprised cal inertness at post mortem examination) despite the phe- a well-characterised cohort of SPMS patients with a nar- notypic definition of “secondary progressive MS”. Indeed, row range of EDSS scores. A limitation is that there was our findings suggest that non-recent WML in SPMS patients no healthy control group and so it is difficult to know how who are EDSS ≤ 6.5 (i.e., not EDSS 10 [dead]) exhibit ongo- ‘normal’ the NAWM really is at this stage of the disease. ing cellular activity that is not consistent with the pathologi- The MS-SMART study is designed to compare two-year cal definition of “chronic”, and may represent a targetable progression in four randomised groups (three candidate drug substrate for repair therapies. treatments and one placebo) and the main analysis will be 1 3 Journal of Neurology (2018) 265:1795–1802 1801 Table 2 Principal metabolite ratios relative to creatine (mean ± std resulting in institutional units. Other studies [3, 6, 15] error) with respect to brain tissue type in 42 patients with SPMS have avoided these steps and reported ratios with respect to creatine, which was found or assumed to have a similar NAWM WML GM concentration in different tissues. In fact, creatine is often Choline/creatine 0.25 ± 0.01 0.29 ± 0.01 0.17 ± 0.01 assumed to be a stable benchmark for hypothesis testing in NAA/creatine 1.80 ± 0.02 1.55 ± 0.05 1.66 ± 0.03 MRS studies. Our work suggests that this is not necessarily Myo-inositol/creatine 0.57 ± 0.01 0.73 ± 0.02 0.53 ± 0.01 the case, since taking ratios with respect to creatine changed Glx/creatine 1.46 ± 0.02 1.26 ± 0.05 1.64 ± 0.02 our findings for NAA and Glx. We, therefore, advise caution for any MRS analysis based on the assumption of creatine All pairwise comparisons are statistically significant at p < 0.01 being a stable reference. NAWM normal-appearing white matter, WML white matter lesions, GM grey matter The current analysis takes no account of the position of the voxels within the MRSI slice, and thus effectively aver - ages tissue metabolite concentration values across the whole slice. Future methodological refinements could include regional analysis and absolute quantification of metabolites. Although the latter requires the use of calibrated phantoms and is time consuming, interpreting the measurements in terms of absolute concentrations (e.g., millimoles per litre) rather than Institutional Units should make it easier to com- pare results between centres. The complex nature of MRS/ MRSI studies requires the detailed reporting of acquisition and analysis methods. Conclusion MS-SMART is one of the largest studies of SPMS, with 440 patients recruited across 13 UK centres. Longitudinal changes in clinical status, structural features and metabolites will be investigated at the 2-year endpoint. Preliminary find- Fig. 6 Principal metabolite ratios (mean ± std error) relative to cre- ings at baseline in our MRSI sub-group show higher choline, atine with respect to brain tissue type. Columns are shaded grey, white and black to indicate GM, NAWM and WML, respectively. creatine and myo-inositol in WML compared with NAWM, Further details in Table 2 indicating higher myelin turnover, higher metabolic rate and increased glial activity respectively. These suggest that the comparison of clinical status, structural features and metabo- lesions may have continuing, abnormal metabolism despite lites for each of the candidate drugs relative to placebo at the these patients being in a progressive phase of their disease end of the trial. Interestingly, Obert et al.  have recently during which lesions are often assumed to be ‘chronic’ and reported a reduction in NAWM NAA in 15 SPMS patients not active. This finding may have wider implications for the over the course of a 2-year study. Our finding that metabo- understanding of pathobiology of non-recent lesions in vivo, lite values and ratios were not related to the EDSS scores and for stratification for studies that evaluate repair thera- at baseline in our study group is not surprising given the pies. Adequately powered longitudinal studies are necessary study design. to establish the usefulness of MRSI and other quantitative We used a single slice of MRSI, nominally placed imme- MRI techniques in monitoring disease progression and eval- diately above the lateral ventricles. Depending on the exact uating potential treatments. placement of the slice and the size and shape of the patient’s Acknowledgements This work was reported in preliminary form as brain, the potential number of brain voxels varied. However, Marshall I et al., Proceedings of the 25th Annual Meeting of the Inter- reduced numbers were mainly due to rejection based on the national Society for Magnetic Resonance in Medicine, Honolulu, USA, Cramer Rao bounds and visual quality assessment. The latter 2017; 4639. This independent research is awarded by the Efficacy and is time consuming and necessarily subjective, thus leading Mechanism Evaluation Programme (EME: reference 11/30/11), funded by the Medical Research Council (MRC) and the Multiple Sclerosis to interest in developing machine-learning approaches to Society (MS Society), and managed by the National Institute for Health spectral quality control . Research (NIHR) on behalf of the MRC-NIHR partnership. Imaging We corrected the ‘raw’ metabolite measurements using (Edinburgh Clinical Research Facility reference E131282) was car- assumed T1 relaxation times and for loading of the coil, ried out at the Edinburgh Imaging Queen’s Medical Research Institute 1 3 1802 Journal of Neurology (2018) 265:1795–1802 facility, (http://www.ed.ac.uk/edinbur gh-imaging ), University of Edin- magnetic resonance spectroscopy at 3 T. Brain 128:1016–1025. burgh, part of the SINAPSE collaboration (http://www.sinapse.ac.uk ). https ://doi.org/10.1093/brain /awh46 7 MJT was funded by the National Health Service Lothian Research 8. Obert D, Helms G, Sattler MB, Jung K, Kretzschmar B, Bahr and Development Office. CJW was supported in this work by NHS M, Dechent P, Diem R, Hein K (2016) Brain metabolite changes Lothian via the Edinburgh Clinical Trials Unit. The MRSI sequence in patients with relapsing-remitting and secondary progressive was provided by Siemens Healthcare as a Works in Progress package. multiple sclerosis: a two-year follow-up study. PLoS One. https ://doi.org/10.1371/journ al.pone.01625 83 9. Scheenen TWJ, Klomp DWJ, Wijnen JP, Heerschap A (2008) Compliance with ethical standards Short echo time H-1-MRSI of the human brain at 3T with mini- mal chemical shift displacement errors using adiabatic refocus- Conflict of interest On behalf of all authors, the corresponding author ing pulses. Magn Reson Med 59(1):1–6. https ://doi.org/10.1002/ states that there are no relevant conflicts of interest. mrm.21302 10. Provencher SW (1993) Estimation of metabolite concentrations Ethical standards The MS-SMART study was approved by the Scot- from localized in vivo proton NMR spectra. Magn Reson Med land ‘A’ Research Ethics Committee (reference 13/SS/0007), and there- 30(6):672–679. https ://doi.org/10.1002/mrm.19103 00604 fore, complies with the standards laid down in the 1964 Declaration 11. Simpson R, Devenyi GA, Jezzard P, Hennessy TJ, Near J (2017) of Helsinki and later amendments. All patients gave written informed Advanced processing and simulation of MRS data using the FID consent. appliance (FID-A)An open source, MATLAB-based toolkit. Magn Reson Med 77(1):23–33. https ://doi.org/10.1002/mrm.26091 12. Michaelis T, Merboldt KD, Bruhn H, Hanicke W, Frahm J (1993) Open Access This article is distributed under the terms of the Crea- Absolute concentrations of metabolites in the adult human brain tive Commons Attribution 4.0 International License (http://creat iveco in vivo: quantification of localized proton MR spectra. Radiology mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- 187(1):219–227 tion, and reproduction in any medium, provided you give appropriate 13. Avants B, Gee JC (2004) Geodesic estimation for large deforma- credit to the original author(s) and the source, provide a link to the tion anatomical shape averaging and interpolation. Neuroimage Creative Commons license, and indicate if changes were made. 23:S139-S150. https://doi.or g/10.1016/j.neuroimag e.2004.07.010 14. Caramanos Z, DiMaio S, Narayanan S, Lapierre Y, Arnold DL (2009) H-1-MRSI evidence for cortical gray matter pathology that is independent of cerebral white matter lesion load in patients with References secondary progressive multiple sclerosis. J Neurol Sci 282(1– 2):72–79. https ://doi.org/10.1016/j.jns.2009.01.015 1. Kurtzke JF (1983) Rating neurologic impairment in Multiple Scle- 15. Parikh J, Thrippleton MJ, Murray C, Armitage PA, Harris BA, rosis—an Expanded Disability Status Scale (EDSS). Neurology Andrews PJD, Wardlaw JM, Starr JM, Deary IJ, Marshall I (2015) 33(11):1444–1452. https ://doi.org/10.1212/wnl.33.11.1444 Proton spectroscopic imaging of brain metabolites in basal gan- 2. Miller DH, Thompson AJ, Filippi M (2003) Magnetic resonance glia of healthy older adults. Magn Reson Mater Phys Biol Med studies of abnormalities in the normal appearing white matter and 28(3):251–257. https ://doi.org/10.1007/s1033 4-014-0465-0 grey matter in multiple sclerosis. J Neurol 250(12):1407–1419. 16. Schuff N, Ezekiel F, Gamst AC, Amend DL, Capizzano AA, https ://doi.org/10.1007/s0041 5-003-0243-9 Maudsley AA, Weiner MW (2001) Region and tissue differ - 3. Leary SM, Davie CA, Parker GJM, Stevenson VL, Wang LQ, ences of metabolites in normally aged brain using multislice 1H Barker GJ, Miller DH, Thompson AJ (1999) H-1 magnetic reso- magnetic resonance spectroscopic imaging. Magn Reson Med nance spectroscopy of normal appearing white matter in primary 45(5):899–907. https ://doi.org/10.1002/mrm.1119 progressive multiple sclerosis. J Neurol 246(11):1023–1026. https 17. Wijnen JR, van Asten JJA, Klomp DWJ, Sjobakk TE, Gribbestad ://doi.org/10.1007/s0041 50050 507 IS, Scheenen TWJ, Heerschap A (2010) Short echo time H-1 4. Cucurella MG, Rovira A, Rio J, Pedraza S, Tintore MM, Mon- MRSI of the human brain at 3T with adiabatic slice-selective refo- talban X, Alonso J (2000) Proton magnetic resonance spec- cusing pulses; reproducibility and variance in a dual center set- troscopy in primary and secondary progressive multiple scle- ting. J Magn Reson Imaging 31(1):61–70. https: //doi.org/10.1002/ rosis. NMR Biomed 13 (2):57–63. https://doi.org/10.1002/ jmri.21999 (sici)1099-1492(200004)13:2<57::aid-nbm609>3.3.co;2-x 18. Kapeller P, Brex PA, Chard D, Dalton C, Griffin CM, McLean 5. Llufriu S, Kornak J, Ratiney H, Oh J, Brenneman D, Cree BA, MA, Parker GJM, Thompson AJ, Miller DH (2002) Quantitative Sampat M, Hauser SL, Nelson SJ, Pelletier D (2014) Magnetic (1)H MRS imaging 14 years after presenting with a clinically resonance spectroscopy markers of disease progression in multiple isolated syndrome suggestive of multiple sclerosis. Multi Scler J sclerosis. JAMA Neurol 71(7):840–847. https ://doi.org/10.1001/ 8(3):207–210. https ://doi.org/10.1191/13524 58502 ms822 oa jaman eurol .2014.895 19. Kyathanahally SP, Mocioiu V, Pedrosa de Barros N, Slotboom J, 6. Tedeschi G, Bonavita S, McFarland HF, Richert N, Duyn JH, Wright AJ, Julia-Sape M, Arus C, Kreis R (2017) Quality of clini- Frank JA (2002) Proton MR spectroscopic imaging in multiple cal brain tumor MR spectra judged by humans and machine learn- sclerosis. Neuroradiology 44(1):37–42 ing tools. Magn Reson Med. https ://doi.org/10.1002/mrm.26948 7. Srinivasan R, Sailasuta N, Hurd R, Nelson S, Pelletier D (2005) Evidence of elevated glutamate in multiple sclerosis using 1 3
Journal of Neurology – Springer Journals
Published: May 30, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera