MAGMA Magr~tic Re~mar, ce Materials Lrt Physic, Biol{~" and Mt~dMne

E L S E V I E R Magnetic Resonance Materials in Physics, Biology and Medicine 13 (2001) 127-133 ..................................... www.elsevier.com/locate/magma

Regional metabolite levels of the normal posterior %ssa studied by proton chemical shift imaging

Damien Galanaud, Yann Le Fur, Frangois Nicoli, Baatrice Denis, Sylviane Confort-Gouny, Jean-Philippe Ranjeva, Patrick Viout, Jean Pelletier,

Patrick J. Cozzone * Centre de R&onance Magn&ique Biologique et M&licab..', CNRS UMR6612, FacultO de, MOdec&e, 27, Bvd Jean Moulin,

F- 13005 Marseille, France

Received 6 March 2001; received in revised form 14 May 2001; accepted 15 May 2001

Abstract

MR spectroscopy of the posterior fossa is pitted with numerous technical difiqculties. It is, however, of great clinical interest in the study of the degenerative diseases and tumors of this area. We have developed a method to perform 2D CSI of this area, by using a sagittal slice and a careful positioning of outer volume saturation. We performed this acquisition in 30 healthy volunteers to determine the normal metabolic ratios in five voxels of this area (mesencephalon, pons, medulla oblongata, vermis, cerebellar white matter). The main technical difficulty was magnetic field inhomogeneity in the lower brainstem generated by dental alloys. However, 88% ot" the voxels were of sufficient quality to be analyzed. The statistically significant regional variations were a higher NAA/Cr ratio in the pons than in the medulla oblongata, higher Cho/Cr in the pons than in the mesencephalon and higher Cho/Cr in the cerebellar white matter than in the vermis. We conclude that 2D CSI of the brainstem, although technically delicate can be performed in most patients. �9 2001 Elsevier Science B.V. All rights reserved.

Kevwords: Brain metabolites; Posterior fossa; Chemical shift imaging: Magnetic resonance spectroscopy

1. Introduction 2. Subjects and methods

MR spectroscopy of the posterior tbssa has always been a difficult task to perform due to the relatively small size of this area and the proximity of bone structures. Thus, only single voxel spectroscopy has been performed in this area, using volumes of interest often spanning over multiple regions of the brainstem. This leads to an averaging of the metabolite concentra- tions in these different areas that may reduce sensitivity in the detection of focal diseases. To overcome this problem and to explore multiple regions of the poste- rior fossa, we have developed and applied a method to perform chemical shift imaging of this region. Addi- tionally, we have determined the metabolic ratios in five different voxels of interest in 30 healthy volunteers.

* Corresponding author~ Tel.: + 33-491-32-4469; fax: + 33-491-25- 6539.

E-mail address: patrick.cozzone@medecine.univ-mrs.fr (P.J. Coz- zone).

2.1. Subjec ts

We studied 37 healthy volunteers used as control subjects in three research protocols at our institution. These studies were approved by the Ethics Committee of La Tirnone hospital in Marseille. All subjects gave their informed consent. They had to be devoid of general or neurological disease. Lack of any neurologi- cal disease was further assessed by clinical examination performed by a neurologist and absence of any signal abnormality on F L A I R sequences. Seven subjects (19%) were excluded from the study, due to insufficient quality of the spectroscopy results, i.e. when less than three out of the five volumes of interest gave satisfac- tory signal quality. Quality of the spectra was judged adequate when the choline and creatine signals were clearly separated in at least half of their height. Data from 30 volunteers (17 female and 13 male) were thus

1352-8661/01/$ - see front matter !;~;' 2001 Elsevier Science B.V. All rights reserved. PII: S 1 352-866 !(0 1)001 35-1

128 D. Gahmaud et al./ Magnetic Resonance Materials in Physics, Biolo&v amt Medicine I3(2001) 127-133

analyzed. Mean age at examination was 29 years, rang- ing from 18 to 77 (S.D. = 14 years).

2.2. MR protocol

All examinations were performed on a 1.5 T Magne- tom Vision Plus system (Siemens, Erlangen). The study protocol included at least a sagittal Tl-weighted se- quence ( T R = 2 0 0 ms, T E = 10 ms), a diffusion- weighted sequence (axial or sagittal), a fast FLAIR sequence (TR = 8000 ms, TI = 180 ms, TE = 110 ms) and the CSI acquisition. Additional sequences were also perfomled, depending on the protocol to which the subjects belonged.

The CSI data are .obtained by using a home-designed acquisition weighted, hamming shape, pulse sequence (Fig. 1A) [1-4]. This acquisition weighted technique has been used successfully to perform CSI on the 31p

nucleus [5-7], on the 13C nucleus [8,9] or on 4D CSI (2D CSI, 2D COSY) [10] but, to our knowledge has never been used before in proton CSI. The water/lipid suppression is described by Tkac et al. [11]. All the

pulses are generated using the Matpulse software [12]. For this experiment, the field of view is 230 mm; pseudocircular encoding steps (21 x 211) are used, result- ing in a nominal spatial resolution of" 11 ram. The slice thickness is 15 mrn. The spatial resolution is defined as the width of the spatial response function at 64% of its maximum [10], resulting in an effective resolution of 22 mm and a voxel volume of 5.7 ml (Fig. I B). Due to the pseudocircular spatial ,/c sampling, voxel shape is cylin- drical. The TR is 1500 ms, the TE is 136 ms; 524 FID are acquired within a total acquisition time of 13 min. After positioning six outer volume suppression slices (OVS) used for lipid suppression, a Mapshim is con- ducted (courtesy of Siemens, Erlangen). A small correc- tion of the shim is perfl)rmed when necessary and an optimization of the water suppression is done, adjusting the 90 ~ RF pulse of the water suppression scheme. This procedure is routinely performed in less than 10 min.

The saturation slices are located as follows (Fig. 2A, B): �9 One antero-inFerior slice is located parallel to the

brainstem and must saturate the orbits and the clivus.

Fig. 1. (A) Pulse sequence used to obtain 2D CS1; (B) spatial response function.

D. Galanaud et al . / Magnetic Resonance Materials in Physics, Biology and Medicine 13 (2001) 127-133 129

Fig. 2. (A) Location of the outer volume saturation slices. The most important one is the anterior band that eliminates lipid signals from the clivus and orbits. (B) FLASH T2* image with saturation bands applied. The non-shaded areas are the anatomic structures l;'Jr which spectroscopic data are recorded. (C) Location of the five voxels analyzed in this study: 1, mesencephalon; 2, pons: 3, medulla oblongata; 4, vermis: 5, cerebellar nuclei and white matter. (D) Echo planar acquisition in a patient with numerous dental fillings leading to strong distortion of the magnetic field at the

level of the pons. Sagittal CSI will not yield usable data in this location.

�9 Five saturation slices eliminate the lipid signals from the superior and posterior part of the scalp.

�9 No lateral saturation slice is necessary. The OVS slices delineate a trapezoidal area of inter-

est including the brainstem, cerebellar vermis and white matter, the third and fourth ventricle, the corpus callo- sum, the cingular gyrus and parts of the internal fron- tal, parietal and occipital cortex (Fig. 2B).

ventricle contains mostly cerebellar white matter with negligible partial volume effects). Statistical analysis was performed on a personal computer using the StatView software (Abacus, CA). Ratios were com- pared using paired Student's t-test.

3. Results

2.3. Data processing

Data are processed using a proprietary software de- veloped on IDL (Interactive Data Language, Research System Inc., Boulder, CO). The t~llowing processing steps are applied to each acquisition: �9 Zero order phase correction. �9 Pixel selection on a conventional 256 x 256 MR

image; the software then performs a voxel shift in order to put the center of the voxel on this pixel.

�9 Additional water suppression on the selected spectra using HLSVD-MRUI FORTRAN code [13,14].

�9 Spectra fitting using AMARES-MRUI FORTRAN code [15]. Metabolic values have been recorded at five different

locations (Fig. 2C and Fig. 3): mesencephalon, pons, medulla oblonga, cerebellar cortex (vermis) and central cerebellum. This latter voxel, centered on the fourth

Eighty-eight percent (132/150) of the spectral data were judged of sufficient quality to be analyzed. Spectra of poor quality were most frequently located in the pons (,7 = 7) and medulla oblongata (t7 = 10) and were caused by mislocation of the outer volume suppression slices, leading either to lipid contamination or suppres- sion of the area of interest and artifacts caused by dental alloys (Fig. 2D).

The metabolic ratios are summarized in Table 1 and Fig. 4. The NAA/Cr ratio ranged from 1.20 (vermis) to 2.33 (pons). The Cho/Cr ranged from 0.85 (vermis)to 1.47 (pons). The NAA/Cho ratio ranged from 1.35 (cerebellar white matter) to 1.65 (mesencephalon). Stan- dard deviations (S.D.) for all metabolic levels were the highest in the medulla oblongata (NAA/Cr S.D. = 0.50 for 1.92 ratio), and the lowest in the cerebellum (NAA/ Cr S.D. = 0.13 for 1.20 ratio in the vermis). There were no significant gender- or age-related differences in the

130 D. Ga,~maud et al./Magnetic Resonance Materials in Physics, Biology and/vIedicine I3 (2001) I27-I33

metabolic ratios in any voxel. The only significant differences were: �9 A higher NAA/Cr ratio in the pons than in the

mesencephalon ( P = 0 . 0 4 ) and medulla oblongata (P = 0.007).

�9 A higher Cho/Cr ratio in the pons than in the

mesencephalon ( P = 0 . 0 0 3 ) and medulla oblongata (P = 0.002).

�9 A higher Cho/Cr ratio in the cerebellar white matter than in the vermis (P = 0.002).

�9 The other NAA/Cr , Cho/Cr and NAA/Cho ratios were not significantly different.

Fig. 3. Spectra obtained in the five voxels of interest. TR = 1500 ms, TE= 136 ms, voxel volume= 5.7 ml. Due to higher magnetic field homogeneity and better signal to noise ratio, the cerebellum yields spectra of better quality than the brainstem.

D. Gahmaud et ell./Magnetic Resonance Materials in Physics, Biotc, gy and Medicine 13 (2001) 127-133 131

Table 1 Metabolic ratios

Location NAA/Cr Cho/Cr NAA/Cho

Mesencephalon 2.11 (0.34) 1.29 (0.19) 1.65 (0.30) Pons 2.33 (0.50) 1.47(0.29) 1.62 (0.41) Med. oblongata 1.92 (0.50) 1.31 (0.27) 1.48 (0.34) Vermis 1.20 (0.13) 0.85 (0.12) 1.43 (0.19) Cerebellar WM 1.30 (0.22) 0.96 (0.14) 1.35 (0.22)

4. Discussion

MR spectroscopy of the posterior fossa is a challeng- ing task, especially in the brainstem, due to the small size of the area of ~ interest and its close relationship with bone structures, some of them particularly rich in

Fig. 4. Metabolic ratios and standard deviations in the five studied locations, mes, mesencephalon; med. obl., medulla oblongata: cer. WM, cerebellar white matter. *Statistically significant difference (P < 0.05).

air and lipids (the clivus and the mastoid). It has, however, a large potential interest, especially in the study of degenerative processes and tumors involving this brain area. MR spectroscopy of the brainstem has been performed as early as 1991 [16], and applied to the study of amyotrophic lateral sclerosis [17], multi system atrophy [18], glial tumors or Gulf war syndrome [19]. MR spectroscopy of the cerebellum has been used to study spino cerebellar ataxia [20], cerebellar ataxia [21] and alcoholism [22]. The technical difficulties of the metabolic studies of this area lead most authors to privilege single voxel spectroscopy over chemical shift imaging. Only Tedeschi et al. performed CSI of the posterior fossa [23] in an axial plane (angle 20 ~ from the orbito-meatal line). This technique is, however, of lim- ited interest in the study of this area, as only one location is accessible in the brainstem in each acquisi- tion (usually the mesencephalon or upper pons). Nor- mal metabolic ratios for this area are not readily available and authors have been using normal ratios of small numbers of control subjects for a given voxel location. As treatment of the data and voxel locations differ among studies, no meta-analysis of these data can be performed.

In order to study the maximum number of locations in the posterior fossa, we have developed a method to perform CSI on a medial sagittal plane. In addition to the whole brainstem, this method enables us to study in the same acquisition, parts of the vermis and white matter of the cerebellum, the corpus callosum, the cingular gyrus and the medial part of the frontal, parietal and occipital cortex (Fig. 2). This study plane has two main limitations: 1. To avoid fat contamination, the anterior volume

suppression must be located as close as possible to the brainstem. Compromises must sometimes be made, leading to suppression of the inferior part of the medulla oblongata or of the genum of the corpus callosum.

2. In some patients dental alloys create strong mag- netic field inhomogeneity, leading to unusable vox- els or examinations. This effect is maximal in the pons and medulla oblongata, the cerebellum being usually not affected. It is related neither to the number of treated teeth nor to their location. In a patient with numerous dental fillings, a smart way to determine if the CS! acquisition is worth per- forming is to record an echo planar image in the same plane as the CSI; in the presence of strong artifacts (Fig. 2D), metabolic data of the brainstem will be unusable and a single voxel acquisition should be preferred.

Standard deviations of the three metabolic ratios are highest in the lower part of the brainstem. This may be explained by the greater magnetic field inhomogeneity generated by the proximity of the bone structures and

132 D. Galanaud et al . / Magnetic Resonance MateriMs in Physics, Biology and Medicine 13 (2001) 127-I33

dental filling material inducing broadening of the spectra and less accurate signal fitting. The small size of the medulla oblongata, sometimes even dimin- ished by misplaced OVS slices, further reduces the signal to noise ratio and the accuracy of the meta- bolite quantification. On the contrary, the metabolic ratios of the voxels located in the large cerebellum, distant from bone structures, have the lowest standard deviations.

Comparison with previous spectroscopic studies of the posterior fossa in normal subjects is impaired by high variability in the acquisition techniques, echo times and post processing methods. Only Cwik et al. pert;armed a single voxel acquisition in the pons and medulla oblongata at similar echo times (120 ms) [24]. Their values for NAA/Cr and Cho/Cr ratios (respec- tively 1.95 and 1.58) are similar to those reported in this study; NAA/Cho in their control population (1.25) is somewhat lower than in our control subjects. This difference is, however, not significant. At a markedly longer echo time (272 ms), Tedeschi et al. found values similar to those reported here for the three metabolic ratios in a voxel including the upper pons and mesen- cephalon (NAA/Cr 2.68, Cho/Cr 1.43, NAA/Cho 1.87) [18]. A single voxel study performed in the same area at identical echo time (272 ms) reports significantly different metabolic ratios (NAA/Cr 6.65, Cho/Cr 3.96) [19]. Differences in acquisition sequence, voxel size and quantification methods may explain these vari- ations.

5. Conclusion

Chemical shift imaging of the brainstem and cerebellum using a sagittal plane is feasible in selected patients. This method enables a separate and simulta- neous analysis of the three regions of the brainstem, of the cerebellar vermis and white matter. The method may not be fully efficient in people with numerous teeth fillings. If the magnetic susceptibility artifacts are too large, a single voxel acquisition should usually be pre- ferred.

We found higher NAA/Cr and Cho/Cr in the pons than in mesencephalon and medulla oblongata. The other brainstem ratios were not significantly different. The cerebellar voxels differ only by a higher Cho/Cr in the cerebellar white matter. Dispersion of the values is the largest in the lower part of the brainstem, probably due to technical limitations.

In this first study we have applied this novel sequence to the exploration of the normal posterior fossa. Fur- ther work is underway to extend its use to patients presenting degenerative diseases or tumors of this area and to study other normal or pathological structures of the midline.

Acknowledgements

This work is supported by grants from Centre Na- tional de la Recherche Scientifique (UMR CNRS no. 6612 and Programme Imagerie Mddicale), Institut Uni- versitaire de France and Assistance Publique-H6pitaux de Marseille.

References

[10l

[111

[121

[13]

[14]

[15]

[16]

[171

[1] Mareci TH, Brooker HR. High-resolution magnetic resonance spectra from a sensitive region defined with pulsed field gradi- ents. J Magn Reson 1984;57:157-63.

[2] Mareci TH, Brooker HR. Essential considerations for spectral localization using indirect gradient encoding of spatial informa- tion. J Magn Reson 1991;92:229-46.

[3] Brooker HR, Mareci TH, Mao J. Selective Fourier transforma- tion localization. Magn Reson Med 1987;5:417-33.

[4] Hugg JW, Maudsley AA, Weiner MW, Matson GB. Compari- son of k-space sampling t\~r multidimensional MR spectroscopic imaging. Magn Reson Med 1996;36:469-73.

[5] Le Fur Y, yon Kienlin M. Acquisition weighted chemical shift imaging of the isolated rat heart at 11.7 Tesla. SMRM (third scientific meeting) and ESMRMB (twelfth annual meeting). Nice, France. 19-25 August, 1995.

[6] Hetherington HP, Luney DJE, Vaughan LJ, Pan JW, Ponder SL, Tscendel O, et al. 3D 31p spectroscopic image at 4.1 T. Magn Reson Med 1995;33:427-31.

[7] Pohmann R, von Kienlin MV. Acquisition-weighted 31p-csI of the human heart. Int. Soc. Magn. Reson. Med. 7th Meeting. Philadelphia 22-28 May, 1999.

[8] K6chenberger W, Heidenreich M, Chandrakumar N, Bowtell R. Kimmich R. Dynamic study of the movement of ~3C labelled sucrose in castor bean seedlings by MRI with cyclic J cross polarized transfer (CYCLCROP). Int. Soc. Magn. Reson. Med 6th, Sydney, 1998.

[9] Weidensteiner C, Lanz T, Horn M, Neubauer S, Haase A, Von Kienlin M. Three-dimensional ~3C-spectroscopic imaging in the isolated infarct rat heart. J Magn Reson 2000,!43:17-23. Von Kienlin M, Ziegler A, Le Fur Y, Rubin C, Decorps MRC. 2D-Spatial/2D-spectral spectroscopic imaging of intracerebral gliomas in rat brain. Magn Reson Med 2000;43:211-9. Tkac I, Starcuk Z, Choi IY, Gruetter R. In vivo ~H NMR spectroscopy of rat brain at 1 ms echo time. Magn Reson Med 1999;41:649-56. Matson G. An integrated program for amplitude-modulated RF pulse generation and re-mapping with shaped gradients. Magn Reson Imag 1994;12:1205-25. Pijnapel WWF, Van den Boogart A, De Beer R, Van Ordrnond D. SVD-based quantification of magnetic resonance signal. J Magn Reson 1992;97:122-134. Vanhamme L, Fierro RD, Van Huffel S, De Beer R. F a s t removal of residual water in proton spectra. J Magn Reson 1998; 132:197- 203. Vanhamme L, Van den Boogart A, Von Huffel S. Improved method for accurate and efficient quantification of MRS data with use of prior knowledge. J Magn Reson 1997;129:35-43. Detre JA, Wang ZY, Bogdan AR, Gusnard DA, Bay CA, Bingham PM, et al. Regional variation in brain lactate in Leigh's syndrome by localized I H magnetic resonance spectroscopy. Ann Neurol 1991;29:218-21. Pioro EP. MR spectroscopy in amyotrophic lateral sclerosis/mo- tor neuron disease. J Neurol Sci 1997;152(Suppl 1):$49-53.

D. Ga&naud et al./Magnetic Resonance MateriaLs" in Physics, Biology and Medicine 13 (200I) 127-133 133

[18]

[191

[20]

[21]

Tedeschi G, Litvan I, Bonavita S, Bertolin 9 A, Lundbom N, Patronas N J, et al. Proton magnetic resonance spectro- scopic imaging in progressive supranuclear palsy, Parkinson's disease and corticobasal degeneration. Brain 1997;120:1541- [22] 52. Haley RW, Marshall WW, McDonald GG, Daugherty MA, Petty F, Fleckenstein JL. Brain abnormalities in Gulf War syndrome: evaluation with ~H MR spectroscopy. Radiology [23] 2000;215:807-17. Mascalchi M, Tosetti M, Plasmati R, Bianchi MC, Tessa C, Salvi F, et al. Proton magnetic resonance spectroscopy in an Italian family with spinocerebellar ataxia type 1. Ann Neurol [25,] 1998;43:244-52. Davie CA, Barker GJ, Webb S, Tofts PS, Thompson AJ, Harding AE, et al. Persistent functional deficit in multiple sclero-

sis and autosomal dominant cerebellar ataxia is associated with axon loss [published erratum appears in Brain 1996; Aug;119(Pt 4):1415]. Brain 1995;118:1583-92. Martin PR, Gibbs SJ, Nimmerrichter AA, Riddle WR, Welch LW, Willcott MR. Brain proton magnetic resonance spec- troscopy studies in recently abstinent alcoholics. Alcohol Clin Exp Res 1995;19:1078-82. Tedeschi G, Bertolino A, Massaquoi SG, Campbell G, Patronas NJ, Bonavita S, et al. Proton magnetic resonance spectroscopic imaging in patients with cerebellar degeneration. Ann Neurol 1996;39:71-8. Cwik VA, Hanstock CC, Allen PS, Martin WR. Estimation of brainstem neuronal loss in amyotrophic lateral sclerosis with in vivo proton magnetic resonance spectroscopy. Neurology 1998;50:72-7.