7Li 2D CSI of human brain on a clinical scanner

MAGMA Magffletic Resonance Materials ha Ph)~ics, Bh')iol.D'ar~l Medichae ELSEVIER Magnetic Resonance Materials in Physics, Biology and Medicine 13 (2001) 1-7 www.elsevier.com/locate/magma 7Li 2 D CSI of human brain on a clinical scanner Franck Girard a,* Tetsuya Suhara a,b Takeshi Sassa b,c Yoshiro Okubo b,c Takayuki Obata a,b Hiroo Ikehira a, Yasuhiko Sudo a,b Masahisa Koga a, Hiroshi Yoshioka a, Katsuya Yoshida a a Division of l~ledical Imaging, National hTstitute of Radiological Sciences, Chiba, J~bpan b CREST (CoreResearch j'or Evolutional Science and Technot;)gy) of Japan Science and Technology Corporation, Tokyo, Japan Department of Neuropsychiatrv. Faculty of Medicine, Tokyo Medical & Dental University, Tokyo, Japan Received l l August 2000; received in revised form 19 March 2001; accepted 20 March 2001 Abstract Lithium salts have been widely used in the treatment of mood disorders, but the mechanism of action is still not clear. In this work, a methodology for two-dimensional Lithium-7 imaging on clinical systems is presented. The data were acquired using a phosphorus volume head coil that was re-tuned for the Lithium-7 frequency. A spectroscopic sequence was used to acquire the fi-ee induction decay (FID) alter volume excitation using a hard pulse. The results obtained on the head of patients undergoing lithium treatment (n = 7, 0.6 mEq/l average serum level) demonstrate that images of adequate signal to noise ratio (100:1) can be obtained in acceptable imaging times (55 min)using the proposed methodology. The distribution of 7Li appears uniform in the brains of the patients studied. ~-~ 2001 Elsevier Science B.V. All rights reserved. t~29.'words: MR imaging; Lithium-7; Psychiatry I. Introduction The treatments of bipolar affective disorders and recurrent episodes of mania intensively use lithium (Li) salts since the first proof of their effic~icy for these indications in 1949 [1]. However, although a great deal of work has been done on the subject, the exact mechanism of action of this ion in the brain is still very unclear [2-5]. Proposed mechanisms include brain neurotransmitter metabolism, effects on electrolytes and endocrine and receptor sensitivity [6]. This implies that even within the generally accepted therapeutic concentration range for serum lithium levels (0.5-1.0 mM in the US and Europe, 0.3-0.5 m M in Japan [7]), inter-individual differences are considerable, and both clinical response and adverse effects vary widely. A noninvasive means to measure brain lithium levels might aid in the treatment of lithium toxicity. * Corresponding author. Present address: GE Medical Systems, University Hospital Zfirich, Department of Magnetic Resonance, R'Mnistrasse 100, CH-8091 Zfirich, Switzerland. Tel.: + 41-1-2553052; t2ax: + 41-1-2554506. E-mail address: franck.girard@med.ge.com (F. Girard). The measurement of Li level in the brain has been conducted on rats and humans post-mortem [8] or during brain operations [9]. However, considerable inter-individual differences have been found in the results, up to 2.5-fold, ranging from 0.25 to 0.89 mM/1 [10] probably due to ion redistribution at death or during sample preparation [11]. N M R is known as a valuable non-invasive in-vivo tool and its applications to psychiatry are rapidly expanding. Monitoring of either endogenous metabolites containing 31p o r 13C, or exogenous agents (like psychoactive ~9F-containing drugs [12]) is becoming more and more common. The 7Li isotope is also a good candidate for N M R studies, as it has potentially interesting parameters: although it has a spin 3/2, and is a quadrupolar species, it shows relatively narrow lines and has a receptivity of approximately 27% of 1H, with a high natural abundance (92.6%). An accurate estimation of the 7Li concentration could therefore, eliminate the need for post-mortem analysis. This attractive nucleus suffers from three major drawbacks, hampering an accurate determination of brains' concentrations. Even though serum concentra- 1352-8661/01/$ - see front matter 9 2001 Elsevier Science B.V. All rights reserved. PlI: S 1 3 5 2 - 8 6 6 1 ( 0 1 ) 0 0 1 1 7 - X 2 K Girard et al./Magnetic Resonance M21teriat~ in Physics, Biology and Medicine 13 (2001) I - 7 tion is 100-fold higher than the typical serum concentrations of most polycyclic drugs, it is still very low from the MRS point of view: typical concentrations of the main proton metabolites range from 2 mM/1 (Cho) to 8-10 mM/1 (Cr) and 10-12 mM/1 (NAA) [13]. Imaging is hard to contemplate because of low concentration of 7Li. Second, lithium is a weak quadrupole that relaxes slowly. It may exhibit multiexponential relaxation behavior in-vivo. This behavior is due to the partitioning into different compartments, and to its spin, quadrupolar nature [14]. Third, the determination of T1 in-vivo will be difficult, implying that a monoexponential fitting instead of a multiexportential one will be used most of the time [7]. This very long T 1 is the most severe limitation for accurate quantification of Lithium in the brain. Despite these drawbacks, lithium-7 MR studies have been done, using first non-localized spectroscopy to detect the Li signal in-vivo [15]. Localized spectroscopy of Li has been performed [16], and first attempt to produce Spectroscopic Images (SI) has been made [17]. The SI approach allows one to obtain simultaneously spectra from slices (1D) or bars (2D) inside the organ, and to determinate the spatial distribution of metabolite, even in highly heterogeneous organs. It is well suited for the study of motionless organs as brain, despite the fact that SI is a very motion-sensitive technique. However, the SI data set on humans [17] was presented as an array of spectra, with a low-resolution matrix of 8 x 8. To our knowledge, no earlier work has presented 7Li images on human brain. In a clinical context, metabolic images are easier to interpret than an array of spectra, still providing the spectroscopic information. For this reason, this work has focused on the development of images from a spectroscopic data set, with a resolution high enough for a future clinical use. The clinical interpretation is the ultimate goal of this study. Therefore, the first part of the paper will present the work realized on phantoms, and discuss the validity of our sequence, while the second part will develop human applications. Images obtained with the SI technique on human brains will be shown. afterwards to superimpose the SI data too. Spectroscopic Imaging of Lithium was performed afterwards. As the resonance frequencies between 31p and 7Li at 1.5 T are very close (25.8 MHz for 31p vs. 24.8 MHz for 7Li), a 31p birdcage-type head coil tuned to the 7Li frequency was also used, enabling a coverage of the entire brain with a high signal to noise ratio. This coil was already in place when acquiring the scout images. No special disturbances could be seen on these images. Phantoms with various concentrations ranging from a concentration similar to that of human in-vivo (0.5 raM) to 1 mM were used. The phantoms consisted of flat-bottom bottles, of 025-cm height and 8-cm diameter. Different phantoms with different concentrations were prepared, each time with the same kind of bottles. These bottles were filled to approximately 3/4 of their heights, this corresponding to approximately 500 ml. A 10 Hz-width at halfheight was achieved on the water line, ensuring a good shimming of the phantoms. Seven patients with manic depressive disorders (four females and three males), their ages ranging from 21 to 83 years with mean age of 51 years, were studied. They were not hospitalized at the time of the experiment. Informed written consents were obtained from them prior to the examination. These patients were taking 800 mg per day of lithium four times a day at regular intervals. They were taking tablets of lithium (Taisho Pharmaceuticals, Japan), each tablet being dosed at 200 mg. The examinations were conducted in the morning or at the beginning of the afternoon between 2 and 3 h after the last ingestion of lithium. This ensured that brain lithium concentration was always comparable from one subject to another, as the brain lithium concentration has been found to undulate during the day [~8]. The mean concentration level of lithium in brain was gauged at 0.5 mEq/1 (ranging fi-om 0.3 to 0.7 mEq/1). This estimation was done on the basis of a blood analysis prior to the examination, on which a partition coefficient was applied [19]. The serum concentrations of our volunteers were between 0.5 and 1.2 mEq/1. These values were multiplied by a correction coefficient of 0.6 to estimate mean brain lithium concentration. Some of them also took other drugs (phenobarbital, chlorpromazine 12.5 mg, and flumitrazepam 2 mg). 2. Materials and methods 2.2. Methods 2.1. Materia,~ For both phantoms and patients examinations, the SI sequence was a Free Induction Decay (FID) sequence. A Philips Gyroscan horizontal whole-body MR imaging system operating at 1.5 T was used for all experiments. A whole-body coil was used for the scout 1H images. These images were used for positioning the slice of the brain under investigation, and were used 2.2.1. Phanwms studies The experimental parameters used for the SI of phantoms were as follows" TR = 250 ms, with a spectral bandwidth of 2000 Hz, field of view ( F O V ) = 400 ram. F. Girard et al./Magnetic Resonance Materials in Physics, Biology and Medicine I3 (2001) I - 7 3 1024 data points, four phase cycles, and no volume selection. The flip angle used for acquisition was 30 ~. The number of averages was chosen to obtain a total scan time of approximately 1 h, which was the maxim u m time we allowed ourselves for in-vivo examinations, and was either equal to 100 for a 16 x 16 matrix size or to 300 for an 8 x 8 matrix. 7'.1 was measured for the l m M phantom using an inversion-recovery experiment with 8 delays, ranging between 60 and 12720 ms (no linear progression in the delays), a TR of 6 s, 100 averages, leading to a total scan time of 1 h and 20 rain. There was no volume selection. tra in the spectroscopic presentation. Images for phantoms and human brain are presented in Figs. 1-4. After post-acquisition processing of the data on the Sun station, the images were viewed on a Silicon Graphics workstation Indigo-2 Extreme, with the Dr View software (Asahi Medical, Tokyo, Japan). This software permits the manipulation of images, allowing superimposition of different images. The ~H scout image was put beneath the Spectroscopic Image obtained to visualize the distribution of Lithium in the brain. 2.2.2. Patients stua'ies The first experiments on phantoms are presented in Fig. 1. Matrix sizes chosen were 32 x 32, like on the image shown, or 16 x 16. Even though a matrix such as 32 x 32 is not permitted for patients studies, due to the long acquisition time necessary with such low concentrations in-vivo, these phantoms data provide clear evidence of the good quality of our sequence. This data set took 1.5 h to obtain. Coronal and axial images were acquired. Fig. l a shows the axial image of a bottle u s e d as the phantom, and its Fourier interpolation (Fig. l b). The shape of the bottle is clearly visible in both acquisition directions, axial (Fig. la and b) and coronal (Fig. lc and d). Fig. 2 shows a spectroscopic image of the head of a patient, with an 8 x 8 matrix size. The superposition on a 1H scout image allows the visualization of the distribution of lithium in-vivo. Spectra corresponding to the volume of interest are also presented on the right part of the figure, showing the signal to noise ratio reached during those experiments. For patient examination, the orientation of the SI image was axial. Lithium SI was performed after a ~H scout image, used for visualization of the slice of the brain under investigation, and acquired with a wholebody coil. Experimental parameters were TR = 250 ms, 100 averages, 256 frequency-encoded points, FOV = 400 mm, a spectral bandwidth of 2000 Hz and a flip angle of 20 ~. The matrix sizes were either 12 • 12 zero-filled to 16 x 16, or 6 x 6 zero-filled to 8 x 8. This significantly reduced the scan time, due to the cut of the profiles with highest kx and ky values, although some residual information was lost. The total experiment time with those parameters was 55 min. There was no volume selection. A T~ correction must be applied to the data when a quantitative estimation of the concentration of 7Li invivo is planned. For this, T~ measurements have been performed on four patients. These patients were four males, their mean age being 48 years. Experimental conditions were as follow.s: inversionrecovery experiment with 8 delays, ranging between 60 ms and 12720 ms (no linear progression in the delays), a T R of 6 s, 100 averages, leading to a total scan time of 1 h and 20 min. There was no volume selection. 3. R e s u l t s 2.2.3. Post -processing The data were processed on a Sun workstation, with the software Xunspec provided by Philips. The processing sequence consisted was as follows: 9 t: Gauss-multiply, with a line broadening of 5 Hz, exponential multiplication with a line broadening of - 1 Hz, Fourier transform. 9 kx direction: Spatial apodization using a cosine filter, Fourier transform. 9 /'~vdirection: Spatial apodization using a cosine filter, Fourier transform. | t: modulus. With the data set obtained, an ira_age was created. A Fourier interpolation was applied. This procedure is fully equivalent to the calculation of intermediate spec- Fig. 1. 32 x 32 7Li Spectroscopic image of a phantom. Coronal (lc) and axial (la) slices are presented together with the respective Fourier smoothing (ld & lb) of these data. TR was 150 ms (only for this experiment) with a spectral width of 2000 Hz, a flip angle of 90~ 48 measurements tor a total scan time of 1 h 35min. Lithium concentration in this phantom was equal to 10 mM. 4 F Girard et al./Magnetic Resonance Materials in Physics, Biology and Medicine 13 (2001) I - 7 Fig. 2. 8 x 8 spectroscopic 7Li image image of human head, The in-vivo concentration of lithium is approximately equal to 0.6 mEq/1. The right part of the image shows spectra corresponding to the voxels indicated by the arrows. Experimental parameters for those experiments were TR 250 ins, a flip angle equal to 20~ 300 averages, no volume selection, and a spectral width of 2000 Hz. The total experiment time was 55 min. Fig. 3 represents the same superposition between 7Li S I image and ~H scout image as in Fig. 2, but for a 16 x 16 matrix size. The grey scale was ranging between 0 to > 0.2 m M (black) and 0.8 to > 1 m M (white). The three grey levels visible in the image c o r r e s p o n d to ranges from 0.2 to > 0.4, 0.4 to > 0.6, and 0.6 to > 0.8 raM. The resolutions reached are roughly equal to 4.3 cm in the case of 12 x 12 zero-filled to 16 x 16 matrix, and of 8.6 cm for a 6 x 6 matrix zero-filled to 8 x 8. Typical in-vivo TI curves are shown in Fig. 4. T1 m e a s u r e m e n t s on our patients gave values between 4.0 +_ 0.1 s and 4.3 +_ 0.5 s, in good accordance with earlier published data [17,19]. Each value of the intensity of each 7Li peak obtained after F o u r i e r t r a n s f o r m was c o m p u t e d versus time, and the points obtained were then fitted to a t h r e e - p a r a m e t e r fit, following the m o d e l 'fit = K0 + K1 x e x p ( - K2 x Tl)'. showing the relative c o n c e n t r a t i o n is the ultimate goal, a metabolic image is m o r e easily understandable than an array of spectra. Lithium 7 N M R can be this potential tool, and SI can help to the u n d e r s t a n d i n g of the distribution of the lithium. It should be c o m p a r e d 4. Discussion A l t h o u g h m a n y studies have been carried out in-vivo on patients treated with Lithium salts, the exact mechanism of action of this ion for the t r e a t m e n t of psycho atTective diseases remains largely u n k n o w n [6,11,15]. T h e r e is a need for an in-vivo m e a s u r e of Li distributiom and when a clinical i n t e r p r e t a t i o n of the data Fig. 3. 16 x 16 spectroscopic image. The superposition with the proton scout image has been done as described in the text. Experimental parameters are the same as for the 8 x 8 matrix, except the number of averages reduced from 300 to 100. F. Girard et al./Magnetic Resonance Materials in Physics, Biology and Medicine 13 (2001) I - 7 0.15 ..=. o.mj 1 - 0.08 ~ o//~ ///, j ..........~............... _J.............................! 0.00 1 OxlO" 2 4 6 8 ..... I 10 ! 12 Time (sl Fig. 4. Inversion-recovery T1 curves measured on patients. Experimental parameters were 8 delays, ranging between 60 and 12720 ms, a T R of 6 s, 100 averages, leading to a total scan time of 1 h and 20 min. There was no volume selection. with single-voxel experiments (STEAM-PRESS), which allow shimming o n a volume of interest, and give a numerical concentration. However, if N scans are needed to obtain one spectrum with a single voxel technique, the SI approach allows N spectra from N voxels to be obtained in the same experiment time. Due to the long acquisition time inherent to SI to obtain a sufficient signal to noise (S/N) ratio, it is difficult to achieve a resolution higher than the one obtained with the 16 x 16 matrix on this system with patients. This coarse resolution ensures, however, a sufficient S/N ratio in each voxel (Fig. 2) and leads to an acceptable localization, as shown on the images. Indeed, very few signal can be seen outside the position of the phantoms or the head. The remaining pixels outside the head on human data are probably arising from the head itself, but are seen outside because of the low resolutions used. The gray level distribution of the voxels placed within the brain lie between those seen for the 0.5 and the 1 mM phantoms. The differences in gray level value between the individual voxels indicate that the concentration of Li, excluding one or two voxels, is rather constant. Our observations are consistent with the ones of Komoroski et al. who also reports few variations of Li concentration in an earlier paper [16]. The quantification of the signal obtained is an important point. Many parameters should be taken into account when considering the data from a quantitative point of view. A quantitative analysis requires knowledge of whether or not characteristics of lithium can yield a reduced signal intensity. Indeed, under conditions of restricted mobility, quadrupolar nuclei with spin greater than 1 can have reduced signal intensities [16,19]. Lithium is a quadrupolar nucleus with a spin higher than 1/2 (3/2), but it fortunately exhibits narrow lines. However, a loss of signal due to this quadrupolar interaction is predictable. Gullapalli et al. [6] made an evaluation of this loss, on red blood cells. They t:ound that the signal was less visible at low concentrations (1 m M , which is roughly the highest concentration reachable in-vivo in Japan), with an intensity of the signal o-,>, of the total signal. The approximately equal to oAo.. quadrupolar 'character' of this ion should be taken into account and will lead to a reduction of the signal observed. The data shown here are heavily T~-weighted because of the very short T R used. A correction must be applied to compensate for this. The knowledge of T~ is therefore essential. Few earlier studies determined the value of T~ of lithium in-vivo [17,19] and proposed values range between 3.4 s and 7 s. These studies proposed m_ono or multiexponential T~ values, on humans and animals (puppies and rats). Renshaw et al. [21] measured those T~ values in the puppy brain, and applied a multiexponential treatment of the data. They proposed respective values of 3.5 s for the fast component (30%), and 6.6 s for the slow component (70%). These authors are, however, the only ones who found multi-compartment T~ values, and others just proposed one T~ relaxation time, on rats [11] or on patients [17]. Kushnir et al. [20] applied both mono and multiexponential treatment of the data, and did not find significant differences. In this study, multiexponential treatment of the data has also been done, but did not bring significant modification of our data, and we decided to keep a monoexponential value for our T~. The values we measured are close to those already proposed by other authors (4 vs. 4.6 s[l 7] or 3.5 s [20]). Contribution from the muscles can lead to an overestimation of the T~ value. However, as pointed out earlier, the use of a volume head coil should ensure that this contribution is not dominant. Renshaw et al. calculated the ratio of saturation factors of Lithium [15], using the following expression: Ratio = (1-exp(-4/STl) = 0.58 (1 - e x p ( - 4/BrT1) where S T~ is the T~ value of the Lithium solution (13.1 s), and BrT~ is the T~ value of the puppy brain (6.6 s). They found a ratio value equal to 0.58, but one should remember they made this estimation at 1.8 T. However, this value has been commonly accepted and used by various authors [16,22]. We preferred a calculated value of 0.42, keeping a value of 13 s for the S TI factor, but with a value of 4.14 s for the BrT~ ~ctor, corresponding to the mean value of the different T~ measured on our patients. The saturation factors of water measured at 2 different TR, were equal to 0.95 for the brain, and 0.82 tar the phantom. The ratio of saturation/"actors of water in the phantom versus the brain was established at 0.88, very close to the value of 0.89 proposed by Kato et al. [22] in their calculations. A few problems should still be answered properly before using this method for quantification from a routine point of view. This study does not distinguish between intracellular and extracellular components of lithium. It is impossi- 6 F. Girard et al./M'agnetic Resonance Materiab in Physics, Biology and Medict'ne 13 (200I) I - 7 ble to distinguish all components, but each compartment exhibits a different lithium concentration, and thus, different relaxation times. So far, the signal to noise ratios reached are not high enough to distinguish the different compartments. It is also very hard to be completely sure that the signal obtained only comes from the brain. Contamination may come from bones or muscles. However, it is unlikely that the Li signal comes from other places: bones' components exhibit such low TI values that they are normally non-observable by standard N M R techniques. The contribution of muscles to the signal is present, as noticed by Gonzales et al. [23], and not negligible, but small. A small correction will therefore, be needed when doing quantitative measurements. Moreover, Kato et al. performed 3~p experiments [22] showing the signal they obtained mainly came from the brain, and not from muscles. Finally, only a negligible concentration of 7Li is present in lipid tissues [24]. Another possible evolution of this work is to perform localized Spectroscopic Imaging of the lithium signal in the human brain. Combining localization methods with encoding schemes can help restricting the volume over which the Bo field is a~usted, and reduce places where large variations in the magnetic susceptibility can appear. This can give an even more accurate idea of the distribution of the lithium in-vivo. Work is currently in progress to realize this. The problem with these experiments is the low signal to noise ratio obtained, due to the localization scheme. This drawback could probably be overcome with higher fields. The gain in signal expected could compensate for the loss of signal due to the localization scheme. The images presented here show the feasibility of obtaining SI data in-vivo within a reasonable scan time. The repartition of the lithium in the brain can be monitored, and this can eventually yield to a better understanding of the mechanism of lithium's action in the brain. Acknowledgements FG is a fellow from the Science and Technology Agency of the Japanese Government (EU-STA 95). The authors would like to express their gratitude to Dr R. Lamerichs and Dr P. Luytens (Philips Netherlands) for their help with the 7Li patch software. Many thanks to Dr V. L. Marcar, Zfirich, for helpfial comments on the manuscript. References [1] Cade JFJ. Lithium salts in the treatment of psychotic excitement. Med J Aust 1949;36:349-52. [2] Kennedy ED, Challiss RA, Nahorski SR. Lithium reduces the accumulation of inositol polyphosphate second messenger following cholinergic stimulation of cerebral cortex slices. J Neurochem 1989;53:1652-5. [3] Harvey BH, Carstens ME, Taljaard JJF. A novel hypothesis for the psycho-modulating effects of lithium: the role of fatty acids, eicosanoids and sub-cellular second messengers. Med Hypoth 1990;32:51-8. [4] Stern TA, Lydiard RB. Lithium therapy revisited. Psychiatr Med 1987;4(1):39-68. [5] Wood AJ, Goodwin GM. A review of the biochemical and neuropharmalogical actions of lithium. Psychol Med 1987;17(3):579-600. [6] Gullapalli RP, Hawk RM, Komoroski RA. A 7Li NMR study of visibility, spin relaxation, and transport in normal human erythrocytes. Magn Reson Med 1991;20:240-52. [7] Kato T, Takahashi S, Inubushi T. Brain lithium concentration measured with lithium-7 magnetic resonance spectroscopy: a review. Lithium 1994;5:75-81. [8] Francis R J, Traill MA. Lithium distribution in the brains of two manic patients. Lancet 1970;2:523-4. [9] Terhaag B, Scherber A, Schaps P, Winker H. The distribution of lithium into cerebrospinal fluid, brain tissue and in man. Int J Pharmacol Sci 1978;16:333-5. [10] Sachs GS, Renshaw PF, Lafer B, Stoll AL, Guimaraes AR, Rosenbaum JF, Gonzalez RG. Variability of brain lithium levels during maintenance treatment: a magnetic resonance spectroscopy study. Biol Psychiatr 1995;38:422-8. [II1 Ramaprasad S, Newton JEO, Cardwell D, Fowler AH, Komoroski RA. In-vivo VLi NMR Imaging and localized spectroscopy of rat brain. Magn Reson Med 1992;25:308-18. [121 Durst P, Schuff N, Crocq M-A, Mokrani M-C, Macher J-P. Noninvasive in-vivo detection of a fluorinated neuroleptic in the human brain by 191: nuclear magnetic resonance spectroscopy. Psychiatr Res 1990:35(2):107-14. Provencher SW. Estimation of metabolites concentrations from [13] localized in-vivo proton NMR spectra. Magn Resort Med 1993;30(6):672-9. [14] Renshaw PF, Sachs GS, Gonzalez R. In-vivo MRS measurement of lithium levels in brain. In: Nasraltah HA, Pettegrew JW, editors. NMR Spectroscopy in Psychiatric Brain Disorders. Washington, DC: American Psychiatric Press, 1995:179-98. Renshaw PF, Wicklund S. In-vivo measurement of Lithium in [15] humans by Nuclear Magnetic Resonance Spectroscopy. Biol Psychiatr 1988;23:465-75. [16] Komoroski RA, Newton JEO, Walker E, Cardwetl D, Jagannathan NR, Ramaprasad S, Sprigg JR. In-vivo NMR spectroscopy of Lithium-7 in humans. Magn Reson Med 1990;15:347-56. [17] Komoroski RA, Newton JEO, Sprigg JR, Cardwell D, Mohanakrishnan P, Karson CN. In-vivo Nuclear Magnetic Resonance study of Lithium pharmacokinetics and chemical shift imaging in psychiatric patients. Psychiatr Res: Neuroimaging 1993;50:67-76. Plenge P, Stensgaard A, Jensen HV, Thomsen C, Mellerup ET, [181 Henriksen O. 24-Hour lithium concentration in human brain studied by Li-7 magnetic resonance spectroscopy. Biol Psychiatr 1994;36(8):511-6. [19] Gyulai L, Wicklund SW, Greenstein R, Bauer MS. Measurement of tissue lithium concentration by lithium magnetic resonance spectroscopy in patients with bipolar disorders. Biol Psychiatr 1991;29:1161-70. [201 Kushnir T, Itzchak Y, Valeski A, Lask M, Modai I, Stokar S, Navon G T1 relaxation times and concentration of lithium-7 in the brain of patients receiving lithium therapy. NMR Biomed 1993;6:39-42. F. Girard et al./Magnetic Resonance Materials in Physics, Biology and zl4edicine 13 (2001) I - 7 [21] Renshaw PF, Haselgrove JC, Bolinger L, Chance B, Leigh JS. Relaxation and imaging of lithium in-vivo. Magn Reson Imaging 1986;4:193-8. [22] Kato T, Takahashi S, Inubushi T. Brain lithium concentration by VLi and 1H magnetic resonance spectroscopy in bipolar disorder. Psychiatr Res: Neuroimaging 1991;45: 53-63. 7 [23] Gonzalez RG, Guimaraes AR, Sachs GS, Rosenbaum JF, Garwood M, Renshaw PF. Measurement of human brain lithium in-vivo by MR spectroscopy. AJNR 1993;14:1027-37. [24] Ozawa H, Nozu T, Aihara H, Akiyama H, Sasajima M. Metabolic fate of lithium salts by single and repeated administration and the behavioral effects. Folia Pharmacol Japonica 1976;72:433-43.