Chemical Imaging: Magnetic Resonance Spectroscopy: The Basics



Magnetic Resonance Spectroscopy enables the detection and quantification of a wide range of cerebral metabolite compounds in vivo. Proton (1H) magnetic resonance spectroscopy in particular holds great potential for characterizing the pathophysiology of disease states and may provide important biomarkers particularly at the level of inhibitory and excitatory amino acid neurotransmitter systems. However, the field of in vivo magnetic resonance spectroscopy contains a multitude of concepts that may be unfamiliar to many readers. The present chapter is an overview of the basic fundamentals of proton magnetic resonance spectroscopy data acquisition and post processing strategies, which are currently are employed by investigators worldwide for characterizing the metabolic profile of the central nervous system and its associated disease states.


Magnetic Resonance Spectroscopy Magnetic Resonance Spectroscopic Imaging Amino Acid Neurotransmitter Central Nervous System Drug Magnetic Resonance Spectroscopy Data 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Magnetic resonance spectroscopy (MRS) is a noninvasive medical imaging modality that holds great promise for assisting the central nervous system (CNS) drug discovery process. MRS measurements yield spectral data that ultimately can be analyzed and interpreted to provide unique information regarding disease processes, pathophysiology and drug response in vivo. 1H-MRS investigations are performed at many institutions although an attractive aspect of in vivo MRS is the potential range of other nuclei that can be exploited. The specific MR-active nucleus under investigation depends on the instrumental hardware available as well as the biological application and questions to be addressed. Table 1 summarizes the biologically important MRS-active nuclei that are of particular interest to drug discovery, and gives nucleus-specific MR properties such as the gyromagnetic ratio (γ) and natural abundance. The MR-sensitivity of a nucleus is directly proportional to both of these variables. The table also provides examples of “biological relevance” for each nuclide and several biochemical compounds are listed for each potential application.
Table 1

Summary of MR-active nuclei relevant to biological investigations in vivo. Biological relevance and molecular examples are provided for each isoptope

Isotope (abbreviation)

Gyromagnetic ratio

(γ; MHz/T)

Natural abundance (%)

Biological relevance

Molecular examples


Proton (1H) and Carbon (13C)





Amino acids/




Cellular Energetics

Membrane and

lipid metabolism

Aspartate (Asp)

γ-aminobutyric acid (GABA)

Glutamate (Glu)

Glutamine (Gln)

Glycine (Gly)

N-acetylaspartyl glutamate (NAAG)

Ascorbic acid (Asc)

Glutathione (GSH)

N-acetyl aspartate


Adenosine-triphosphate (ATP)

Creatine/Phosphocreatine (Cr/PCr)

Lactate (Lac)

Glucose (Glc)

Pyruvate (Pyr)

Succinate (Suc)

Choline (Cho)

Glycerophoshocholine (GPC)

Phosphocholine (PCho)

Ethanolamine (Etn)

Phosphoethanolamine (PEtn)

Lipids (Lip)


Phosphorous (31P)



Cellular Energetics

Membrane and

lipid metabolism

ATP, PCr, PCho,

Inorganic phosphate (Pi)

Glycerophosphoethanolamine (GPEtn), PEtn, Glycerophosphocholine (GPC)

Membrane phospholipid (MP)






Drug-detection and metabolism

19F-containing agents

(e.g., fluoxetine, fluvoxamine)






Drug quantification

7Li-containing agents

(e.g., lithium carbonate)


The proton (1H) is by far the most widely studied nucleus for both preclinical and clinical MRS investigations in vivo. The 1H nucleus has the highest sensitivity owing to its high γ and high natural abundance, and 1H-MRS can be performed using the hardware typically supplied with preclinical and clinical MRI systems such as 1H-tuned radiofrequency (RF) amplifiers and coils. In addition, the 1H nucleus is ubiquitous in nature and many biologically relevant organic compounds contain proton sub-groups that are detectable via 1H-MRS. Molecular species that are pertinent to CNS drug discovery and are that are detectable using 1H-MRS include the amino acid neurotransmitters Glu, Gln, GABA and Gly (see table 1 for metabolite acronyms). The neuroscience product pipelines of major pharmaceutical companies currently include glutamatergic, glutaminergic, GABAergic and glycinergic agents, which are under evaluation for treating a wide variety of neurologic disease and psychiatric disorders including affective disorders, Alzheimer’s disease, anxiety, pain, schizophrenia, stroke and substance abuse.

The importance of MRS studies involving other types of nuclei should not be discounted. 31P-MRS, for example, can provide unique information regarding NTP concentration and cellular energetics. Membrane phospholipids and their precursors such as PCho and PEtn can also be measured using 31P-MRS and these types of studies hold great potential for monitoring tumor phospholipid metabolism in response to therapeutic intervention. Although hindered by low natural abundance and low sensitivity, 13C-MRS lends itself nicely to labeled substrate infusion studies, which are performed at a handful of sites worldwide. Administration of 13C-enriched glucose or acetate leads to a downstream 13C-labeling of glutamate and glutamine at specific carbon positions, and tricarboxylic acid enzymatic rate constants can be extracted. Several CNS drug compounds, such as the selective serotonin reuptake inhibitors (SSRIs) paroxetine and fluoxetine, are fluorinated and thus detectable using 19F-MRS techniques. Localized 19F-MRS measurements could enable the in vivo monitoring of drug localization and catabolism and help elucidate drug pharmacokinetic parameters including half life and clearance.

All medical imaging modalities suffer from drawbacks and limitations and MRS is no exception. MRS is an insensitive measurement technique and a brain metabolite must be present at a sufficient level, typically in the millimolar (mM) concentration range, to be detected and quantified. Low-sensitivity measurements suffer from a low signal-to-noise ratio (SNR) and signal enhancement is crucial for accurate quantification. For MRS measurements SNR is typically enhanced by averaging packets of data recorded in discrete periods over time (signal averaging) and/or increasing the size of the MRS volume element (voxel). Unfortunately, these two potential methods compromise spatial and temporal resolution, respectively. As outlined later in this chapter, quantification of 1H MRS data is hampered by a very low spectral resolution particularly at magnetic field strengths (B0) associated with clinical MR systems. Another disadvantage of MRS is that resulting data must be processed offline and specialized software often is required for reliable quantification. Also, purchase of additional hardware such as dedicated radiofrequency (RF) coils and RF amplifiers is essential for MRS measurements involving nuclei other than the proton. These types of hardware along with the more useful (and complex) MRS techniques are not generally provided as standard on MRI/MRS scanners and their correct implementation requires some degree of local MRS expertise. This expertise is also critical for establishing reliable quantification protocols and ensuring meaningful interpretation of spectral data.

Nevertheless, MRS benefits from a range of advantages unparalleled by many other forms of medical imaging. MRS measurements do not involve delivery of ionizing radiation to biological tissue in contrast to other chemical imaging methods such as positron emission tomography (PET) and single photon emission computed tomography (SPECT). Therefore, MRS measurements can be incorporated into repeated measures study designs and are safe for investigations involving pediatric populations. A primary advantage of 1H MRS as applied to CNS drug discovery is its ability to quantify several amino acid neurotransmitters, coagonists and their derivatives. Therefore, MRS has the potential to establish the mechanism(s) of novel drug candidates at the level of individual neurotransmitter systems. Such pharmacodynamic responses might serve as important biomarkers of functional effects on amino acid neurotransmitter systems and provide crucial data for determining the biological mechanisms underlying treatment response. A large amount of literature has emerged demonstrating that MRS-detectable between-group differences do exist for many CNS disorders and MRS-detectable drug effects are beginning to emerge for several pathophysiologies.

Proton Magnetic Resonance Spectroscopy

In vivo 1H-MRS can be implemented and performed at most MRI centers and the types of metabolites that can be detected through 1H-MRS will be most applicable to CNS drug discovery. The main objective of this chapter is to give a brief and highly conceptual introduction to some of the fundamental aspects of in vivo 1H MRS. The reader is directed to other theoretical sources that provide more complete descriptions of MRS concepts (Salibi and Brown 1998; de Graaf 2007). This chapter initially describes the basic steps that are required for the execution of 3D localized water solvent-suppressed 1H-MRS experiments, and these sections are intended to introduce the uninitiated reader to some of the technical jargon often encountered in the MRS literature. The basic physical features of a 1H-MRS spectrum including chemical shift, signal amplitude and scalar spin–spin coupling is then characterized using real data acquired from human brain in vivo. The same dataset is used to highlight some of the key metabolite proton resonances that can be detected using standard 1H-MRS methods and examples of MRS spectral fitting algorithms are briefly described. Importantly, this chapter will outline several alternative 1H-MRS methods that can be used for resolving amino acid neurotransmitter resonances that are difficult to isolate using standard approaches.

In Vivo 1H-MRS Measurement

MRI and Shimming. The acquisition of low-resolution MR images is an important first step to any 1H-MRS examination, as these can be used to assess and ensure optimal subject positioning within the RF coil and MRI scanner. The main magnetic field (B0) homogeneity over the sample (subject) then can be optimized through a process known as B0 shimming, which involves passing relatively small current through small electromagnets or “shim coils” that are carefully positioned within the magnet bore at the scanner installation. The shim coils provide small-amplitude auxiliary magnetic fields, which are iteratively adjusted using computer controlled algorithms in order to compensate for B0 inhomogeneities. A high degree of B0 homogeneity across the sample is particularly crucial for MRS measurements as this resonance line width and enhances spectral resolution.

Spatial Localization. In vivo 1H-MRS measurements require a form of spatial locali­zation such that spectral data can be recorded from a preselected and well defined region-of-interest (ROI). Historically, localization was crude and achieved by positioning the so-called “surface” RF coils superficially to the ROI. Nowadays it is much more common to utilize B0 field gradients for spatial localization and the approaches can be broadly classified into two common types: single-voxel and multiple voxel 1H-MRS. Single-voxel 1H-MRS methods achieve 3D localization through the sequential execution of three RF “pulses” of finite bandwidth, each of which is executed in the presence of a linear B0 gradient applied along one of the three orthogonal axes. Each RF pulse – gradient combination selects a slice with the thickness being proportional to the RF pulse bandwidth/B0 gradient strength. The interpulse timings together with the RF flip angles constitute what are referred to as “pulse sequences” and determine whether a spatially localized stimulated echo or a double spin echo originates from the intersection of the three selected slices. The two sequences routinely used for 1H-MRS applications are the stimulated echo acquisition mode (STEAM; (Frahm et al. 1989)) and point-resolved spectroscopy (PRESS; (Bottomley 1987)) methods. The advantage of the STEAM sequence is that a low echo time can be achieved potentially giving access to increased metabolic information. The PRESS sequence on the other hand benefits from increased inherent measurement sensitivity due to the acquisition of a spin echo. Whether STEAM or PRESS is used, the echo signals are detected and digitized in the form of free-induction decays (FIDs), which yield spectral data following Fourier transformation (FT).

Multiple voxel 1H-MRS methods such as magnetic resonance spectroscopic imaging (MRSI) are usually examples of B0 gradient-based localization methods but combine spectroscopic and imaging concepts to yield multiple localized spectra. In MRSI a single slice is usually selected along one of the orthogonal axes through the combination of a RF pulse applied with a linear B0 gradient. Pulsed B0 gradients subsequently applied along the two other orthogonal axes act to phase encode the nuclear spins along those dimensions. The phase encode process is repeated in successive acquisitions until the desired spatial resolution has been achieved and a grid of multiple spectra are afforded following the necessary spatial and spectroscopic FTs.

RF Pulse Power Calibration. Once the single-voxel element or a MRSI slice has been positioned within the ROI the RF pulse power is automatically optimized, which ultimately ensures reliable peak quantification. “Local” B0 shimming then should be performed over the ROI until the unsuppressed water signal line width is minimized. Finally, a method for solvent water suppression is activated and calibrated until the water resonance is reduced to a level where metabolite resonances can be readily quantified. Water suppression schemes typically employ a train of frequency-selective RF pulses that are applied immediately prior to the spatial localization scheme. The RF pulses are tuned to excite the water resonance with minimal excitation of surrounding metabolite resonances, and each pulse is followed by a strong B0 gradient that acts to dephase or “crush” the generated transverse water magnetization.

MRS Sequence Parameters. MRS is an insensitive measurement and the detected signal is weak. For in vivo measurements this problem is made worse by the low metabolite concentrations to be measured (mM) and the electronic thermal noise associated with MRI scanner hardware. It can be shown that signal-to-noise ratio (SNR) can be increased by averaging FIDs from a number of excitations (NEX). This is because the genuine MR signals coverage linearly whereas the noise increases by √NEX. Therefore, quadrupling NEX increases the observed SNR by a factor of two. The scanner operator is responsible for setting the NEX together with the time between NEX (repetition time, TR) and the time between excitation and onset of FID detection (echo time, TE). The product of TR and NEX gives the total MRS measurement, which should be kept relatively short for in vivo applications. In general, short TE’s give access to increased metabolite information at the expense of increased spectral complexity (see below).

The In Vivo Brain 1H-MR Spectrum

Figure 1a shows an axial T1-weighted MR image recorded at 4.0 T from a healthy adult human subject and the red box corresponds to a section of a 2 × 2 × 2 cm3 MRS voxel positioned predominantly within the predominantly gray matter of the parietal occipital cortex. Figure b shows a 1H-MRS spectrum that was obtained from the same MRS voxel without solvent water suppression. The resonance line observed in Fig. 1b corresponds to a single resonance frequency and is solely attributable to tissue water protons. If the proton nuclei within different molecular species (e.g., cerebral metabolites) exhibited the same identical resonance frequency as water, then a single MR peak would be observed and little information could be extracted from the resulting spectral data.
Fig. 1 (a)

An axial high-resolution T1-weighted MR image recorded from a healthy adult human head at a field strength of 4.0 T. The MR image shows a section of a 2 × 2 × 2 cm MRS voxel (red box) positioned within predominantly gray matter of the parietal-occipital cortex. (b) A water unsuppressed 1H-MRS spectrum recorded from the voxel shown in Fig. 1a where only a single water proton resonance is seen at 4.7 ppm. (c) A water-suppressed 1H-MRS spectrum recorded from the same voxel (PRESS; TR = 2,000 ms, TE = 30 ms, NEX = 256). The vertical scaling was increased 80-fold compared to the corresponding water unsuppressed dataset and metabolite resonances are observable between the 0.5 and 4.2 ppm chemical shift range. (d) LCModel analysis of the water-suppressed 1H-MRS dataset presented in Fig. 1c. See text for details

Chemical Shift and the Chemical Shift Scale. Fortunately, the resonance frequency of nuclear spins is highly sensitive to the local electronic environment. In some situations, electrons present within the local molecular structure act to shield the nuclear spins from the main B0 field and the effective field experienced by the nucleus under investigation is effectively reduced. If a nuclear spin experiences a reduced B0 field then it will exhibit a slightly lower MR frequency. Electronic ‘deshielding’ can also occur where electronegative atoms or functional groups within a given molecular species attract electron clouds away from a nuclear spin. In that situation, the local B0 will be enhanced leading to a slightly higher MR frequency. These effects act to produce a spread of MR frequencies for a given nuclide and the shift effect due to chemical environment is known as the chemical shift. The x-axis scale used for displaying MRS spectra is termed the chemical shift or ppm scale and is convenient for several reasons. The resonance frequency, ω 0, of the proton nucleus is directly proportional to the static B0 field as governed by the fundamental Larmor equation, ω 0 = γB0 (units: rad.s−1). For example, at static B0 field strengths of 1.5 and 4.0 T, the proton resonance frequency would be approximately 64 and 170 MHz, respectively. By convention the chemical shift scale utilizes the same idea for a basic reference, tetramethylsilane (TMS) or related analogs, whose four methyl proton groups would give rise to a single resonance frequency (ω ref) referenced to 0 ppm. The chemical shift (σ) of other proton nuclei is then calculated using a simple formula given by: σ = 106(ω − ω ref/ω ref). For example, water has a higher MR frequency to TMS and the difference is approximately 300 and 800 Hz at B0 field strengths of 1.5 and 4.0 T, respectively. Hence, the water proton resonance resides at a chemical shift of 4.7 ppm regardless of the applied B0. Note that when presenting MRS data the chemical shift scale (and frequency) increases from left-to-right.

Cerebral Metabolite 1 H-MRS Resonances. A water-suppressed 1H-MR spectrum recorded from the voxel displayed in Fig. 1a is then displayed in Fig. 1c. Note the increased spectral complexity owing to the additional multiple metabolite resonances observed between the 0.5 and 4.2 ppm chemical shift range. The water resonance has been attenuated to a level where access to neurochemical information becomes available and the figure clearly demonstrates the problem of severe spectral overlap associated with in vivo 1H-MRS applications. The spectrum presented in Fig. 1c is dominated by large-amplitude metabolite resonances positioned at 2.0, 3.0, and 3.2 ppm, and those peaks correspond to methyl (CH3) protons of NAA, the sum of PCr and Cr (total Cr) and Cho-containing compounds, respectively. The peak centered at 3.9 ppm is attributable to the methylene (CH2) proton nuclei of total Cr. Being enriched in viable neurons, NAA is generally accepted as a marker of neuronal viability and a growing body of evidence also supports NAA’s role as a molecular water pump. 1H-MRS NAA decreases have been observed in multiple sclerosis and degenerative diseases such as Alzhemier’s disease (AD), possibly relating to depleted neuronal density and/or neuronal dysfunction. Total Cr on the other hand is related to energy metabolism and PCr serves as a precursor for ATP synthesis. The total Cr peak is relatively stable across age and disease types such that it is often used as a normalization reference for expressing metabolite levels. Individual contributions from Cho, GPC and PC make up the 3.2 ppm CH3 peak with changes generally inferring impaired membrane metabolism. Increased total Cho signal has been reported in tumors, AD and multiple sclerosis. Note that the area underneath each MR resonance is directly proportional to the metabolite concentration and NAA, which is the highest concentrated metabolite in brain matter, produces the most dominant single resonance observed in 1H-MR spectra recorded from healthy brain. Metabolite peak integral is also proportional to the number of proton nuclei giving rise to that specific resonance. Hence, owing to their unique local electronic environments, the CH2 and CH3 protons of Cr compounds produce resonances at discrete chemical shift positions with a peak integral ratio of 2:3.

Scalar Spin–Spin (J) Coupling. The three individual proton nuclei comprising the CH3 group of Cho, Cr or NAA are both chemically and magnetically equivalent. The CH3 proton nuclei of Cho, Cr and NAA therefore co-resonate at their respective chemical shift positions yielding simple line shape structures that are typically referred to as singlet resonances. In contrast, other types of proton nuclei are both chemically and magnetically inequivalent. Quantum mechanical selection rules, which are outside the scope and aim of this chapter, dictate that magnetically inequivalent nuclei within a given chemical structure are able to interact and “sense” one another’s effective spin state through a process known as scalar spin–spin (J)-coupling. J-coupling is influenced and propagated by the electrons within covalent bonds and leads to a splitting of resonances into more complex lines distributed about a specific chemical shift position. J-coupling operates over relatively small covalent bond distances and its strength decreases with distance. We are generally concerned with two- and three-bond J-couplings that are usually denoted 2 J HH and 3 J HH, respectively, where the subscript “HH” refers to homonuclear proton–proton coupling. Metabolite resonance splitting through J-coupling effects leads to the observation of multiplet patterns for many proton groups in many of the metabolites detectable via 1H-MRS. For example, the 1H-MRS 2.0-3.0 ppm chemical shift region includes J-coupled resonances arising from Asp, GABA, Gln, and Glu. For an excellent source of proton chemical shift and J-coupling information see Govindaraju et al. (2000). J-coupling effects clearly increase the severity of spectral overlap and is a major problem for the quantification of metabolites of interest in CNS disorders including GABA, Gln, Glu and Gly. This necessitates the requirement for (i) the acquisition of high quality well-shimmed 1H-MRS datasets with sufficiently low signal line width and (ii) rigorous spectral quantification procedures and/or alternative MRS measurement strategies.

Analyzing In Vivo 1H-MRS Data

Spectral Fitting and Quantification. MRS-derived metabolite levels are expressed as metabolite ratios (e.g., metabolite:creatine) or absolute concentrations using tissue water as an internal reference standard. The complexity of 1H-MRS data dictates that simple peak integration is a grossly insufficient approach for extracting the peak area. Instead, an unbiased and automated method for spectral fitting that uses a priori information should be used.

The 1H-MRS spectral data presented at the top of Fig. 1d shows an expanded chemical shift region of the same spectrum presented in Fig. 1b. The black spectrum is the raw spectral data as obtained without additional signal processing prior to FT whereas the red spectrum that is overlaid represents as an estimated spectral fit that was carried out using a commercially available MRS fitting package (Linear Combination (LC)-Model; (Provencher 1993)). The spectral fit can be qualitatively evaluated by the residual (raw data minus spectral fit) shown at the top of Fig. 1d. This software performs its analyses in the frequency (spectral) domain and fits a series of ‘basis’ functions to in vivo MRS data. The basis functions are the individual metabolite spectra recorded from concentrated chemical solutions or “phantoms” or individual metabolite spectra that are simulated using a prior knowledge of chemical shift and J-coupling information from a molecular model. Note that it is crucial to employ the same choice of MRS pulse sequence and TE for acquiring or simulating individual metabolite spectra in order to closely match the resulting in vivo data. Furthermore, it is entirely up to the user as to how many basis functions should be included in the analysis and a total of thirteen simulated metabolite spectra were included in the basis set with the relative contributions of each to the final composite fit shown at the bottom of Fig. 1d. The macromolecule (MM) is a baseline function estimated by LCModel and accounts for resonances that are primarily attributable to the side chains of low molecular weight cytosolic proteins. The CH3 peaks of NAA and Cho are clearly well-fitted and the dual peak nature of the 3.0 and 3.9 ppm resonances due to Cr and PCr components should be noted. The remaining metabolite resonances show complex line shapes and are distributed over several resonance frequencies including the J-coupled metabolite peaks arising from Asp, Glu, Gln, and GABA. Note that NAA and Cho also contained J-coupled metabolite peaks (for an excellent source of 1H-MRS chemical shift and J-coupling data for common brain metabolites see (Govindaraju et al. 2000)).

For each metabolite species, spectral fitting algorithms should provide a measure that provides some statistical insight into the relative reliability of the individual fit. In general, these measures often infer a high degree of reliability for fitting NAA, Cr and Cho CH3 peaks although metabolite ratios and/or absolute concentrations returned for low-concentration J-coupled metabolite species such as GABA should be treated with extreme caution. For these metabolites, different 1H-MRS acquisition strategies might be utilized for improved peak resolution and quantification.

1H-MRS Metabolite Spectral Editing

A major area of 1H-MRS research at many institutions worldwide has involved the development of novel acquisition strategies that aid the robust quantification of amino acid neurotransmitter species in vivo. CNS drugs may be designed to increase synaptic availability of a given amino acid neurotransmitter by inhibiting catabolic enzymes or through blockade of transporter proteins whereas other CNS drug candidates might be potent agonists or antagonists at postsynaptic receptors. Regardless of their specific mechanism of action, monitoring amino acid modulation via 1H-MRS could provide a biomarker for drug response and deliver unique insights into the mechanism of drug action and efficacy following the administration of a pharmacotherapy.

Notably a handful of 1H-MRS techniques have evolved for measuring GABA, for example, which is the major inhibitory neurotransmitter within the mammalian brain. J-difference editing 1H-MRS has become a popular approach for quantifying GABA (Rothman et al. 1993; Mescher et al. 1998) and makes use of the J-coupling constant operating between the coupled GABA protons. The J-difference edited MRS pulse sequences introduce additional RF pulses whose frequency bands are selectively tuned to a specific GABA resonance. Using otherwise identical MRS sequence parameters (e.g., TR, TE, NEX) a series of signal averages are acquired with and without the application of the additional RF pulses. Posthoc subtraction of the “with” from the “without” spectra affords a difference spectrum that retains a GABA 3.0 ppm signal with the cancelation of most other metabolite peaks including the large creatine 3.0 ppm singlet peak. Figure 2a shows an example of a “J-difference edited” 1H-MR spectrum recorded from the same voxel shown in Fig. 1a. The data shows that additional Gln, Glu, MM and NAA resonances are “co-edited” in the final J-difference spectrum and there is some MM-contamination of the GABA 3.0 ppm resonance. Nevertheless, this type of approach provides a more robust and reliable GABA quantification procedure particularly when compared to data acquired using conventional 1H-MRS methods as shown in Fig. 1b. These editing techniques have successfully demonstrated significant increases of brain GABA levels in epileptic subjects and depressed patient populations following administration of GABAergic anticonvulsants (Petroff et al. 1999) and antidepressants (Sanacora et al. 2002), respectively.
Fig. 2 (a)

GABA J-difference edited 1H-MRS spectrum recorded from the voxel shown in Fig. 1a. In addition to the GABA resonance that is clearly resolved at 3.0 ppm, co-edited proton resonances from Gln, Glu, MM and NAA have been labeled. The spectrum was recorded using a PRESS-based sequence implementation (TR = 2,000 ms, TE = 68 ms, NEX = 256) as described by Mescher and Garwood (Rothman et al. 1993). (b) LCModel analysis of a GABA J-difference edited 1H-MRS dataset. Only the Gln, Glu and GABA basis function contributions to the composite fit are displayed for presentation purposes

A second method of metabolite-editing known as multiple-quantum coherence (MQC) filtration also takes advantage of the J-coupling phenomenon. MQC filtration methods use specific MRS sequences that act to promote quantum mechanical energy level transitions, which are permitted for nuclei within J-coupled spin systems but are forbidden for uncoupled nuclei. As such, a specific J-coupled metabolite resonance can be targeted and resolved whereas the undesired and uncoupled proton spins are simultaneously suppressed. The “single-shot” nature of MQC-editing is a major advantage of this approach compared to J-difference editing and MQC-editing methods have been used to quantify GABA (Keltner et al. 1997) and GSH (Zhao et al. 2006) in vivo. However, the total signal yield and sensitivity associated with MQC filters is relatively low.

The 1H-MRS data presented and methods discussed thus far have been one-dimensional in nature and characterized by a single frequency axis (i.e., the chemical shift axis). Another potential way to access increased metabolite information is through the use of two-dimensional (2D) 1H-MRS methods and, for these methods, the relevant MRS sequences introduce a timing variable that is systematically incremented between blocks of signal averages and following a 2D FT yield spectral data characterized by two frequency axes (e.g., chemical shift vs. J-coupling, chemical shift vs. chemical shift). The spectral resolution is effectively increased as the metabolite proton resonances are spread over a 2D surface, and a number of spatially localized variants of 2D 1H-MRS have been implemented for in vivo applications with GABA (Ke et al. 2000), Gln and Glu (Schulte and Boesiger 2006; Hurd et al. 2004) being examples of the resolvable metabolites. Specifically, 2D 1H-MRS methods have demonstrated significant GABA elevations in the rat brain following anticonvulsant treatment (Welch et al. 2003). Altered glutamatergic systems have also been linked to psychiatric illness including anxiety and depression and 2D 1H-MRS might be useful for monitoring the efficacy of novel drug candidates that target metabotrophic glutamatergic receptors (Kugaya and Sanacora 2005).

2D 1H-MRS variants also have been proposed for improving the detection reliability of the 3.55 ppm Gly singlet resonance, a peak whose resolution is compromised by the large overlapping and strongly J-coupled resonances of mI (Prescot et al. 2006). Single-shot 1D 1H-MRS Gly-detection methods were recently introduced, which attenuate the mI resonances through optimized multiple-refocusing strategies (Choi et al. 2008). Gly is a coagonist at the N-methyl d-aspartate (NMDA) receptor subtype, and NMDA receptor dysfunction is associated with psychiatric disorders including schizophrenia. Novel drug species are emerging that enhance NMDA receptor function by blocking the type-1 Gly transporter protein and increasing synaptic Gly levels. Hence, Gly-detection 1H-MRS strategies will serve as invaluable tools for monitoring glycinergic drug response.

Finally, the overall performance of metabolite editing procedures, whether J-difference editing, MQC-filtration methods or 2D 1H-MRS, might be enhanced by combining the data acquisition method with an automated and unbiased spectral fitting algorithm. For example, GABA J-difference edited 1H-MRS data could be analyzed with LCModel, where phantom-derived or simulated J-difference edited basis functions could be used for fitting the Gln, Glu, GABA and NAA contributions (see Fig. 2b). It is also worth adding that many of the spectral editing procedures described are not provided as stock sequences by scanner manufacturers and that local research physicists generally are responsible for implementation and development of editing methods. This often means that the exact choice of metabolite-editing methods at a given facility is institution and research group-specific.


  1. Bottomley PA (1987) Spatial localization in NMR spectroscopy in vivo. Ann N Y Acad Sci 508:333–348PubMedCrossRefGoogle Scholar
  2. Choi C, Bhardwaj PP, Seres P, Kalra S, Tibbo PG, Coupland NJ (2008) Measurement of glycine in human brain by triple refocusing 1H-MRS in vivo at 3.0 T. Magn Reson Med 59(1):59–64PubMedCrossRefGoogle Scholar
  3. de Graaf RA (2007) In vivo NMR spectroscopy. Wiley, ChichesterCrossRefGoogle Scholar
  4. Frahm J, Bruhn H, Gyngell ML, Merboldt KD, Hanicke W, Sauter R (1989) Localized high-resolution proton NMR spectroscopy using stimulated echoes: initial applications to human brain in vivo. Magn Reson Med 9(1):79–93PubMedCrossRefGoogle Scholar
  5. Govindaraju V, Young K, Maudsley AA (2000) Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed 13(3):129–153PubMedCrossRefGoogle Scholar
  6. Hurd R, Sailasuta N, Srinivasan R, Vigneron DB, Pelletier D, Nelson SJ (2004) Measurement of brain glutamate using TE-averaged PRESS at 3 T. Magn Reson Med 51(3):435–440PubMedCrossRefGoogle Scholar
  7. Ke Y, Cohen BM, Bang JY, Yang M, Renshaw PF (2000) Assessment of GABA concentration in human brain using two-dimensional proton magnetic resonance spectroscopy. Psychiatry Res 100(3):169–178PubMedCrossRefGoogle Scholar
  8. Keltner JR, Wald LL, Frederick BD, Renshaw PF (1997) In vivo detection of GABA in human brain using a localized double-quantum filter technique. Magn Reson Med 37(3):366–371PubMedCrossRefGoogle Scholar
  9. Kugaya A, Sanacora G (2005) Beyond monoamines: glutamatergic function in mood disorders. CNS Spectr 10(10):808–819PubMedGoogle Scholar
  10. Mescher M, Merkle H, Kirsch J, Garwood M, Gruetter R (1998) Simultaneous in vivo spectral editing and water suppression. NMR Biomed 11(6):266–272PubMedCrossRefGoogle Scholar
  11. Petroff OA, Hyder F, Mattson RH, Rothman DL (1999) Topiramate increases brain GABA, homocarnosine, and pyrrolidinone in patients with epilepsy. Neurology 52(3):473–478PubMedGoogle Scholar
  12. Prescot AP, de BFB, Wang L, Brown J, Jensen JE, Kaufman MJ, Renshaw PF(2006) In vivo detection of brain glycine with echo-time-averaged (1)H magnetic resonance spectroscopy at 4.0 T. Magn Reson Med 55(3):681–686Google Scholar
  13. Provencher SW (1993) Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med 30(6):672–679PubMedCrossRefGoogle Scholar
  14. Rothman DL, Petroff OA, Behar KL, Mattson RH (1993) Localized 1H NMR measurements of gamma-aminobutyric acid in human brain in vivo. Proc Natl Acad Sci USA 90(12):5662–5666PubMedCrossRefGoogle Scholar
  15. Salibi N, Brown MA (1998) Clinical MR spectroscopy: first principles. Wiley-Liss, CanadaGoogle Scholar
  16. Sanacora G, Mason GF, Rothman DL, Krystal JH (2002) Increased occipital cortex GABA concentrations in depressed patients after therapy with selective serotonin reuptake inhibitors. Am J Psychiatry 159(4):663–665PubMedCrossRefGoogle Scholar
  17. Schulte RF, Boesiger P (2006) ProFit: two-dimensional prior-knowledge fitting of J-resolved spectra. NMR Biomed 19(2):255–263PubMedCrossRefGoogle Scholar
  18. Welch JW, Bhakoo K, Dixon RM, Styles P, Sibson NR, Blamire AM (2003) In vivo monitoring of rat brain metabolites during vigabatrin treatment using localized 2D-COSY. NMR Biomed 16(1):47–54PubMedCrossRefGoogle Scholar
  19. Zhao T, Heberlein K, Jonas C, Jones DP, Hu X (2006) New double quantum coherence filter for localized detection of glutathione in vivo. Magn Reson Med 55(3):676–680PubMedCrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media, LLC 2010

Authors and Affiliations

  1. 1.Department of Veterans Affairs, The Brain InstituteThe University of Utah School of MedicineSalt Lake CityUSA

Personalised recommendations