Physical Address

304 North Cardinal St.
Dorchester Center, MA 02124

Single Voxel MR Spectroscopy in the Spinal Cord: Technical Challenges and Clinical Applications


5.1.1

Magnetic Resonance Spectroscopy

5.1.1.1

Introduction

In the large majority of applications MRI focuses on detecting proton signal, in particular the signal from water protons. It therefore provides information about anatomical structure and the biophysical state of tissue water.

Magnetic resonance spectroscopy (MRS) can be seen as an alternative or complementary technique to MRI as it provides chemical and biophysical information that can be extracted from molecules other than water, and from nuclei other than 1 H (with 13 C and 31 P and 23 Na being the most common). The main targets of in vivo 1 H MRS are metabolites, including neurotransmitters such as glutamate (Glu), aspartate, or gamma-aminobutyric acid (GABA) in the central nervous system (CNS).

In fact whilst the first 2D (MRI) image was produced in 1973, the discovery of MRS, originally referred to as nuclear magnetic resonance (NMR), long predates the use of MRI and led to the pioneers, Felix Bloch and Edward Purcell, being awarded the 1952 Nobel Prize in Physics.

The aim of MRS is to quantify the concentrations of specific molecules and compounds containing the nucleus of interest in well-defined volumes of interest (VOIs) in the sample or subject. As almost all spinal cord MRS studies have been performed on 1 H, the rest of this section will assume that 1 H is the detected nucleus.

5.1.1.2

Basic Principles of MRS

A number of textbooks and reviews on the principles of MRS are available. Only a very brief overview is presented here.

  • Metabolite . Indicates any substance produced during metabolism

  • ppm (parts per million) . The unit used to measure chemical shift. The proton chemical shift range is 0–15 ppm. By using these units, the chemical shift for a specific proton does not depend on the B 0 of the scanner. At 3 T, 1 ppm corresponds to ∼128 Hz

  • Chemical Shift Imaging . MR technique aimed at producing images of metabolite concentrations. It works by effectively acquiring an MRS spectrum from each of many physical locations over a 2D grid or over a 3D volume

  • Shielding constant . Determined by the electronic environment of a nucleus. Indicates the discrepancy between the applied magnetic field and the local magnetic field experienced by an individual proton

  • Spectral resolution . By definition, the resolution in an MR spectrum (frequency step between neighboring points on the horizontal axis) is determined in the first instance by the inverse of the actual data readout duration. By extension this term is used to indicate the ability to resolve neighboring peaks in a spectrum (e.g., it increases/decreases when peaks are narrower/broader)

5.1.1.2.1

Chemical Shift

The physical principle underlying MRS is that a proton will experience a slightly different magnetic field depending on its chemical environment. This is because the electrons spinning around the nucleus create their own tiny magnetic field, shielding the nucleus from the main magnetic field B 0 .

The keyword here is shielding and refers to the difference between an externally applied magnetic field ( B 0 ) and the actual field B 0k experienced by each particular proton (identified by the subscript “k”):


B 0 k = B 0 · ( 1 − σ k )

where the shielding constant σ k is independent of B 0 .

Given the different shielding constants, the resonance frequencies (Larmor frequencies [ ω Lk = γ B k ]) of protons belonging to different molecules will vary. The signal collected from a VOI in the sample/subject is Fourier transformed to display the signal on a frequency axis: different chemical environments will appear at different places (frequencies) in the spectrum. The frequency difference from a reference compound is dubbed chemical shift , expressed in parts per million (ppm) (i.e., at 3 T 1 ppm is ∼128 Hz). By convention, the frequency axis on which metabolite resonances are displayed is inverted such that lowest frequencies are on the right. The 0 ppm point is where the methyl (CH 3 ) protons of DSS (4,4-dimethyl-4-silapentane-1-sulfonic acid) resonate: these protons are highly shielded by their electron-dense environment, and most other metabolite protons experience less shielding (i.e., greater magnetic fields), have positive chemical shift, and appear to the left of DSS on the spectra.

An example of human brain spectrum from parietal gray matter is shown in Figure 5.1.1 . For instance the methyl protons of N -acetyl-aspartate (NAA) resonate at 2.01 ppm and free water protons at 4.7 ppm.

FIGURE 5.1.1, Spectrum acquired at 3 T from a 15.6 ml VOI in the parietal GM of a 32-year-old female. TR = 2 s, TE = 30, NSA = 256. SNR ∼ 95, FWHM 4.7 Hz; the peaks from the most prominent brain metabolites are labeled. The spectrum is shown as displayed in the standard output from LCModel with estimated metabolite concentrations on the right-hand side.

5.1.1.2.2

Spin–Spin Coupling

Many biological molecules contain several protons. Because protons possess a small magnetic moment (spin) themselves, they influence each other, thus they affect the magnetic field experienced by nearby protons; this effect is known as spin–spin coupling . a

a Heteronuclear coupling also exists, e.g., between 1 H and 13 C.

In the liquid state this spin–spin interaction is typically canceled through space interaction due to rapid molecular tumbling. In chemical bonds, instead, spin–spin interaction through electrons does not average to zero and is responsible for the “splitting” of resonance lines. This coupling is expressed in units of Hz as it is independent of the externally applied B 0 and only varies with the number (and type) of bonds between the protons.

Each proton will be affected by the spin state of nonmagnetically equivalent protons b

b Magnetically equivalent nuclei have the same chemical shifts (chemically equivalent), plus they have same coupling constants with all other coupled nuclei. For instance, CH 3 protons are magnetically equivalent; therefore, if uncoupled they will appear as a singlet. They have a different chemical environment from CH protons, thus they have a different chemical shift.

in the same molecule. Due to the different chemical shifts and coupling constants, each molecule will display one or a number of peaks (or “resonances”) on the MRS frequency axis, its MRS “signature”.

For instance, any proton that is not coupled to another proton would show a “singlet” resonance, i.e., a single peak in the MRS spectrum. This is not the case for coupled spins, for example in lactate, a group of magnetically equivalent protons (CH 3 ) is coupled to the CH proton. The proton in CH can be in one of two states, parallel and antiparallel to B 0 , each affecting the CH 3 protons slightly differently. The CH 3 singlet therefore experiences a “splitting” and appears in a spectrum as a “doublet” of peaks, each with half of the area it would have had if there had been no coupling. The CH proton resonance is in turn coupled to the three protons of CH 3 and affected by their individual spin states, which means that the CH spectrum will therefore show four lines (quartet) with intensity ratios 1:3:3:1 (corresponding to one configuration for all spins up, three possible configurations for two up one down, three configurations for two down one up, and one configuration for all down). For a CH proton coupled to a CH 2 group, we would see instead a “doublet of doublets” that would form a “triplet” of peaks with ratios (1:2:1).

Spin–spin couplings and the multiplets they produce in relation to the underlying molecular structure are thus extremely useful, together with chemical shift information, to identify and assign resonances.

High resolution spectra allowing split resonances to be fully resolved can often be obtained on ex vivo tissue extracts or homogenized solutions at high magnetic field (7 T–14 T).

In vivo MRS spectra have limited resolution and in many cases the line splitting cannot be directly observed. In any case, knowledge of the theoretical splitting allows better modeling of the observed signal.

5.1.1.2.3

Aim of MRS Spectral Analysis

Once an MRS spectrum is collected the aim of the MRS spectroscopist is to correctly identify the resonances in the spectrum and assign them to the corresponding molecules. The greater the concentration of a particular molecule in the VOI, the greater the amplitude of its corresponding MRS signature (resonances). By measuring peak amplitudes, after correct assignment it is possible to calculate the absolute or relative concentration of the molecules from which these peaks originate.

This procedure can give direct and detailed information on the biochemistry within the VOI in a wholly noninvasive fashion. Whilst single-point MRS gives a snapshot view of metabolite concentrations at a specific moment in time, repeated measurements can also be carried out to provide, for instance, dynamic metabolic information.

5.1.1.2.4

Features of MRS Spectra

The main complication in the spectral analysis is that the MRS signatures of many molecules do partially overlap, and this causes an uncertainty associated with each frequency (as explained following), which makes the assignment procedure potentially equivocal.

One of the characteristics of each resonance peak, in fact, is its linewidth (often calculated as the full width at half maximum, FWHM) that depends on its T 2 ∗ (apparent transverse relaxation time constant), related to the compounded effect of local and macroscopic magnetic field inhomogeneities. While the effects of inhomogeneities that are stationary in time can be reversed or accounted for with appropriate acquisition techniques, local magnetic field fluctuations are randomly variable in time and are inherent to the system under study.

As in MRI, MRS data quality can also be characterized by a signal-to-noise ratio (SNR) value, expressing the ratio between resonance peaks arising from a chosen compound and random noise occurring equally at all spectral frequencies.

In addition to random noise, the analysis will be complicated by signals arising from molecules and compounds that are not expected or taken into account explicitly (often this is the case for extremely broad resonances from macromolecules resulting in nonflat spectral baselines).

The ability to discriminate resonances will increase as peak linewidths are minimized and SNR maximized. Linewidths can be improved by optimizing the uniformity of the local magnetic field over the VOI (shimming) prior to the acquisition of MRS data (see Section 5.1.2.2 ). SNR will depend on the acquisition parameters: choice of echo time (TE) and repetition time (TR), acquisition bandwidth (BW), as well as the volume of the VOI (VOI volume ) and the number of spectral averages ( N avg ) collected.

Similarly to MRI, for a singlet resonance peak this can be generally expressed as:


SNR ∼ N avg BW · VOI volume · e − TE T 2 · ( 1 − e − T R T 1 )

with T 1 and T 2 being the longitudinal and transverse relaxation times, respectively.

In other words, SNR increases linearly with VOI volume , but only with the square root of N avg . Knowledge of T 1 of the substance of interest can allow optimization of SNR per unit time, though in some instances long TRs are preferable to provide insensitivity to pathological T 1 changes. Reduced TE also allows better SNR, however the contribution to the spectra of short-TE macromolecules results in rolling spectral baselines and complicates quantification. Lower readout bandwidths give an SNR advantage, however, the BW cannot be arbitrarily chosen and needs to be large enough to span across the frequencies of resonances from all detectable metabolites.

5.1.1.2.5

Basics of Localized MRS Data Acquisition

The first MRS spectra ever acquired were collected as a rapidly decaying signal or free induction decay (FID). This was measured immediately following a nonspatially selective excitation of a sample in an external (uniform and stationary) B 0 magnetic field. The excitation consisted of a magnetic field ( B 1 ) perpendicular to the B 0 direction and oscillating or rotating at the Larmor frequency corresponding to it. It is therefore common to refer to this magnetic field as a radiofrequency (RF) pulse. In this experiment the FID, contains information from the whole sample experiencing the applied magnetic field B 1 .

However, often it is necessary to define the volume from which the signal originates. The simplest way to achieve a spatial localization of the collected signal is by making sure that the excitation pulse is “on resonance” only for a selected VOI as follows.

The B 1 pulse can be designed to excite only a narrow slab of a sample exactly as in MRI by applying at the same time a magnetic field that has the same direction of B 0 but varies in space, i.e., a magnetic field gradient. Employing three subsequent B 1 pulses with gradients along three orthogonal directions defines a 3D VOI where the three slabs overlap. One of the most commonly employed sequences is equivalent to a double-echo spin-echo and employs B 1 pulse with typical flip angles of 90°, 180°, 180° and was originally suggested in 1984 with the acronym of PRESS (point resolved spectroscopy). If the three B 1 pulses are all producing 90° flip angles another extremely common MRS sequence is produced, commonly referred to with the acronym of STEAM (stimulated echo acquisition mode ), where it is only the resulting stimulated echo that contains the localized spectral information desired. (Note: See Section 5.1.3.5 for a discussion of the relative merits of these two sequences in relation to spinal cord MRS.)

Signal from multiple VOIs from a slab or volume of tissues can also be collected; the technique is called chemical shift imaging (CSI) or MR spectroscopic imaging (MRSI). For instance, in PRESS-based CSI, the three overlapping slabs define a large VOI that is further subdivided into smaller elements by means of phase encoding as in conventional MRI. c

c This “phase encoding” is applied along one, two, or three orthogonal directions for 1D-CSI, 2D-CSI, and 3D-CSI, respectively.

5.1.2

1 H MRS in the Spinal Cord

5.1.2.1

Introduction

Twenty years passed after the first human in vivo MRS experiment, before localized MRS in the spinal cord was performed. The group that first investigated it was based in the Institute of Neurology in University College London in 1997–2000. It took a further four years for other groups to pick up the challenge. Currently there are still only a few groups worldwide that routinely undertake spinal cord MRS, and one of the reasons for this is that it is technically challenging.

A selection of studies that have reported spinal cord MRS methodology are listed in Table 5.1.1 and will be referred to in more detail in the next sections.

TABLE 5.1.1
Selection of Spinal Cord MRS Papers
Publication Year B 0 RF Coil Sequence Water Supp. Cardiac Trigger TR (s) TE (ms) NSA Voxel Size Position Shim Method Subject Group
Gomez-Anson 2000 1.5 T Volume PRESS CHESS – 3 30/144 256 9 × 60 × 40 – – Healthy
Cooke 2004 2 T Surface PRESS CHESS Yes (400 ms) 3 30 256 9 × 7 × 35 Above C2 B 0 mapping Healthy
Kendi 2004 1.5 T Phased array PRESS MOIST – 1.5 35 196 6 × 6 × 50 C3–C7 Auto MS
Kim 2004 1.5 T Flexible surface STEAM CHESS – 2 30 256 – Spinal tumor mass – Spinal mass lesions
Blamire 2007 2 T Surface PRESS CHESS Yes (400 ms) 3 30 256 9 × 7 × 35 Center C3 B 0 mapping MS
Ciccarelli 2007 1.5 T Saddle PRESS CHESS Yes (150 ms) 3RR 30 192 6 × 8 × 50 C1–C3 Manual MS
Edden 2007 3 T Flexible surface 1D-PRESS CSI Dual HS pulses No 2200 144 64 FOV 10 × 12 × 90 Medulla-C3 FASTMAP Healthy
Henning 2008 3 T 12 Channel spine PRESS CHESS Yes (300 ms) 2 42 512 6.5 × 8.5 × 27 C2–C3,Thoracic lumbar B 0 mapping Healthy + Various pathologies
Holly 2009 1.5 T – PRESS – Yes 1.5/3 30 256 10 × 10 × 20 Center C2 Manual Cervical Spondilotic myelopathy
Ciccarelli 2010 1.5 T Saddle PRESS CHESS Yes (150 ms) 3RR 30 192 6 × 8 × 50 C1–C3 Manual MS
Marliani 2010 3 T 8 Channel spinal PRESS CHESS – 2 35 396 7 × 9 × 35 C2–C3 Auto non- triggered MS
Carew 6 2011 3 T Volume PRESS CHESS – 2 35 256 8 × 5 × 35 Above C2 B 0 mapping ALS at risk
Kachramanoglou 2011 3 T 8 Channel PRESS CHESS – 3 30 ∼160 – C1–C3 B 0 mapping MS and brachial plexus
Elliott 2011 3 T – PRESS – – 3 30 ∼220 10 × 10 × 20 C1–C3 Auto Whiplash
Hock 2011 3 T 16 Channel PRESS VAPOR/None Yes >2.5 30 512 6 × 9 × 35 – FASTERMAP Healthy

5.1.2.2

Challenges

The main factors limiting the quality of spinal cord MRS spectra are:

  • adequate VOI positioning, avoiding CSF contamination

  • motion of cord and surrounding CSF

  • resulting challenge of shimming an elongated VOI surrounded by CSF

  • low SNR with respect to conventional brain studies due to the small cross-sectional area of the cord; longer acquisition times also incur more subject motion.

5.1.2.3

Main Metabolites Observed in the Spinal Cord

The relatively low SNR achievable in spinal cord MRS in sessions of reasonable duration (e.g., up to 20 min) does not generally allow for the detection of as many metabolites as in the brain where around 15 metabolites can be measured in less than 10 min acquisitions, depending on location and VOI volume. However most publications quantify the most prominent features in spectra from the CNS, which are outlined following:

  • NAA, often reported together with NAAG from which it cannot be easily differentiated in vivo; NAA is selectively localized within neurons, and a decrease is commonly used as a marker of neuronal injury; reduced relative NAA concentrations are said to reflect axonal loss and/or metabolic dysfunction.

  • total Choline (tCh) (including contribution from free choline, glycerol-phosphoryl-choline [GPC] and phosphorylcholine [PC]). Changes in tCh levels are generally associated with membrane composition with increased signal associated, for example, with cancer, ischemia, and head trauma.

  • total Creatine (Cr) (creatine plus phosphocreatine) levels are related to energy metabolism as phosphocreatine behaves as reservoir for the generation of adenosine triphosphate (ATP). Cr is relatively stable across age or disease and is thus often used as reference compound.

  • myo-Inositol (mIns) mIns is considered to be a glial marker; its increase can be related to astrocytic activation and proliferation.

  • Glutamate (Glu) is the major excitatory neurotransmitter in mammals. Despite is relatively high abundance it is particularly difficult to quantify, since its signal is spread over many low-intensity resonances. It is often quantified together with glutamine (Gln), an important component of metabolism (Glu + Gln = Glx). Few spinal cord MRS studies have reported Glu or Glx levels.

ppm Metabolite Properties
2.01 N -acetylaspartate Neuro-axonal marker
2.0–2.4 Glutamate/Glutamine Neurotransmitters
3.03 Creatine/Phosphocreatine Energy metabolism
3.22 Choline compounds Cell membrane marker
3.56 Myo-inositol Glial cell marker

5.1.2.4

Short Summary of Findings from Clinical Applications

Looking again at Table 5.1.1 , it is apparent that the main clinical application of spinal cord MRS has been to multiple sclerosis (MS), with a few reports on other pathologies including amyotrophic lateral sclerosis (ALS), cervical spondylotic myelopathy, spinal mass lesions, tumors, whiplash, and brachial plexus avulsion. The most commonly and consistently reported finding is that of reduced NAA/Cr concentration ratio in pathological spinal cord compared to healthy subjects. In MS, reduced NAA concentration (concentration is denoted by []) was also observed, together with the NAA/Cho ratio whilst Cho/Cr and mIns/Cr increased. It has also been reported that reduced NAA at the onset of an MS relapse can be followed by a partial recovery over time with greater increase at one month correlated to greater recovery. NAA was also shown to correlate with cerebellar scores of neurological assessment.

mIns was measured to be slightly higher in MS patients than controls, but the difference was not statistically significant; however, it did correlate with EDSS. mIns/Cr was significantly elevated in MS vs. controls in another study, whilst lower mIns/Cr than in controls was found in a case series of five patients with neuromyelitis optica (NMO), potentially reflecting astrocytic damage.

5.1.3

Technical Issues

5.1.3.1

Anatomy/Geometry of the Cord

On axial images at the level of the cervical spinal cord it can be seen that the cord has an oval cross section, with an LR axis of approximately 8–11 mm and an AP axis of 6–9 mm. It is “padded” all around by CSF in a thickness of 2–8 mm.

Typically the minimum VOI volumes used in single-voxel MRS are around 4–8 ml, however, rarely voxels larger than 2–2.5 ml can be achieved from the spinal cord.

As summarized before in Eqn (5.1.2) , if the VOI volume is halved, four signal averages are necessary to recover the original SNR, so the VOI needs to be maintained as large as possible, as long as it does not span outside the cord.

The obvious approach to increasing the VOI dimensions for spinal cord is to extend it in the HF direction, however, given the VOI has a cuboidal shape, the maximum usable length depends on the curvature of the cord in the chosen section. In diseases that involve neuronal loss, atrophy of the spinal cord also limits the size of the VOI.

Other considerations affecting voxel positioning include B 0 inhomogeneity (see Chapter 2.2 or Section 5.1.3.7 of this chapter on shimming).

5.1.3.2

Main Field Strength

MR scanners of higher and higher fields are increasingly popular and widespread. The highest B 0 currently used for in vivo studies in humans is 9.4 T, though 11.7 T scanners are currently being built. As described in Chapter 2.4 , one benefit of higher field strength is that signal to noise (SNR) increases approximately linearly with field strength. Moreover, higher field strength also offers higher spectral resolution—the capability to resolve two adjacent spectral peaks, thereby making it possible to investigate other metabolites such as GABA which is overlapped by NAA and Cr at 3 T but resolved at 7 T given the larger chemical shift. Similarly, at 7 T the higher field results in the added information from Glu and Gln, which cannot be resolved at 3 T without spectral editing pulse sequences. At the same time, with increased B 0 we also observe increased longitudinal relaxation times T 1 , increased magnetic susceptibility differences between tissue/air/bone structures causing increased susceptibility-dependent effects, causing line broadening in MRS as well as chemical shift positioning errors (see following).

The first few spinal cord MRS studies have been conducted at lower clinical field strengths of 1.5 T and 2 T, whilst since 2007 3 T has been more commonly employed. Higher field strengths are employed predominantly for brain applications. In addition to the lack of high field studies on the spinal cord there are limited sites with coils that are built for spinal cord purposes. In addition to hardware constraints for spectroscopy at 7 T or higher, field strengths would also need to overcome issues related to chemical shift positioning errors, which would make MRS VOI placement in the spinal cord more challenging.

These chemical shift errors are related to the limited effective bandwidth of the selective excitation and refocusing pulses used for VOI localization and result in signal from different metabolites arising from different spatial locations, much the same as in the fat appearing shifted from the water signal in conventional MRI. For instance the main resonance used for NAA quantification occurs at 2.01 ppm whilst for mIns it appears at 3.56 ppm, a 1.55 ppm difference; this corresponds to 99 Hz at 1.5 T ( ω L = 64 MHz) and twice as much at 3 T. Using for instance a 1 kHz bandwidth refocusing pulse, defining one of the short dimensions of the VOI, say 6 mm, it will result in a spatial shift between NAA and mIns VOIs of 0.6 mm and 1.2 mm (20%) at 1.5 T and 3 T, respectively, and 1.8 mm at 4.7 T. In other words, the greater the field strength, the greater the chance that the signal from one of the metabolites comes at least partially from outside the cord and could include CSF contributions. Whilst using higher bandwidths would drastically reduce this issue, in practice realistic pulse bandwidths are restricted as higher bandwidths are associated with higher tissue heating, especially at higher B 0 s (see, for instance, Kinchesh 2005, JMR , for issues in carrying out MRS at 4.7 T). In addition to chemical shift–associated problems, magnetic susceptibility effects further exacerbate shimming issues within the cord, and hence even though an increase in SNR is theoretically possible at higher fields, unless MRS acquisition sequences are carefully and cleverly optimized, spectral quality in the spinal cord may suffer.

5.1.3.3

Coil Selection

Similarly to MRI, in spinal cord MRS the choice of RF coil is particularly important; a thorough description of spinal cord coils is given in Chapter 2.1 . A few considerations though are worth repeating with particular reference to MRS. The small anatomy and the shape of the surrounding area (i.e., neck) must be considered, as some coils do not have adequate sensitivity for this region. Simple “scout” imaging protocols can give information on the cervical cord area and potential sensitivity drop-offs as distance from the coil increases.

The sensitivity of the coil, the ability to produce accurate flip angles for correct localization, and the filling factor of the coil (the fraction of the coil detection volume filled with sample) all need to be considered to get the highest SNR and hence quicker scans as a result. Following, two types of coil are considered and the advantages and drawbacks of each discussed.

5.1.3.3.1

Surface Coils

Surface coils, which both transmit (Tx) and receive RF signal (Rx), were traditionally favored in MRS protocols due to their high local sensitivity. However the sensitivity falls off rapidly (for a loop, as distance 3 ) and is roughly only acceptable within one radius of the coil. This means the anatomically deep cervical spine, which lies approximately 7–8 cm from the surface of the neck, is only within the sensitive region of relatively large surface coils. Flat surface coils are unlikely to offer a good filling factor due to the small width and curvature of the neck when lying in a supine position. This curvature of the neck also leaves a gap between the tissue and the coil, exacerbating shimming issues.

Given the rapid drop off in the B 1 profile of surface coils, when they are used for excitation, adiabatic pulses (pulses that are insensitive to B 1 variations) must be used to give more accurate excitation and inversion profiles over the sensitive region. However, these adiabatic pulses are usually much longer than conventional RF pulses and result in: (1) significantly increased minimum TE, thus reduced SNR and reduced number of metabolites detectable before dephasing due to T 2 relaxation, and (2) increased heating (specific absorption rate, SAR), especially at medium and high magnetic field strengths.

The suitability of a surface coil largely depends on the diameter and shape of the coil available. Some studies have used flexible surface coils, which offer better filling factors and can be volunteer specific in shape.

It is advised to compare suitable surface coils against a volume coil in the first instance .

The quality of structural images needed for voxel placement must also be taken into account. These will suffer due to B 1 inhomogeneities and/or signal drop off if these are acquired with a surface coil, resulting in images with spatially variable SNR.

Alternatively to overcome the Tx B 1 -inhomogeneity issues whilst keeping the sensitivity of the surface coil, RF coils, which are only active during signal detection (Rx), may provide a solution (see following).

Surface coils have been chosen over volume coils at all field strengths 1.5 T–3 T and provided successful metabolite detection in scan times comparable to volume coils. However, more recently they are used less and less due to the availability of volume coils and modern array coils where the individual elements can be chosen and combined on a subject-to-subject basis.

5.1.3.3.2

Volume and Phased Array Coils

Volume coils generally offer much better spatially uniform inversion and excitation profiles. The main drawback of such coils when used in combined Tx/Rx mode is that signal is received from the entire coil whose sensitive volume is much greater than the VOIs used in MRS. Hence, most of the coil is not sensitive to the region of interest and simply adds noise to the spectrum.

Many modern scanners offer a single body–Tx coil for excitation. Whilst the resulting B 1 is very uniform, the large diameter (typically 50–60 cm) causes the coil to be very inefficient for a small VOI, and the maximum bandwidth that can be obtained is much lower than if, for instance, a head-only Tx/Rx coil was used.

The issues related to choosing a single surface or volume head coil for both Tx and Rx have effectively been overcome by the widespread use of Rx-only phased array coils. Phased array coils consist of a number of different coils that all receive like small surface coils that are highly sensitive only to the surrounding area. At the same time, given that the coil array spans a larger area, a relatively high and uniform spatial sensitivity is achieved across the whole area covered.

Roughly speaking, if we replaced a cylindrical “volume coil” with smaller coils placed on the same surface, the sensitivity would increase with the number of coils used close to the surface itself; in the middle of the cylinder, the sensitivity would remain approximately the same as for the equivalent volume coil.

In a realistic situation, for single voxel MRS, only a few of the coils would be sensitive to the VOI. The others would only add noise to the spectrum. To overcome this potential drawback, usually the raw spectrum from each coil is weighted according to the received signal from an MRS reference scan of the VOI with no water suppression and appropriately phased. In this manner coils further away from the VOI will be assigned lower weighting than those closer and most sensitive to the VOI. Some MRS sequences have an option to turn on sensitivity weighting, although often this may not be available for MRS, but only for imaging, so this should be checked when selecting the appropriate coil. A number of studies have successfully detected signal from the main metabolites using a phased array coil even without using sensitivity weighting.

5.1.3.3.3

Filling Factor and Shimming

An inherently inhomogeneous B 0 field surrounds the cervical cord area. To overcome the adverse effects of this, active shimming and dielectric foams and beads can help to alleviate this similarly to imaging. When employed for spinal cord MRS, rather than giving a more homogenous signal over the area scanned, these methods and materials are likely to result in better-resolved, higher-SNR spectra.

You're Reading a Preview

Become a Clinical Tree membership for Full access and enjoy Unlimited articles

Become membership

If you are a member. Log in here