Introduction

Ordered biological tissues and substructures can be studied via deuterium (²H) magnetic resonance because its strong quadrupolar interaction produces a frequency doublet which is sensitive to the effect of ordering on the time-averaged direction of the local electric field gradient with respect to the magnetic field^1,2. The low natural abundance of deuterium (0.015%) means that performing such measurements in vivo without deuterium enrichment can be challenging, but achievable at high field³. However, tissues typically contain multiple water compartments and spectra can be complicated by the superposition of ordered (anisotropic) and disordered (isotropic) signal components. One method of eliminating the signal from isotropic components is by employing a double quantum filter (DQF)^4,5, but this also significantly reduces the detectable signal. Therefore, taking advantage of a parallel study on immune cell proteomics, in which healthy human subjects ingested deuterium oxide (D₂O) to produce ~1.5% ²H enrichment over six weeks, we measured ²H DQF spectra from muscle in the presence of the enhanced signal.

Methods

Deuterium spectroscopy and chemical shift imaging (CSI) were performed on lower legs and forearms in four human volunteers, using in-house-built ²H coils resonating at 19.6 MHz, interfaced to a Philips Achieva 3T imaging system. A 16cm-diameter saddle coil was used for the lower leg, and a 15cm-diameter Helmholtz coil was used for the forearm. All measurements were performed with the limb parallel to the magnetic field direction.
Deuterium spectra were acquired from a 2 cm axial slice of the lower leg or forearm, by using hard pulses in combination with outer-volume saturation. DQF spectra were created by using the sequence in Figure 1, which produces an anti-phase spectrum whereby the quadrupolar doublet peaks acquire a relative phase of 180$$$^{\circ}$$$ (see Figure 2). This relative phase means that the signal vanishes for any components if their splitting frequency is small. DQF spectra are shown in subsequent figures, phased to produce a symmetric line-shape (labelled as dispersion spectra in anti-phase in Figure 2).
DQF spectra were acquired for various values of the creation time, $$$\tau$$$, (see Figure 1) and T_R/T_E=1000/0.58 ms, sampling bandwidth 3000 Hz, 1024 samples. The number of averages varied, depending on available signal, between 28 and 128. FIDs were processed using Matlab scripts and spectra were fitted to two independent Lorentzian line shapes.
Deuterium CSI data were also acquired from a single 2 cm axial slice, again using outer-volume suppression. CSI data were acquired for the usual pulse-acquire sequence, and using the anti-phase DQF sequence with $$$\tau=6.5$$$ ms. The in-plane resolution was $$$10\times 10$$$ mm², T_R/T_E=378/3.9 ms, 256 samples were acquired at a bandwidth of 750 Hz. For the DQF CSI, 16 averages were acquired, whereas only 4 were needed for the pulse-acquire CSI.

Results

Figure 3 shows ²H CSI data overlaid on ¹H gradient-echo images of the lower leg, for both the DQF CSI and the pulse-acquire CSI sequences. Individual DQF and pulse-acquire spectra from three voxels are also shown in more detail. The characteristic ‘sinc-like’ shape (Figure 2) can be seen in the DQF spectra, while the in-phase doublet pattern is seen in some voxels in the pulse-acquire data.
Figure 4 shows examples of four, phase-corrected, DQF spectral datasets (three lower leg, one forearm) acquired from 2 cm axial slices, acquired with a range of $$$\tau$$$-values. Each spectrum was fitted to two independent Lorentzians (as illustrated in Figure 2) from which the mean amplitude was plotted as a function of $$$\tau$$$. Examples of these DQF ‘build-up’ plots are shown in Figure 5 (two forearms, two lower legs).

Discussion

In the ideal case of perfect flip angles and the signal (central frequency) being exactly on-resonance, the DQF build-up curves, shown in Figures 4 and 5, follow the damped sinusoidal form $$$A\sin ( 2\pi\nu_q\tau ) \exp ( -2\tau/T_2 )$$$, where $$$\nu_q$$$ is the reduced quadrupolar splitting frequency⁴. Although the curves in Figure 5 are fitted to this form, it is clear that they do not perfectly model the data, most likely due to flip-angle errors, small resonance offsets, and spread of quadrupolar frequencies⁴. Fits to eight DQF datasets produced a mean and standard deviation of $$$\nu_q=39.7\pm 11.5$$$ Hz. However, in all curves, a global maximum is seen at approximately $$$2\tau\approx 15$$$ ms, which corresponds approximately to $$$\nu_q\approx 33$$$ Hz. As a comparison, mean values obtained by fitting all DQF spectra in each dataset produced $$$\nu_q=29.5\pm 3.7$$$ Hz, which is in closer agreement with values similarly measured in the lower leg by Gursan et al.³
From the spectra in Figure 3, we also observe that the splitting frequencies tend to be larger in the tibialis anterior, which is primarily believed to result from the muscle group being more closely aligned with the magnetic field direction.

Conclusions

We have shown that ²H DQF measurements can be made on a 3T scanner in skeletal muscle in human subjects and that although the resulting spectra are reduced in amplitude, the elimination of isotropic water components simplifies the line shapes and separates out signals from ordered compartments. Further work is required to understand the precise origin of the quadrupolar splitting in muscle in relation to its structure and the existence of bound water compartments⁶.

Acknowledgements

This research was funded by the NIHR Nottingham Biomedical Research Centre. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care. DJC’s Ph.D. studies are funded by the Precision Imaging Beacon at the University of Nottingham. The pilot study into immune cell responses in cancer is supported by Cancer Research UK [A29820].