Introduction

Deuterium (²H) has a 0.01% natural abundance, which is low in comparison to that of proton (¹H), but comparable to the concentrations of other ions, such as intracellular ²³Na. The resulting lower sensitivity of ²H-MRI compared to ¹H-MRI has resulted in a paucity of research in this area. There has been a resurgence of interest in ²H-MRI with the advent of deuterium metabolic imaging (DMI) and the imaging of ²H-labelled glucose. ²H-labelled molecules that measure metabolic activity are particularly interesting for measuring the Warburg effect in brain tumours. Recent work has shown MRI of ²H-labelled water and [6,6′-²H₂]glucose at concentrations as high as 5%. ²H-labelled glucose has been shown to metabolise into lactate and glutamine/glutamate in the normal human brain and in brain tumours. ²H-MRI with [6,6′-²H₂]glucose has the potential to rival FDG-PET imaging without the need for ionising radiation (1).

One of the challenges of DMI is that it is difficult to optimise and standardise ²H-MRI in the absence of endogenous signal. Here we demonstrate natural abundance ²H-MRI, which can be used for optimization and quality assurance of hardware, pulse sequence, and reconstruction methods. This work will also aid the acquisition of signal following the administration of ²H-labelled molecules.

Methods

We modified a 16-rung birdcage coil to resonate at 19.8 MHz for ²H-MRI (Figure 1) for a 3T MRI system (MR750, GE Healthcare, Waukesha, WI). The birdcage connected to a quadrature ²H transmit/receive switch (Clinical MR Solutions, Brookfield, WI). The 28 cm diameter birdcage included copper conductor rungs with 8 cm length, 1 cm width, and with 2 cm end-ring widths.

We imaged natural abundance ²H in the brains of two cadaveric sheep and in three normal volunteers with 3D chemical shift imaging (CSI). Volunteer imaging occurred with informed consent and local ethical approval. The 3D density-weighted CSI sequence acquired either 2 averages of a 12x12x12 matrix size, 2405 FID acquisitions, scan time 9.5 minutes; or 1 average of a 16x16x16 matrix size, 5408 transients, scan time 10.8 minutes. Both used field-of-view = 240 mm3, TR = 120 ms, flip angle = 45°, bandwidth = ±2.5 kHz, with 512 spectral points between -128 and +128 ppm per voxel, and both were zero-filled once in the time domain and once spatially in the X and Y planes for analysis. For viewing images based on the maximal spectral intensity, an additional 4x4 interpolation was applied with 3-point spline interpolation using SciPy (2).

3D gradient echo T₁-weighted images were also obtained with the ¹H body coil without repositioning of the subjects.

Results

Figure 2 shows sample spectra from a 5x5x2 cm³ region in the center of the brain. The spectra are dominated by a central heavy water peak, and several spectra show a possible fat peak at around -4ppm relative to water.

Figure 3 shows the maximum signal in two axial slices of a normal volunteer. The cerebral ventricles demonstrate hyperintensity.

Figure 4 shows a sagittal view of the brain with the voxel-wise mean of the CSI spectra, demonstrating good delineation of the brain in 3D.

Figure 5 shows the MRF-derived ¹H density and relaxation maps, which should relate to the endogenous images obtained with ²H.

Discussion

Despite the low natural abundance of ²H, we demonstrated that endogenous 2H signal of normal brains can be imaged in vivo using MRI.

This work demonstrates that the evaluation of ²H methods can be performed without the need for increased signal following the administration of ²H-labelled molecules. Natural abundance imaging simplifies the development and optimization of new methods, such as the evaluation of more sensitive radiofrequency coils.

Baseline measurements with natural abundance imaging also establish the feasibility of performing dynamic metabolic measurements. Administration of ²H-labelled molecules would significantly increase the signal-to-noise above this baseline, e.g., using [6,6′-²H₂]glucose. Resolution of isotopically-labelled metabolite peaks from the nearby naturally abundant heavy water signal will require more careful optimization of field homogeneity than that achieved in the current study using only linear shim adjustment.

Conclusion

We have demonstrated natural abundance deuterium imaging in vivo in normal human brains, which can be useful for methodological evaluation of pulse sequences and radiofrequency coils, and for baseline measurements before dynamic imaging.

Acknowledgements

This work acknowledges support from the National Institute of Health Research Cambridge Biomedical Research Centre, Addenbrooke's Charitable Trust, Cancer Research UK, and from the European Union's Horizon 2020 research and innovation programme under grant agreement no. 761214.