Hyperpolarized nanodiamond with long spin-relaxation times

The use of hyperpolarized agents in magnetic resonance, such as 13C-labelled compounds, enables powerful new imaging and detection modalities that stem from a 10,000-fold boost in signal. A major challenge for the future of the hyperpolarization technique is the inherently short spin-relaxation times, typically <60 s for 13C liquid-state compounds, which limit the time that the signal remains boosted. Here we demonstrate that 1.1% natural abundance 13C spins in synthetic nanodiamond can be hyperpolarized at cryogenic and room temperature without the use of free radicals, and, owing to their solid-state environment, exhibit relaxation times exceeding 1 h. Combined with the already established applications of nanodiamonds in the life sciences as inexpensive fluorescent markers and non-cytotoxic substrates for gene and drug delivery, these results extend the theranostic capabilities of nanoscale diamonds into the domain of hyperpolarized magnetic resonance.

The orange shaded region is the size range, and the dashed orange line is the median particle size as specified by the supplier.

Feasibility of imaging with hyperpolarized nanodiamonds
We expect to perform hyperpolarized ND imaging in a preclinical scanner with a small tip angle 2D fast spin echo (FSE) sequence, with 1 H-13 C co-registration. Overlaying 13 C images of functionalized nanodiamonds on high resolution 1 H anatomical images would provide information in a similar format to those seen in PET/MRI 1 . Here, we present calculations estimating a pixel signal-to-noise ratio (SNR) of ∼ 11 for a nanodiamond concentration of 1 mg mL −1 . This SNR value is for 2 mm × 2 mm sized pixels in a 5 mm slice assuming significant polarization loss during transfer from polarizing cryostat to imager. We note that nanodiamond concentrations of 2 mg mL −1 have previously been used in vivo 2 .
First, we consider the fundamental limit to MRI resolution set by the observed transverse coherence time T * 2 , which is reached when the frequency line-width of the signal is approximately equal to the frequency separation between each pixel 3 : where dz is the pixel length, γ the nuclear gyromagnetic ratio and G is the peak gradient strength. Our nanodiamond samples have T * 2 ∼ 250 µs, which corresponds to a fundamental resolution limit of 0.25 mm in a preclinical scanner or 2.5 mm in a whole body MRI scanner (assuming typical peak gradient strengths for these systems of G = 500 mT m −1 and G = 50 mT m −1 respectively).
Next, we consider the pixel SNR that would be possible from our hyperpolarized nanodiamond samples in an imaging experiment. Our 2 µm ND samples have a polarization after DNP at 4 K, P DNP , of ∼ 8%. The thermal polarization, P thermal , at B 0 = 7 T and T = 300 K, is 0.0006%, as given by the Boltzmann distribution: Where γ is the gyromagnetic ratio and k B is Boltzmann's constant. The free induction decay (FID), after a π/2 pulse, from a 0.1 g, thermally polarized, 2 µm ND sample was acquired in our 7 T spectroscopic probe under the matched filter condition, t acq /T * 2 = π/2, where t acq is the acquisition time. The Fourier transform of this FID has SNR π/2-thermal = 35.
Preliminary transfer measurements between the hyperpolarizer and 7 T detection magnet have shown a sample transfer efficiency, η, of 10%. Hence, we predict that, after sample transfer of a hyperpolarized sample to our 7 T spectrometer, we will have an SNR post transfer, SNR π/2-PT , of 45 000, Scaling this SNR value to give an expected sensitivity in a preclinical imaging experiment is inherently nontrivial due to the difficultly of estimating noise associated with coil resistance and losses arising from the sample 4-6 . Here we make an SNR estimate for a preclinical scanner on the assumption that our SNR is limited primarily by coil resistance, which is generally true at B = 7 T for mouse coils 7 .
The SNR of a pickup coil scales as: where Br Ir is the magnetic field strength of the pickup coil per unit current, T is the coil temperature and R c is the coil resistance. Br Ir at the centre of an optimised saddle coil is given by: where N is the number of turns in the coil, µ 0 is the permeability of free space, l is the coil length and d is its diameter 8 . If all power dissipation occurs in the coil, then we can estimate the coil resistance from where L c is the coil inductance and ω 0 is the resonance frequency. Typically Q is ∼100 at 75 MHz for preclinical imaging and spectroscopic NMR probes. Assuming a homogeneous field across the saddle coil, we estimate from Faraday's law that the coil's inductance scales approximately as 8,9 Assuming a 1 turn 40 mm diameter, 60 mm long saddle coil is used for mouse imaging, the ratio Br Ir is reduced by 95% compared to the 2 turn 6 mm diameter, 13 mm long coil in our spectroscopic NMR probe. Therefore, the expected SNR after DNP and transfer to the imager is SNR π/2-imager = 2100. This result is very similar to that obtained when the resistance is simply scaled by the ratio of the wire lengths in the coils.
We envision using a small tip angle 2D FSE sequence, similar to that used in Ref. 10 Where N is the number of pixels across an N × N image, N 0 is the number of pixels across the object, θ is the tip angle (setting the acquisition time t acq ∼ T * 2 ). If the 0.1 g of nanodiamond powder in our sample is uniformly distributed through a 40 mm × 40 mm × 5 mm phantom there is a nanodiamond concentration of 16 mg mL −1 . For a 32 × 32 pixel image with 2 mm × 2 mm resolution and 5 mm slice thickness, N = 32, N 0 = 20. For tip angles 10 • and 90 • , this gives SNR pixel = 30 and SNR pixel = 170 respectively. Normalizing this value, we predict SNR pixel = 11 at 1 mg mL −1 for a 90 • tip angle.
Hence, we estimate that there will be sufficient SNR for hyperpolarized nanodiamond imaging. In practice, the measured SNR will deviate from these values depending on the actual sensitivity of the detection coil and polarization lost during sample transfer. We have also not considered the loss of spin coherence due to T 2 effects during the acquisition sequence. These effects will cause some degradation of the SNR at higher spatial frequencies.
A range of linewidth narrowing sequences developed for solid imaging may also help to improve image quality 3 .