Sub-second Proton Imaging of ¹³C Hyperpolarized Contrast Agents in Water

Abstract

Indirect proton detection of ¹³C hyperpolarized contrast agents potentially enables greater sensitivity. Presented here is a study of sub-second projection imaging of hyperpolarized ¹³C contrast agent addressing the obstacle posed by water suppression for indirect detection in vivo. Sodium acetate phantoms were used to develop and test water suppression and sub-second imaging with frequency selective RF pulses using spectroscopic and imaging indirect proton detection. A 9.8 mM aqueous solution of ¹³C PHIP hyperpolarized 2-hydroxyethyl-¹³C-propionate-d_2,3,3 (HEP), <P>~25% was used for demonstration of indirect proton sub-second imaging detection. Balanced 2D FSSFP (Fast Steady State Free Precession) allowed recording proton images with FOV = 64×64 mm² and spatial resolution 2×2 mm² with total acquisition time of less than 0.2 s. In thermally polarized sodium 1-¹³C-acetate, ¹³C to ¹H polarization transfer efficiency of 45.1% of the theoretically predicted values was observed in imaging detection corresponding to an 11 fold of overall sensitivity improvement compared to direct ¹³C FSSFP imaging. ¹³C to ¹H polarization transfer efficiency of 27% was observed in imaging detection corresponding to a 3.25 fold sensitivity improvement compared to direct ¹³C FSSFP imaging with hyperpolarized HEP. The range of potential applications and limitations of this sub-second and ultra-sensitive imaging approach are discussed.

1. Introduction

Hyperpolarization techniques increase MR sensitivity by several orders of magnitude (1). In principle, many nuclear spins can be hyperpolarized including ¹⁹F, ¹H, ¹⁵N, and ¹³C. Abundant protons in biologically relevant molecules such as lactate have the largest magnetic moments and MR Larmor frequencies and therefore, should be superior from a detection perspective. However, biomedical use of hyperpolarized in vivo contrast agents require a significant time scale on the order of tens of seconds for contrast agent manipulation and in vivo delivery from the site of injection to the volume of interest (e.g. tumor) (2). The delay between synthesis of the hyperpolarized state and delivery in vivo is long compared to the typical lifetime of protons (~ 2 sec), therefore precluding direct detection of hyperpolarized protons. For example, if T₁ is 2 s, polarization decreases by ~400 fold after 12 s. As a result, long-lived nuclear spin sites, e.g. carboxyl ¹³C, are prepared by dissolution Dynamic Nuclear Polarization (DNP) (3) directly. In contrast, in parahydrogen induced polarization (PHIP) (4), the initial singlet state of nascent parahydrogen protons is converted to magnetization on a longer lived nuclear spin (e.g. carboxyl ¹³C) (5). Regardless of preparation, this heteronuclear storage nucleus is subsequently detected directly by ¹³C NMR to interrogate metabolism in vivo (6). Examples include 1-¹³C-pyruvate (7), 1-¹³C-succinate (8), 1-¹³C-glutamine (9) and others.

Despite nearly perfect polarization P → 100% achieved by dissolution DNP and PASADENA (Para-Hydrogen and Synthesis Allow Dramatically Enhanced Nuclear Alignment) (10), sensitivity could be further improved by reading out heteronuclear magnetization on protons. Protons are inherently more sensitive under conditions of identical polarization and T₂*, because of the other two gamma factors: greater resonant frequency and larger magnetic moment. This results in a combined enhancement of up to 15.7 fold, (γ_1H/γ_13C)², not accounting for T₂*, RF coil sensitivity, and subject noise.

Other, more practical factors, make ¹³C detection challenging. (γ_1H/γ_13C)² more gradient power is required to achieve the same spatial resolution as in conventional MRI. In addition, all clinical and preclinical MRI scanners are equipped with ¹H detection as a standard feature, while ¹³C requires specialized hardware and software. The latter can actually become a major impediment to the widespread use of hyperpolarized contrast agents.

Proton detection of ¹³C hyperpolarized contrast agents can solve the above challenges by the use of pulse sequences requiring only RF ¹³C excitation. The latter requires significantly less demanding hardware and can potentially provide more than an order of magnitude sensitivity gain. This additional sensitivity gain would be helpful even in hyperpolarized MR, because it can improve spatial resolution, decrease the dose of contrast agents, and therefore potentially allow for deeper penetration of metabolic fluxes.

Candidate polarization transfer sequences for reading out hyperpolarized heteronuclear magnetization can be found from high-resolution NMR, such as DEPT (Distortionless Enhancement by Polarization Transfer), INEPT (Insensitive Nuclei Enhanced by Polarization Transfer), and others. Here, conventional INEPT was chosen (11), and specifically a refocused INEPT sequence, which can theoretically provide detection sensitivity enhancements of up to (γ_1H/γ_13C)², where γ_1H and γ_13C are gyromagnetic ratios of detection and polarization storage nuclear spin sites. The INEPT sequence was chosen due to its relative simplicity and known utility in clinical research (12).

The concept of a more sensitive, indirect proton detection was demonstrated for highly polarized ¹³C PHIP contrast agents (13) as well as DNP polarized contrast agents spectroscopically (14) and more recently using fast imaging sequences (15). However, the vast majority of these experiments were performed in D₂O to suppress background signals from water and other protons. DNP coupled with polarization transfer techniques have been shown to effectively transfer polarization in a multidimensional framework (16–18) including a recent study with DNP hyperpolarized ¹³C-acetate in vivo (17). However, the demonstrated polarization transfer efficiency was rather low (17). It should be noted, however, that the choice of 1-¹³C-acetate in our in vitro studies was determined by the recent use of its hyperpolarized form for studying metabolism in vivo (19–21).

Translation of this approach in vivo requires sequence development for proton imaging of dilute millimolar proton spins in the presence of > 50 M water. The work presented here demonstrates sub-second proton imaging of ¹³C hyperpolarized contrast agent in water medium. A 25% polarized contrast agent at ~10 mM concentration was used to simulate in vivo conditions of highly polarized contrast agent with low concentration and decaying non-equilibrium hyperpolarization.

High polarization transfer efficiency and 2×2 mm² in-plane resolution were achieved. The prospects of pre-clinical translation, method potential and limitations for high and low magnetic fields are discussed from the perspective of metabolic imaging of cancer. Other polarization transfer approaches (22,23) can be used to further maximize the use of nonrenewable hyperpolarization.

2. Materials and Methods

2.1. Phantom Preparation for Boltzmann polarization studies

Sodium acetate and sodium 1-¹³C-acetate (99% atom ¹³C) were obtained from Isotec-Sigma-Aldrich Corporation (Miamisburg, OH), and concentrated to 4.3 M in H₂O and 99.8% D₂O respectively. Spherical phantom plastic containers were fully filled for each sample to a total solution volume of 2.8 mL.

2.2. Phantom preparation for hyperpolarized studies

2-hydroxyethyl-1-¹³C-acrylate-d_2,3,3 (HEA) was hydrogenated using a home-built PHIP polarizer with in situ detection capability operating at 5.9 mT. This device (Coffey, A. M. et al., unpublished results) is similar to our previously published polarizer operating at 47.5 mT using PASADENA (parahydrogen and synthesis allow dramatically enhanced nuclear alignment) (24,25). Production of PHIP hyperpolarized 2-hydroxyethyl-¹³C-propionate-d_2,3,3 (HEP) has been previously described by us (5,24–26) and others (27). Briefly, parahydrogen gas was produced with a custom lab-built generator by pulsing ambient ortho-hydrogen into an iron oxide catalyst filled chamber at 14–20 K (26). This enriched parahydrogen gas was then added to unsaturated molecular precursors using water-soluble rhodium(I) based molecular catalysts (28,29) in a high pressure chemical reactor. 1,4-bis[(phenyl-3-propanesulfonate)phosphine]butane, disodium salt (717347, Sigma-Aldrich-Isotec, OH, USA) and bis (norbornadiene) rhodium (I) tetrafluoroborate (45–0230, Strem Chemicals, MA, USA) were used for hydrogenation catalyst preparation in water medium as described earlier (5). PHIP precursor, HEA, was added to the aqueous solution after catalyst preparation. The concentration for the hyperpolarized product, HEP, was 9.8 mM, with an average hyperpolarization level of 25% at the ¹³C site measured in situ. Approximately 4 mL of aqueous solution containing hyperpolarized HEP was produced and transferred to a plastic syringe in the fringe field of a 4.7 T MR scanner, where ¹³C T₁ = 50 s (30). The hyperpolarized material was injected into a 2.8 mL hollow polypropylene sphere near the bore of the scanner and placed inside the RF coil, with a total delivery time of ~50 s.

2.3. Spectroscopy and Imaging

All spectroscopy and imaging experiments were performed on a Varian 4.7 T animal imaging system using the VNMRJ version 3.2 software suite. The experiments were conducted with a custom-built 38 mm ID two-channel RF coil, dual-tuned to ¹H (200.25 MHz) and ¹³C (50.25 MHz). All the experiments used the shim gradient values obtained from shimming a 2.8 mL sample of deionized water in a plastic spherical phantom container, resulting in a half-height line width of 4 Hz. Once the magnetic field was shimmed, the shim sample was swapped with the experimental samples, with care taken to replicate the sample position in the RF coil and the RF coil position in the magnet as used for shimming so that the magnetic field optimizations were accurately reflected across all the samples. Phantom replacement was necessary for hyperpolarized experiments, because the lifetime of hyperpolarization prevented shimming on individual samples.

All ¹H and ¹³C spectroscopic experiments consisted of a single acquisition with 1 s of acquisition time, 10 kHz spectral width, 10,000 complex points, and calibrated 90° excitation RF pulse. The INEPT delay for sodium acetate was τ_INEPT = 42 ms and the refocusing delay was τ_refocus = 16 ms for ¹H to ¹³C polarization transfer in the ¹³C-CH₃ molecular framework with J²_CH = 6.0 Hz (11). The INEPT delay for sodium acetate was τ_INEPT = 17 ms and the refocusing delay was τ_refocus = 35 ms for one ¹³C to three protons polarization transfer in the ¹³C-CH₃ molecular framework with J²_CH = 6.0 Hz. For hyperpolarized HEP, the polarization transfer from ¹³C to two protons derived from parahydrogen in the ¹³C-CH-CH molecular framework used delays of τ_INEPT = 20 ms and τ_refocus = 16 ms as determined previously (13) based on prior knowledge of the spin-spin couplings (27). In all imaging and spectroscopic experiments, the generic refocused INEPT sequence was augmented by τ_-90y at the end of the second refocus delay as shown in Fig. 1a on the detection channel to prepare magnetization along the z-axis for subsequent detection with frequency selective RF pulses using spectroscopic or imaging acquisition.

Figure 1. ¹³C Spectroscopy and Imaging of sodium 1-¹³C-acetate.

(a) Refocused INEPT sequence. (b) ¹³C reference spectrum of a thermally polarized 2.8 mL phantom of 4.3 M sodium 1-¹³C-acetate in water using a single scan acquisition, and (c) FSSFP projection image. (d) ¹³C INEPT spectroscopy and (e) INEPT enhanced imaging of sodium 1-¹³C-acetate using the sequence shown in (a). All the images were acquired with TR = 2 ms, TE = 1 ms, a field of view (FOV) of 96×96 mm², and an in-plane resolution of 3×3mm², with the same contour scaling levels. The following INEPT delays were used τ_INEPT = 42 ms and τ_refocus = 16 ms for polarization from ¹H→¹³C.

All imaging sequences used Varian’s version of 2D FSSFP with single slice acquisition. Slice thickness was set to 60 mm to encompass the entire phantom. 90° acquisition flip angle and 50 kHz spectral width were used for ¹³C and ¹H imaging. Other imaging parameters such as TR/TE varied depending on the frequency selective excitation pulses. All ¹³C FSSFP images were acquired with 96×96 mm² FOV with 3×3 mm² spatial resolution with τ_90° = 52.5 μs at 51.9 W and TR = 2.00 ms, TE = 1.00 ms with total acquisition time of 64 ms. ¹H τ_90° was 57 μs and ¹H τ_180° was 114 μs at 51.9 W for all non-frequency selective RF pulses. ¹H frequency selective RF pulses were ¹H τ_90° = 4.35 ms at 50.7 mW. Refocused INEPT polarization transfer blocks did not use frequency selective RF pulses. When ¹H frequency selective RF pulses were used during imaging, TR/TE increased from 1.91/0.95 ms (no frequency selection) to 6.20/3.10 ms (with frequency selection) respectively. All ¹H FSSFP images were acquired with 64×64 mm² FOV with 2×2 mm² spatial resolution with total imaging time of 61.1 ms (no frequency selection) and 198 ms (with frequency selection). Refocused INEPT polarization transfer imaging experiments utilized the same delays as described in the spectroscopy experiments.

3. Results

3.1. Heteronuclear Polarization From Three Methyl Protons to ¹³C in Isolated Methyl Group, Sodium 1-¹³C-Acetate Phantom in D₂O: Imaging and Spectroscopy

The experimental polarization enhancement ε_exp can be calculated as the ratio of S_INEPT (see Supporting Information for details) and S_thermal, the signal intensities of the corresponding refocused INEPT and thermal experiments:

$${\epsilon }_{\mathit{exp}}=\frac{s_{\mathit{INEPT}}}{s_{\mathit{thermal}}}$$ (2)

Spectroscopic detection resulted in ε_exp = 3.23 for INEPT transfer from ¹H→¹³C. This translates to 70% polarization transfer efficiency, η = ε_exp/ε_theory, for the refocused INEPT sequence when compared to the theoretical enhancement value, 4.59 (Supporting Information).

When refocused INEPT was coupled with the FSSFP fast imaging sequence at 4.7 T, Fig. 2, the magnitude of the INEPT enhanced image was increased with ε_exp = 3.99, corresponding to η = 87%, Figs. 1d and 1e. Imaging values of ε_exp and η were different from those of spectroscopic acquisition due to the relatively low SNR of ¹³C images acquired with thermal levels of ¹³C polarization.

Figure 2. INEPT Enhanced ¹H FSSFP sequence diagrams.

(a) The RINEPT-FSSFP RF pulse sequence contains a 2 s presaturation pulse (Presat), a refocused INEPT block (RINEPT), and FSSFP imaging block using a low power frequency selective Gaussian pulse for detection. τ_INEPT represents the INEPT delay, while τ_refocus is the refocusing delay, (b) Double RINEPT-FSSFP RF pulse sequence with extra RINEPT block for prepolarization of ¹³C from more polarized protons of thermally polarized phantoms in sodium 1-¹³C-acetate.

3.2. ¹H Background Suppression Using Multiband RF Excitation Pulses and ¹H Pre-saturation: Imaging and Spectroscopy

The approach of using water presaturation along with frequency selective Gaussian RF pulses was adopted to test water suppression in the phantom sample containing 2.8 mL of water, Fig. S1 (see Supporting Information for details). A 2 s presaturation pulse block (square RF pulse block at 1.6 W) coupled with a 50 mW Gaussian detection pulse was applied with an offset frequency of 750 Hz upfield from the water frequency. Spectroscopically, this approach resulted in 99.88% reduction of the background water signal for the water phantom when compared with a baseline single pulse MR spectrum (Figs. S1a and S1b), as determined through signal integration. Additionally, when the water suppression scheme was coupled to FSSFP, the resulting MR images produced the same result (Figs. S1c and S1d). The intensity of the water image was diminished from 101.5 a.u. without water suppression to 1.36 a.u. with suppression.

3.3. Proton Indirect Detection in Sodium Acetate Phantoms: Imaging and Spectroscopy

To examine the ability to indirectly detect protons from stored ¹³C magnetization, a series of experiments were conducted using refocused INEPT coupled with the frequency selective water suppression method discussed above. A single scan MR spectrum and MR image of a 1-¹³C-acetate phantom in water provided a baseline for examining ¹³C polarization transfer efficiency and water suppression. With a peak integral value of nearly 1.19•10⁸ a.u., the water peak from the baseline spectrum (Fig. 3a) overshadowed the acetate signal, 1.26•10⁷ a.u., and resulted in an image where the voxel intensities were dominated by water proton magnetization. Next, ¹H signal from acetate methyl protons was detected after polarization transfer from pre-polarized ¹³C₁ (see above). The sequence (Fig. 2) started with multiband proton presaturation followed by the refocused INEPT sequence block (¹³C₁→¹H), and ended with either spectroscopic or imaging ¹H detection using a frequency selective Gaussian shaped RF pulse as described above. The multiband proton suppression block was different from the one used in Fig. 2 in its use of a series of 50 ms-long RF pulses of identical amplitude but frequency alternating between the resonant frequencies of water and the methyl peaks of 1-¹³C-acetate. The resulting spectrum (Fig. 3c) showed nearly full suppression, 99.83%, of the water peak, while maintaining 40% of the baseline acetate signal integral value. This trend was also confirmed in a follow-up study of a 1-¹³C-acetate phantom in D₂O (Fig. 3e). Indirect detection of ¹³C magnetization using proton MRI images with multiband ¹H presaturation and Gaussian RF excitation pulse is shown in Figs. 3d and 3f with SNR of 144.7 and 197.2 for 1-¹³C-acetate in H₂O and 1-¹³C-acetate in D₂O respectively. Additionally identical spectroscopy and imaging studies were conducted on a sample of sodium acetate with natural abundance enrichment level of ¹³C and the same concentration and volume as used in both the experiments involving ¹³C sodium acetate in water and in D₂O. By testing a natural abundance sample with the INEPT enhanced pulse sequence, the resulting spectrum and image (Figs. 3g and 3h) with little to no signal assured that the sequence is performing properly and that the resulting images (Figs. 3d and 3f) were not from background water protons. The natural abundance sample contains vanishing quantity of ¹³C spin labels, ~1/91^th of that in Figs. 3c–f, which is likely responsible for a small signal seen in Fig. 3g.

Figure 3. Indirect ¹H Spectroscopic and Imaging Detection Using INEPT Enhancement.

Single scan ¹H detection of a 2.8 mL phantom of 4.3 M sodium 1-¹³C-acetate in water (a and b) without background multiband ¹H suppression (Off) or INEPT enhancement (Off). Indirect ¹H detection enhanced by INEPT for spectroscopy, and INEPT enhanced FSSFP for imaging for 2.8 mL phantoms of 4.3 M ¹³C-sodium acetate in water (c and d) and in D₂O (e and f), and also 4.3 M natural abundance sodium acetate in D₂O (g and h). A small amount of thermal naturally abundant acetate signal is detected even with background ¹H suppression and can be observed in the inset of (g) and faintly in the image (h). All the images were acquired using TR = 6.20 ms, TE = 3.10 ms, FOV = 64×64 mm², and an in-plane resolution of 2×2 mm².

From a theoretical perspective, for 1-¹³C-acetate polarization transfer from ¹³C→¹H can be expressed with the equation:

$$P_{1H}^{\mathit{INEPT}}=sin(2J_{HC}{\tau }_{\mathit{INEPT}})•{cos}^2(2\pi J_{HC}{\tau }_{\mathit{INEPT}})•sin(2\pi J_{HC}{\tau }_{\mathit{refocus}})•P_{13C}$$ (3)

For ¹³C-sodium acetate, the J-coupling between ¹³C₁ and ¹H is 6 Hz, and when τ_INEPT =17 ms and τ_refocus =35 ms, P_1H reaches the maximum value of 0.372•P_13C. The value of P_13C was boosted by a refocused INEPT ¹H → ¹³C pre-polarization of ¹³C as described above. This improves P_13C, which without pre-polarization would be 0.251 that of P_1H at equilibrium.

The experimental polarization transfer enhancement for indirect ¹H detection can be calculated similarly to Eq. 2 with modification accounting for artificially boosted ¹³C polarization above the thermal level used in our experiments:

$${\epsilon }_{\mathit{exp}}=\frac{P_{\mathit{INEPT}}^{1H}}{P^{13C}},$$ (4)

where P_13C is nuclear spin polarization of ¹³C immediately before the refocused INEPT transfer procedure and $P_{\mathit{INEPT}}^{1H}$ is experimentally detected ¹H polarization calculated as follows:

$$P_{\mathit{INEPT}}^{1H}=\frac{s_{\mathit{INEPT}}^{1H}}{s_{\mathit{thermal}}^{1H}}•P_{\mathit{thermal}}^{1H},$$ (5)

where $S_{\mathit{INEPT}}^{1H}$ is imaging or spectroscopic signal detected after INEPT transfer, $P_{\mathit{thermal}}^{1H}$ is the equilibrium thermal polarization of ¹H spins, and $S_{\mathit{thermal}}^{1H}$ is corresponding imaging or spectroscopic signal detected using identical acquisition protocol as $S_{\mathit{INEPT}}^{1H}$. For the refocused INEPT transfer from ¹³C → ¹H between the ¹³C₁ and three methyl protons of acetate, ε_exp is 0.367 for spectroscopic acquisition, 0.320 for imaging detection in D₂O and 0.167 for imaging detection in H₂O corresponding to polarization efficiency η values of 98.6%, 85.9% and 45.1% respectively.

3.4. ¹³C and Indirect ¹H Detection of Aqueous ¹³C Hyperpolarized Contrast Agent: Imaging and Spectroscopy

Single scan FSSFP ¹³C imaging was performed to determine the hyperpolarization levels for ¹³C HEP (Fig. 4a). The ¹³C hyperpolarization level was calculated to be %P = 5.3% at 4.7 T (Fig. 4a) after ~50–60 seconds of delivery time from the polarizer to the MRI scanner using Eq. 6, which compares FSSFP images of hyperpolarized ¹³C contrast agent (Fig. 4a) and reference compound (Fig. 1e) with known values of polarization using the same imaging parameters according to:

$$\%P_x=\epsilon ·P_X^0·100\%=\frac{s_X·{\chi }_{\mathit{REF}}}{s_{\mathit{REF}}·{\chi }_X}{P}_{\mathit{REF}}·100\%,$$ (6)

where $P_X^0$ is the nuclear polarization at equilibrium at 298 K and 4.7 T, and the enhancement ε is determined using the second part of this equation with measured signal, S_REF (reference) and S_X (hyperpolarized), and quantity information from reference compound χ_REF and the hyperpolarized agent χ_X of the same X nucleus, where X = ¹H or ¹³C. With B₀ = 4.7 T and T = 298 K, $P_{13C}^0$ is 4.055*10⁻⁶ and $P_{1H}^0$ is 1.61*10⁻⁵. The 5.3% ¹³C hyperpolarization of HEP detected at 4.7 T correlates to an enhancement of 13,200 fold when compared to equilibrium $P_{13C}^0$ of 1-¹³C-acetate at 4.7 T. The hyperpolarized ¹³C image of HEP had maximum SNR of 177. To demonstrate the relatively short lifetime of ¹³C HEP hyperpolarization, a ¹³C single slice FSSFP experiment (Fig. 4b) was followed by an identical imaging acquisition 90 s after the first scan, resulting in no detectable signal in the image, signifying a full decay of ¹³C hyperpolarization in HEP. 90 s delay corresponds to approximately 2xT₁.

Figure 4. Hyperpolarized ¹³C FSFFP Imaging.

(a) ¹³C detection of ¹³C hyperpolarized HEP using sub-second FSSFP, (b) FSFFP image of the “cold” contrast agent 90 s after acquisition shown in (a). Hyperpolarized ¹³C images were acquired with TR = 2 ms, TE = 1 ms, a field of view (FOV) of 96×96 mm², and an in-plane resolution of 3×3 mm². No extrapolation or zero filling was applied to images or spectra, and images are shown in their raw resolution.

For indirect ¹H imaging of ¹³C hyperpolarized HEP upon delivery to the 4.7 T scanner magnet, polarization was transferred from ¹³C to the nascent protons via refocused INEPT using the same imaging sequence with proton multiband presaturation and selective RF excitation pulses as was used in the acetate studies above. A 2×2 mm² in-plane resolution (Fig. 5a) was obtained for hyperpolarized HEP using indirect proton detection. Total experimental time for proton imaging was on the order of 3 s, consisting largely of the multiband proton presaturation period of 2 s. The ¹H hyperpolarized image had an SNR of 256 corresponding to 1.41% (corresponding to ε_exp = 900) hyperpolarization of total polarized ¹H intensity per 0.0098 mM effective concentration, similar to previous studies (13). As with the ¹³C direct detection experiments using ¹³C hyperpolarized HEP, the lifetime of the hyperpolarization was short and hyperpolarization decayed after 90 s to undetectable levels in proton indirect imaging (Fig. 5b).

Figure 5. ¹H Imaging of ¹³C Hyperpolarization.

(a) Indirect FSSFP ¹H imaging of polarization transferred from ¹³C nuclei of ¹³C hyperpolarized HEP to neighboring protons using INEPT, (b) ¹H image of “cold” state, no detectable signal is observed in the indirect ¹H FSFFP image. Indirect proton images were acquired using TR = 6.35 ms, TE = 3.17 ms, FOV = 64×64 mm², and an in-plane resolution of 2×2 mm². No extrapolation or zero filling was applied to images, and images are shown in their raw resolution.

Examining the theoretical polarization transfer efficiency η in HEP from ¹³C to ¹H was not pursued because in addition to the methyl protons of interest, HEP contains a methylene proton as well. However, we note that 1.41% ¹H polarization was achieved after polarization transfer from 5.3% polarized ¹³C in HEP corresponding to 27% polarization transfer efficiency η calculated as P_1H/P_13C.

4. Discussion

4.1. Polarization transfer efficiency

When comparing polarization transfer efficiency from ¹H to ¹³C in 1-¹³C-acetate, η differed in the spectroscopic and imaging detection approaches. The difference between 70% and 87% respectively was likely a manifestation of large error in η of imaging, because of the relatively low SNR of the reference thermal ¹³C image, Fig. 1c. Therefore, the η value derived from the H→¹³C spectroscopic study was used in further calculations of ¹³C→H polarization ε_exp and η.

¹³C→H polarization transfer efficiency η was lower in the imaging studies of 1-¹³C-acetate/D₂O phantom compared to corresponding spectroscopic studies, 85.7% vs. 98.6%. Moreover, η decreased further, down to 45.1% in the case of the 1-¹³C-acetate/H₂O phantom. A likely explanation for such significant decreases in the imaging studies and especially for the case of 1-¹³C-acetate/H₂O is the T₂ effect, because RF pulses are applied every 6 ms and non-equilibrium ¹H polarization transferred from ¹³C is mixed with thermal polarization of protons with relatively short T₂. A potential improvement can be achieved through imaging acceleration using faster acquisition. However, frequency selective RF pulses as implemented here represented the largest fraction of total acquisition time. Therefore, more optimal design and implementation of RF excitation pulses with spectral selectivity would potentially improve the overall SNR of indirect proton imaging detection. The latter can be achieved through a variety of solutions. (i) Higher field MRI systems such as ≥9.4 T increase chemical shift dispersion leading to shorter RF pulses with identical frequency selectivity. (ii) Sinc RF pulses or other waveforms can also potentially lead to shorter RF excitation pulses with similar or better frequency selectivity. (iii) Low field MRI indirect detection potentially allows for significantly lower water background and could potentially eliminate the need for frequency selective RF pulses. For example, it has been shown that high detection sensitivity can be achieved at 0.0475 T, a field that has 100 times less polarized water (25). In that case, simple proton presaturation would perform better and may be sufficient. Regardless of the approach to improve the apparent η of ¹³C→H for indirect proton imaging, it will significantly increase the detection sensitivity presented here by several fold.

A similar trend in η for ¹³C→H polarization transfer in hyperpolarized HEP was observed in corresponding imaging experiments, Table. 1. η of the imaging study was only 27% and clearly has room for potential improvement.

Table 1.

Summary of the results. Note the error is ±1% unless otherwise noted.

Transfer	Spatial Resolution	SNR	13C Polarization	1H Polarization	εexp	εtheory	η	J (Hz)
Sodium 1-13C-acetate/D2O
13C→H	Spectroscopy	6209	1.30*10−5	4.75*10−6	0.367	0.372	98.6%	6.0
H→13C	Spectroscopy	960	1.30*10−5	1.61*10−5	3.23	4.59	70.3%	6.0
13C→H	2×2 mm2	197	1.30*10−5	4.14*10−6	0.319	0.372	85.7%	6.0
H→13C	3×3 mm2	29.8	(1.61±0.3)*10−6	1.61*10−5	3.99±0.6	4.59	87±15%	6.0
Sodium 1-13C-acetate/H2O
13C→H	2×2 mm2	145	1.30*10−5	2.17*10−5	0.167	0.372	45.1%	6.0
13C hyperpolarized HEP
13C→Hb	2×2 mm2	256	(5.3±0.5)*10−2	(1.41*±0.14)*10−2	900±90	-	-	5.6, 7.3
13C	3×3 mm2	177	(5.3±0.5)*10−2	-	13,200±1,300	-	-	-

4.2. Spatial resolution improvement

Despite reduced apparent η values in the ¹³C→H imaging study with 1-¹³C-acetate in water, proton imaging SNR was 145 for a 2×2 mm² in-plane resolution. This correlates with a SNR of 29.8 and 3×3 mm² in-plane resolution where ¹³C was imaged with the same level of ¹³C polarization. We note that the RF coil was not particularly optimized for ¹H detection. Nevertheless, indirect proton detection yielded 11.0 times the SNR per unit of volume. This proves the underlying hypothesis that indirect proton detection of hyperpolarized compounds is a more sensitive means of imaging. To a certain extent, this rather high value of sensitivity enhancement is enabled by three protons rather than one ¹³C carbon, even under conditions of relatively low Eff, and ε_exp ≪ 1. It should also be pointed out that the overall experimentally observed sensitivity enhancement factor of 11.0 described above also reflects ¹H sensitivity loss due to the implemented RF excitation pulses. Approximately 60% signal loss was observed in the implementation of frequency selective pulses during ¹H imaging compared to square RF pulses with τ_90° = 57 μs. This shortcoming should be addressed in future studies. However, if properly addressed, SNR gains in excess of 20 fold are potentially achievable.

4.3. Limitations and future perspectives, and subject RF noise

The ¹³C→H imaging study with ¹³C hyperpolarized HEP in water also yielded relatively low η of 27%. This compares to η = 45.1% in the ¹³C→H imaging study with 1-¹³C-acetate in water. The comparison of imaging SNR of direct ¹³C and indirect ¹H studies, Table 1, shows that indirect detection yielded 3.25 times more SNR. The significant discrepancy with the factor of 11.0 described above can be explained by lower η value and three acetate protons per molecule compared to the one proton per chemical group in perdeuterated PASADENA addition product of HEA (HEP). Moreover, polarization transfer delays could likely be further optimized to improve experimental transfer yields.

It should also be pointed out that proton detection is conducted at significantly higher frequency, which is one of the contributing factors in the (γ_1H/γ_13C)² expression, not accounting for T₂, RF coil sensitivity and subject RF noise. While RF coil sensitivity can be optimized for the detection frequency of interest, T₂ (31) and subject RF noise (32) are fundamentally unavoidable, but have frequency dependence. Therefore, it is likely that the efficiency of improved proton detection can decrease with transition to in vivo studies, larger sample and subject sizes and higher resonance frequencies. While the increased body noise at greater proton frequencies is a clear limitation of this method (33), we point out that it can be significantly alleviated in the low-field MRI regime, where RF losses and body noise on the clinical subjects is negligible (34). As a result, low-field MRI of hyperpolarized contrast agents in humans is an attractive approach from sensitivity perspective.

Specific absorption rate (SAR) is undesirable, because it leads to heat deposition from RF pulses. The pulse sequence described here is indeed RF pulse intense. However, the polarization transfer block uses only a few high power RF pulses. The remainder of RF pulses, corresponding to excitation RF pulses of FSSFP, are extremely low power (~50 mW). Moreover, the sequence is played only once in less than 1 second. Therefore, SAR is not expected to be an impediment for pre-clinical translation of the presented work at magnetic fields of 4.7 T and lower. However, the clinical translation and high-field preclinical translation will likely result in increased RF pulse power, which could raise SAR concerns. This will likely be exacerbated by 3D applications, where additional RF pulses are needed for 3D encoding. However, it should be stressed that the sequence is ultra-short and power deposition will be limited to ~ 1 second. In case if SAR will exceed the safe limits, a potential solution is to replace FSSFP imaging block by less RF-pulse power intense gradient echo sequences, where a small tipping angle RF pulse is used, i.e. only a small fraction (2–5%) of a full 90° RF pulse is utilized.

The RF pulse width used in frequency selective FSSFP component was > 4 ms, which is sufficiently long for potential clinical translation of this sequence. Because the polarization transfer delays are long due to relatively small heteronuclear spin-spin coupling of isolated ¹³C sites, the RF pulses of polarization transfer block can also be potentially increased as needed (up to the width of frequency selective RF pulses of > 4 ms) for clinical translation.

While the body noise has a negligible contribution to overall NMR noise in low magnetic field (34), the body noise dominates the overall noise at higher frequencies. Because there is more body noise at high frequencies as compared to in vitro phantoms, there will be more noise in proton MR compared to ¹³C MR at the same magnetic field. Therefore, the observed SNR gains for indirect proton detection will be reduced in vivo at high magnetic fields and frequencies. However, hyperpolarized MRI sensitivity does not gain from higher fields in general and therefore, translation of this approach will be more beneficial at lower fields, i.e. < 4.7 T.

¹³C→H imaging studies used proton background suppression, taking advantage of multiband pre-saturation RF pulses and frequency selective RF excitation pulses, which resulted in an overall increase of experimental time and more importantly imaging time. For example, imaging time increased from 64 ms (no frequency selective pulse) to 192 ms (with frequency selective pulses). As mentioned above, most of the imaging time was spent on RF excitation pulses. While the latter can be potentially decreased by smarter RF pulse design and implementation, the alternative solution is to increase spatial resolution while decreasing the RF pulse width and maintaining the same background level. For example, the relatively high SNR of 256 with 2×2 mm² in-plane resolution can be traded for improved spatial resolution with 1×1 mm², which is also desirable from a cancer imaging perspective, while maintaining relatively high and certainly acceptable levels of SNR.

We anticipate that this work can be readily translated to in vivo imaging of hyperpolarized contrast agents. Indirect imaging of hyperpolarized 1-¹³C-lactate as a metabolic product of injected hyperpolarized 1-¹³C-pyruvate is particularly attractive using the demonstrated strategy. Because of larger chemical shift separation in the ¹³C dimension between 1-¹³C-pyruvate and 1-¹³C-lactate > 10 ppm, 1-¹³C resonance of lactate can be selectively excited (35) for polarization transfer to its three methyl protons, while retaining hyperpolarized magnetization of 1-¹³C-pyruvate. Additionally, lactate similarly to the acetate presented here, has three methyl protons and similarly large SNR gains are expected, as much as an order of magnitude as presented here. This can be translated to significantly improved spatial resolution or improved sensitivity of hyperpolarized MR, which are still lacking and can benefit from the improvements described here.

Extension of this method to 3D hyperpolarized imaging (36) can be potentially realized, but challenges need to be addressed including the required short imaging time due to imposed T₂ limitations of protons. These limitations can be addressed by decreasing the acquisition time necessary to scan each line of k-space as well as implementing compressed sensing.

A single proton image was recorded for ¹³C hyperpolarized contrast agent here, while multiple acquisitions may be desired for kinetics studies of metabolism in vivo. The latter can be realized by the use of partial polarization transfer from ¹³C to ¹H (23), where only a fraction of long-lived ¹³C hyperpolarized is consumed for a single 2D or potentially 3D image, and multiple polarization transfer can enable acquisition of multiple proton images of ¹³C hyperpolarized contrast media.

5. Conclusion

We have used frequency selective FSSFP method for indirect proton imaging of the aqueous ¹³C hyperpolarized contrast agent HEP. Indirect proton imaging allowed for notable sensitivity improvement. Frequency selective RF pulses combined with proton multiband pre-saturation significantly decreased water and other proton background signals to an acceptable level for quantitative imaging, demonstrating that ¹³C hyperpolarized MR can significantly benefit from indirect proton sub-second imaging.

SNR gains of up to 11 fold from proton imaging versus direct ¹³C imaging were demonstrated with a sodium 1-¹³C-acetate phantom in water, where three methyl protons were used for detection. SNR gain of 3.25 for hyperpolarized HEP was somewhat lower. The demonstrated method can be applied to a number of other hyperpolarized contrast agents including hyperpolarized 1-¹³C-pyruvate and 1-¹³C-lactate, 5-¹³C-glutamine, succinate and others. The demonstrated SNR gains can be further improved significantly by (i) implementation of more suitable frequency selective RF pulses and better chemical shift dispersion at high field or (ii) use of low field MR imaging. We speculate that SNR can be improved by a factor of 2 to 5.

The demonstrated method can potentially be extended to 3D and in vivo application. Moreover, this method of indirect proton detection of hyperpolarized contrast agents solves an important challenge of ¹³C hyperpolarized MR by eliminating the need for ¹³C detection. Conventional proton detection is universally available on clinical MRI scanners and can be utilized for detection of hyperpolarized contrast agents while requiring only ¹³C RF excitation during the polarization transfer sequence. Furthermore, ¹H imaging is significantly less demanding from the perspective of gradient power requirements.

Figure 6.

Molecular structures and polarization transfers.

ABBREVIATIONS

PHIP

parahydrogen induced polarization

HEP

2-hydroxyethyl 1-¹³C-propionate-d_2,3,3

HEA

2-hydroxyethyl 1-¹³C-acrylate-d_2,3,3

INEPT

insensitive nulcei enhanced by polarization transfer

RINEPT

refocused insensitive nuclei enhanced by polarization transfer

DEPT

distortionless enhancement by polarization transfer

APT

attached proton test

FSSFP

fast stead-state free precession

FOV

field of view

η

efficiency

PASADENA

Para-Hydrogen and Synthesis Allow Dramatically Enhanced Nuclear Alignment

DNP

dynamic nuclear polarization