Enabling Clinical Technologies for Hyperpolarized Xenon-129 MRI and Spectroscopy

Abstract

Hyperpolarization is a technique that can increase nuclear spin polarization with the corresponding gains in nuclear magnetic resonance (NMR) signal by 4–8 orders of magnitude. By applying this process to biologically relevant samples, they can be used as exogenous MRI contrast agents. A technique called Spin Exchange Optical Pumping (SEOP) can be applied to hyperpolarize noble gases such as ¹²⁹Xe. Hyperpolarized ¹²⁹Xe is poised to revolutionize clinical lung imaging, offering a non-ionizing, high contrast alternative to CT imaging and conventional proton MRI. Moreover, CT and conventional proton MRI report on lung tissue structure but provide no functional information although substantial progress has been made recently using ultrashort (UTE) TE MRI to obtain surrogate ventilation metrics. On the other hand, by breathing hyperpolarized ¹²⁹Xe gas, strong MRI signal can be obtained to produce functional lung images reporting on lung ventilation, perfusion and diffusion with 3D readout in seconds. In this Review, the physics of SEOP is discussed and the different production modalities are explained in the context of their clinical application. We also briefly compare SEOP to other hyperpolarization methods and conclude this paper with the biomedical outlook for hyperpolarized ¹²⁹Xe applications to lung imaging and beyond.

Entry for the Table of Contents

Hyperpolarized ¹²⁹Xe has a potential to revolutionize clinical imaging, offering a non-ionizing contrast of organ function complementary to the CT imaging and to conventional MRI. Here, the spin physics of SEOP is discussed and the different production modalities are explained in the context of their clinical application. We conclude with the biomedical outlook for hyperpolarized ¹²⁹Xe applications to lung imaging and beyond.

Graphical Abstract

1. Introduction

Magnetic resonance imaging (MRI) is a versatile and commonly used clinical imaging technique for producing highly detailed soft-tissue images. The principle of MRI relies on nuclear magnetic resonance (NMR) of nuclei within the body as a result of applied magnetic fields and radio frequency (RF) pulses. The NMR signal from the nuclei depends on the nuclear spin polarization factor, P, which gives a measure of the extent of alignment of the nuclei’s spins. In order to obtain a greater signal, P must be maximized. In conventional NMR, this is given by:

$$P\approx \frac{\hslash \gamma B_0}{2k_{B}T}$$ (1)

Therefore, in a conventional 3 T clinical MRI scanner with a temperature of 300 K, the maximum polarization value for ¹H is ≈1×10^-5. This low P is the result of very few nuclei being aligned with the magnetic field, and therefore only a small number of nuclei produce a signal during NMR investigations. The low P poses a significant fundamental problem: only with highly abundant nuclei, such as ¹H of water and fat, can suitably strong signals still be obtained for imaging purposes – albeit with spatial-temporal resolution lower than that of Computed Tomography (CT).^[1] MRI is challenging with more-dilute compounds and other magnetically active nuclei, and therefore low P fundamentally reduces the versatility of MRI and NMR.

To overcome this limitation, a technique to enhance the nuclear spin polarization factor, called hyperpolarization, can be used to increase NMR sensitivity by 4–8 orders of magnitude.^[2] The improved sensitivity occurs due to an increased level of alignment of the hyperpolarized (HP) spins, resulting in corresponding gains of the NMR signal. By increasing the signal generated, the uses of MRI can be expanded, allowing clinically useful data to be created in many more scenarios. The stronger signal means investigations can be performed in lower magnetic fields with shorter scan duration times, thereby increasing the versatility of the approach. As a result of the stronger signals, HP molecules have been used for investigations into metabolism,^[3] cancer,^[4] lung function,^[5] and others (Figure 1a).

Figure 1.

Thermal equilibrium polarization produces a small excess of spins in one state. When the sample undergoes hyperpolarization a large excess of spins exists in one state producing a considerably larger signal since more spins contribute.

An area of particular importance is the functional imaging of lungs using HP gases. This developing method offers a new clinical tool that is non-invasive and provides high quality images and data. This technique involves using a HP noble gas such as ³He and ¹²⁹Xe as a contrast agent. The HP gas is inhaled by the patient and diffuses within the lungs allowing for both detailed MRI images and quantitative information to be obtained.^[6]

Using HP gases offers several advantages in MRI compared to traditional methods. Most notably, by imaging an inhaled gas like ³He or ¹²⁹Xe, there is a higher density of NMR-active molecules in the lungs compared to traditional ¹H imaging, which improves the MRI images that can be obtained. Moreover, unlike proton MRI which reports on anatomical features of lung tissues, HP gas MRI reports primarily on void lung compartments. This allows imaging of lung function including gas ventilation, diffusion and perfusion (as discussed in Section 7). The functional lung images can be collected in a single breath hold, reducing the strain on the patient. This not only makes the imaging more convenient for the patient but increases the accessibility of the imaging technique. This is particularly useful for patients with pulmonary disorders who are the main target for the imaging technique as they cannot carry out long breath-holds and cannot be imaged for long periods of time. The first biological images using HP ¹²⁹Xe were obtained in 1994 by Albert et al.^[7] and using ³He in 1995 by Middleton et al.^[8] Historically, MRI’s of the lung were poor due to the weak signals in the ¹H images as a result of the low density within this organ. For this reason, HP ³He was proposed as an imaging technique to produce clinically useful lung images. Since ³He has the highest gyromagnetic ratio of the noble gases, γ (32.4338 MHz/T) compared to proton (42.5775 MHz/T), it was selected in order to allow for a strong signal to be produced during HP lung imaging.^[9]

Despite early work on HP ³He producing high spatial resolution and clinically useful images, the use of the gas was limited due to worldwide shortages^[10] resulting in cost increases that were deemed unsustainable for the envisioned use. Instead, ¹²⁹Xe was found to be a suitable alternative gas for clinical studies. While the gas suffers from a lower γ (11.7769 MHz/T)—and therefore a lower signal—than ³He, hyperpolarization is still able to impart a sufficiently strong signal for the images to be clinically useful. Moreover, the use of ¹²⁹Xe is advantageous compared to other gases due to its lipophilic properties, making it is soluble in barrier tissue, blood,^[11] and others.^[12] The absorption of the ¹²⁹Xe by various biological compartments (e.g., red blood cells) induces a frequency shift, allowing the absorbed and non-absorbed gas to be distinguished via spectroscopy.^[13] This aids the clinical properties of HP ¹²⁹Xe, offering not only high-quality images, but also quantitative information on lung function and gas perfusion rate.^[14] As a result, ¹²⁹Xe has been the focus of both clinical studies and technological developments to optimize the hyperpolarization process of the HP gas production for clinical use, aiming to obtain clinical grade HP gas in a time- and cost- efficient manner.

For noble gases, the most common hyperpolarization method is spin exchange optical pumping (SEOP).^[15] While the fundamental physics of SEOP is unchanged, the approach and technologies can vary significantly depending on the production method employed. HP ¹²⁹Xe can be prepared via two distinct methods to hyperpolarize and collect the product. In the first method, called continuous flow, the ¹²⁹Xe flows through the SEOP cell with a fixed rate before exiting the polarizer and can be cryo-collected for subsequent use. Alternatively, in a method called “batch mode” (or stopped-flow), the ¹²⁹Xe is loaded into the oven where the SEOP process occurs for a set amount of time before the process is stopped and the HP gas is removed. In this review, the development of the different production methods of SEOP will be discussed and their respective advantages and limitations explained. This will be compared with other emerging hyperpolarization production technologies. The promising biomedical outlook for the use of ¹²⁹Xe HP production technology will then be presented.

2. Spin-Exchange Optical Pumping (SEOP)

Optical pumping (OP) was first introduced by Kastler in 1950, who found that an alkali metal vapor’s electrons can be spin-polarized when in the presence of a magnetic field and circularly polarized light.^[16] While a pioneering discovery on its own, it was not until 1960 that it was found that the electron spin polarization possessed by the alkali metal vapor could be transferred to the nuclear spins of noble gases.^[17] This finding produced the SEOP process known today, allowing the production of highly polarized noble gases through the optical pumping of alkali metals. SEOP is composed of two stages. In the first step, a vaporized alkali metal, e.g. Rubidium (Rb), undergoes optical pumping through the application of circularly polarized light, thereby imparting electron spin polarization to the alkali metal. In the second step, the spin polarization of the alkali metal electrons is transferred to the nuclei of the noble gas of interest (e.g., ¹²⁹Xe) to produce a HP noble gas over time.

2.1. Step 1: Optical Pumping of Electron Spins

To perform SEOP, the vaporized alkali metal and noble gas are contained with an optically transparent cell, often with other buffer gases to assist with the hyperpolarization process. Outside the SEOP cell, a high-power laser is used to apply the circularly polarized light tuned to the alkali metal’s D₁ resonant frequency (at ca. 794.7 nm for Rb), which when absorbed by a Rb atom within the SEOP cell, excites its outer-shell electron from the ²S_1/2 (ground) to the ²P_1/2 (excited) state. Additionally, static magnetic field (typically a few millitesla) is applied along the beam direction of circularly polarized laser light. Within both the ground and excited states exists two sublevels denoted m_J = ±1/2 (neglecting the nuclear hyperfine splitting for simplicity). Thus, when a resonant photon is absorbed, the unpaired electron is excited into a specific m_J sublevel of the ²P_1/2 state as dictated by the conservation of the angular momentum, Figure 2. While in the excited state, the electrons of the Rb vapor become evenly distributed between the two sublevels as a result of collisions with other gas-phase species. This equalization results in the ground state sublevels being repopulated by relaxation at roughly equal rates. If the polarized light is continually applied to the Rb, there is a build-up of the population in a single ground state sublevel, usually reaching steady state in a fraction of a second. As a result, the alkali metal becomes electronically spin polarized, Figure 2.

Figure 2.

Optical pumping of Rb electron spins. a) Rb begins in the ground state, with electrons equally distributed between sublevels. Absorption of circularly polarized light excites electrons into the m_j=½ sublevel, conserving angular momentum. Collisional mixing equilibrates the excited sublevel populations and the rates at which the ground sublevels are repopulated by relaxation (b), but selective depletion of one sublevel by the laser leaves the vapor electronically spin-polarized. c) Rb/¹²⁹Xe Spin exchange collisions allow ¹²⁹Xe nuclear spin polarization to accumulate over time.

While most electron spins remain within the m_J=1/2 sublevel of the ground state, some return to the m_J=−1/2 sublevel via relaxation caused by non-angular-momentum-conserving collisions with other gases in the optical cell. As a result, the steady-state polarization achieved via optical pumping is determined by the balance of these competing factors, according to the relation:^[18]

$$\begin{array}{c}{P}_{Rb}\left(z\right)= \frac{{\gamma }_{op}\left(z\right)}{{\gamma }_{op}\left(z\right)+{\gamma }_{sd}}\end{array}$$

where $P_{Rb}\left(z\right)$ is the polarization of the Rb vapor at position z along the optical cell axis, ${\gamma }_{op}\left(z\right)$ is the optical pumping rate (which depends on the photon flux at that point in the SEOP cell) and ${\gamma }_{sd}$ is the Rb depolarization rate (a.k.a., “spin destruction” rate). Therefore, minimizing the spin destruction rate, and ensuring ${\gamma }_{op}\gg {\gamma }_{sd}$ allows for a near-unity Rb polarization to be achieved.

2.2. Step 2: Polarizing Xe Nuclei via Spin Exchange

The electronic polarization of the Rb can be transferred to the nuclear spins of ¹²⁹Xe in the gas mix via two types of gas-phase collisions: binary collisions between the Rb and ¹²⁹Xe atoms, and tertiary collisions involving a third body that result in the formation of Rb/Xe van der Waals “molecules”. When collisions occur between the two species, Fermi-contact hyperfine interactions allow the Rb electronic spins and the ¹²⁹Xe nuclear spins to exchange angular momentum.^[19] Although only a small fraction of these collisions yield successful spin exchange, over time a bulk nuclear spin polarization accumulates in the ¹²⁹Xe nuclear spins, according to:^[20]

$$\begin{array}{c}{P}_{Xe}\left(t\right)= \frac{{\gamma }_{se}}{{\gamma }_{se}+{\mathrm{\Gamma }}_{Xe}}〈P_{Rb}〉\left(1-e^{-\left({\gamma }_{se}+{\mathrm{\Gamma }}_{Xe}\right)t)}\right)\end{array}$$

where $⟨P_{Rb}⟩$ is the average Rb polarization within the SEOP cell, ${\Gamma }_{Xe}$ is the ¹²⁹Xe nuclear spin relaxation rate (1/T₁), e.g. due to collisions with the SEOP walls, and ${\gamma }_{se}$ is the rate of polarization transfer from the Rb to the ¹²⁹Xe via spin-exchange collisions.

To maximize the ¹²⁹Xe polarization attainable via SEOP, several factors should be optimized. Most notably, the polarization of the Rb vapor can be increased by minimizing the spin destruction rate (e.g., by keeping the Xe density low), or by increasing the Rb optical pumping rate by using a higher power laser. Additionally, the ¹²⁹Xe polarization can be increased by mitigating ¹²⁹Xe relaxation in the SEOP cell.

While the fundamental physics of course remains unchanged, the SEOP production methods and technologies vary significantly. HP ¹²⁹Xe can be prepared via many ways falling within two distinct categories to hyperpolarize and collect the HP product. In the first method, called continuous flow, the ¹²⁹Xe flows through the SEOP cell with a fixed rate before exiting the polarizer while being cryo-collected for subsequent use (Figure 3). Alternatively, in a method called “batch mode” or stopped-flow, the ¹²⁹Xe is first loaded into the SEOP cell for a set period of time before the process is stopped and the HP gas is removed. These two approaches are described in greater detail in Sections 3 and 4 respectively.

Figure 3.

Overall design of the original continuous flow (CF) ¹²⁹Xe hyperpolarizer.^[20] λ/4 refers to high-power polarizing beam splitter, and L1 and L2 lences to other beam expanding components of the optical train.

3. Continuous-Flow SEOP Production Method

3.1. CF SEOP: Design Considerations and Initial Efforts

¹²⁹Xe hyperpolarizer designs based on continuous flow (CF) polarization allow for an uninterrupted stream of xenon gas mixture to be passed through the SEOP cell.^[21] A gas mixture containing ¹²⁹Xe and Rb vapor is flowed through the heated SEOP cell and irradiated by circularly polarized light. After passing through the SEOP cell and undergoing hyperpolarization, the ¹²⁹Xe gas is cryo-collected using liquid N₂ to allow the HP ¹²⁹Xe to be concentrated in the solid state (Figure 3). By storing solid HP ¹²⁹Xe at a low temperature and within a high magnetic field, the HP state can be retained for a long period of time due to the temperature and magnetic field dependence of the nuclear spin-lattice relaxation rate, T₁ (many hours at 77 K).^[22] Furthermore, by cryo-collecting the HP ¹²⁹Xe, a higher density of HP product is produced by removing additional buffer gases that have been included in the gas mix (e.g., N₂ and ⁴He).^{[20, 23]} When sufficient quantities of HP ¹²⁹Xe have been accumulated, it is rapidly thawed into the gas phase and expanded into a storage vessel—such as a Tedlar bag—that can be transported to the MRI scanner (e.g., where the Xe can be inhaled by a subject). However, while the collection of the solid ¹²⁹Xe is useful for the removal of buffer gases, care must be taken with this step because the cryogenic separation and accumulation can otherwise lead to significant polarization losses. For example, Kuzma et al.^[24] showed that Xe polarization can be lost while warming the solid ¹²⁹Xe to near its melting point (161.4 K) because the T₁ can decrease dramatically, to the order of seconds. This problem can be mitigated by moving through phase transitions rapidly, in addition to applying a magnetic field^[22] during cryogenic separation to raise the solid-state ¹²⁹Xe T₁. While the relaxation rate of the HP gas is strongly dependent on the magnetic field, it was found that very little change in the T₁ occurs when a magnetic field greater than 500 G is applied^[25] (using, for example, strong permanent magnets placed near the Xe cryo-collection vessel).

In order for CF production to work effectively, the flow rate must be high enough to produce a sufficiently large quantity of HP ¹²⁹Xe for the desired application—e.g., human clinical imaging. However, for the ¹²⁹Xe polarization level to also be high, the spin exchange rate must be high enough to accommodate for the correspondingly low in-cell Xe residence times. High spin exchange rates can be achieved with high rubidium densities (i.e., at elevated SEOP cell temperature), but these high optical densities must be compensated with increased photon flux. Thus, high-power laser diode arrays (LDAs) were the primary enabling technology that led to the development of the first clinical-scale CF hyperpolarizer in 1996 by Driehuys et al. (Fig. 3).^[20] Indeed, this device included a laser system that could produce high power (and therefore high photon irradiance) in a lower cost, compact structure.^[26] The original LDAs had a much broader spectral output than (say) Ar⁺/Ti:Sapphire lasers, as measured by the full width half maximum (FWHM). This characteristic meant a lower proportion of the incident laser flux was resonant with the Rb D1 absorption line.^[25] To compensate for the spectrally broader LDA output, the early CF hyperpolarizers employed high gas pressures (up to 8–10 atm) to pressure broaden the Rb absorption line.^[27] Unfortunately, the ¹²⁹Xe density has to be kept low to maintain high Rb polarization (the loss of Rb polarization from non-angular-momentum-conserving collisions is roughly three orders of magnitude greater for Xe than that for lighter gas such as He and N₂). Instead, by using a gas mixture dominated by ⁴He, the Rb D1 resonance can be broadened,^[28] allowing for a greater proportion of the incident laser light to be absorbed. In addition to the pressure broadening, the Rb atoms’ photon emission needs to be quenched to allow for a high level of polarization to be retained. The optical emission from excited Rb is comprised of unpolarized photons, which cause a loss of electron polarization when absorbed by other Rb atoms. N₂ is often selected as the quenching gas due to its large quenching cross section and ability to absorb energy from electronically excited Rb atoms into its ro-vibrational energy levels.^{[29, 30]} By incorporating the higher power LDA (50 W with 2 nm FWHM) with the lean gas mix into the hyperpolarizer, Driehuys et al. were able to dramatically increase the rate of HP gas production. When using a gas mix of 98% ⁴He, 1% ¹²⁹Xe, and 1% N₂ at 10 atm, around 1 standard liter (sL) of ¹²⁹Xe was produced with a P_Xe of 5%.^[20] This increase in production capacity enabled viable first-in-kind clinical studies.

3.2. Scaling up CF-SEOP, and Commercial Ventures

While a high production rate was established, the P_Xe level was limited by the available optical power of the LDAs at the Rb D1 wavelength. This limitation was addressed by Zook et al.^[31] who used a 210 W LDA to hyperpolarize the Rb. This LDA comprised seven individual fiber array packages (FAPs) that were independently tuned to 795 nm. This approach increased the photon flux and thus P_Rb–and hence P_Xe reached 65% using a 0.6% ¹²⁹Xe gas mix. While this demonstration achieved greatly increased P_Xe, the LDA still suffered from a rough Gaussian frequency distribution with FWHM≈1.6 nm. As a result, a large proportion of the incident photons are not absorbed by the Rb vapor, despite considerable pressure broadening of the Rb D1 line.

Most early CF hyperpolarizers focused on increasing P_Xe by increasing LDA power while working in the high-pressure regime (3–10 bar).^[20] Hersman and Ruset^[15] developed a CF hyperpolarizer that works in the low-pressure regime, where three-body van der Waals interactions dominate—giving rise to higher spin exchange rates.^[32] The higher spin exchange rate allows a higher flow rate and therefore improving the HP ¹²⁹Xe production rate. Moreover, by keeping ¹²⁹Xe concentrations low, high Rb polarization can be obtained with less laser absorption.^[15] Operating in this regime, enabled in part by new design features described below, provided record metrics of HP ¹²⁹Xe production: a maximum P_Xe of 64% was obtained with a flow rate of 0.3 sL of Xe/hr. Increasing the Xe flow rates to 1.2 sL/h and 6 sL/h gave P_Xe of 50% and 22% respectively.

Hersman’s design^[15] features an enormous (1.8-m-long) oil-heated SEOP cell optimized for the low-pressure/high-flow rate regime and employing counter-flow operation for the gas mixture “against” the high-power laser beam, 90 W and spectral bandwidth of 1.5 nm. Allowing the gas mixture to flow against the laser means Rb exiting the cell should receive maximum laser flux at the front of the cell to ensure the highest P_Rb and by extension the highest possible P_Xe, Figure 4a. The large SEOP cell equipped with long-lasting Rb pre-saturator (25 g Rb metal) requires a complex optical setup to ensure the whole cell is uniformly illuminated by the circularly polarized light. In 2008 this polarizer was used for human-scale HP ¹²⁹Xe production for lung imaging studies.^[33] Around 3–4 batches of HP ¹²⁹Xe could be produced daily by the hyperpolarizer in a hospital setting, ranging from 0.5 to 2 sL/batch with P_Xe of ~ 15%—demonstrating the consistency in production of large quantities of HP ¹²⁹Xe, which is important for the future of clinical MRI applications. This hyperpolarizer design was later improved through a series of additional technology advances including a significantly higher-power frequency narrowed LDA.^[34]

Figure 4.

a) Overall design of a continuous flow (CF) ¹²⁹Xe hyperpolarizer developed by Hersman and co-workers. b) The most recent embodiment of the ¹²⁹Xe hyperpolarizer by Xemed LLC, Durham, NH USA. Images courtesy of Dr. Iulian Ruset and Prof. William Hersman.

Xemed LLC (Durham, NH USA) now offers its fully automated XeBox ¹²⁹Xe hyperpolarizer (employing ~1 kW LDA), clinically ready, which does not require operator supervision other than replacing consumables, Figure 4b. The hyperpolarizer operates in a CF regime, energized by a kilowatt-scale laser with automated cryogenic separation/accumulation. This hyperpolarizer is capable of producing up to 6 sL of HP ¹²⁹Xe per hour, selectable by a touchscreen or remotely. In its current configuration, it can hyperpolarize up to four bags of various sizes per batch totaling up to 3 sL of HP ¹²⁹Xe gas. P_Xe reported by users routinely exceed 50% (up to 55%) in the bag (polarizations in the flow are typically 20–25% higher). A recent laser development, SpectraLock, incorporating a novel Atomic Line Filter in the external cavity, will allow higher power, cost reduction, and new polarizer operating regimes or designs.

Although able to produce excellent P_Xe values in volumes useful for a clinical setting, this hyperpolarizer has a complex design that would be difficult and expensive to build and emulate. As a result, Saam et al.^[35] developed a hyperpolarizer inspired by Hersman’s design, aiming to be simpler as well as to keep manufacturing costs below $100,000. This hyperpolarizer^[35] made use of a large (but comparatively shorter, ca. 1 m) SEOP cell with the counter-flow of the gas mixture against the laser light. However, the new hyperpolarizer utilized lower-cost options for heating, SEOP cell manufacturing, and the LDA system. Like the Hersman design, the Utah design introduced the Rb pre-saturator outside the main SEOP region. However, the Utah design used a forced-air heating system. Moreover, the Utah design utilized simple glass wool saturated with Rb in order to introduce the vapor into the gas stream. Finally, the Utah design used a much lower-power LDA with “only” 100 W maximum power. The average P_Xe obtained was 20% at a flow rate of 0.6 sL/h with 30 W laser photon flux. Although achieving lower HP ¹²⁹Xe production rates than those demonstrated by Hersman, these values demonstrate that it is possible to achieve high P_Xe at lower cost and with a simplified design.

The Utah hyperpolarizer was also used to validate a numerical model for CF SEOP (see also Section 5);^[35] both the P_Xe and the in-situ P_Rb were measured and compared to numerical predictions. The P_Rb was measured using optically-detected Rb electron spin resonance spectroscopy (ESR). The relative heights of the hyperfine peaks are directly indicative of the Rb polarization, Figure 5.

Figure 5.

A ⁸⁵Rb ESR spectrum used to determine P_Rb in the SEOP cell of the Utah hyperpolarizer. Relative heights of the hyperfine spectrum lines were used to calculate P_Rb. The spectrum represents Rb with a very low polarization. At higher P_Rb values, only the first 2–3 peaks are visible.

High P_Xe of the inhaled HP gas bolus is particularly important in a clinical setting. For example, image quality comparison using signal-to-noise ratio (SNR) metrics among HP ¹²⁹Xe ventilation images produced by Norquay et al.^[36] showed that using a smaller volume of ¹²⁹Xe—produced at a lower flow rate, but with higher ¹²⁹Xe polarization—yielded comparable images to those acquired with a larger administered dose with lower P_Xe (produced at a higher flow rate). Achieving comparable magnetization—and hence image quality—with a smaller administered dose can be particularly useful when there is a need to either dilute the Xe (to make it easier for patients with impaired lung function to inhale, owing to reduced gas density), or reduce the anesthetic effects of xenon (as Xe mixtures of over 50% can give rise to anesthetic effects^[37]). The hyperpolarizer used by Norquay et al. operates within the mid-pressure range, at 2 bars, compared to Hersman’s hyperpolarizer that operates at <1 bar. Operating within this mid-pressure range means the molecular lifetimes are shorter and the contribution of van der Waals interactions is reduced. As this contribution is inversely proportional to the gas density, operating at high Rb density will ensure that the spin exchange rate remains high when at this pressure. P_Xe of 12% was obtained when operating at a flow rate of 18 L/h at 373 K and a pressure of 2 bars, using a 25 W laser with FWHM of 0.1 nm. The laser FWHM is not significantly greater than the Rb D1 linewidth at this given cell pressure (0.05 nm) ensuring high photon absorption and a high optical pumping rate. A later adaptation of this hyperpolarizer, containing a larger volume SEOP cell and higher laser power, showed it is possible to generate higher P_Xe of 30% at a production rate of around 4 sL of Xe/h for HP gas imaging within a clinical setting.^[38] Experimental results achieved a dose equivalence (DE)^[39] rate of 1.013 L of pure and 100% polarized ¹²⁹Xe/h—three times higher than the highest previously reached value.^[15] Obtaining such high output rates suggests the possibility of performing high-resolution HP gas lung MRI with naturally abundant Xe (at high doses), which has a lower ¹²⁹Xe enrichment of 26.4% than the enriched (typically over 85%) ¹²⁹Xe usually used for these investigations.^[40] Due to the much lower cost of naturally abundant Xe this would further aid the development and clinical adaptation of this technology.

A key organization in the commercial development of Xe hyperpolarizers for clinical use is Polarean, Inc. (Durham, NC), a spinout of GE Healthcare which is actively pursuing regulatory approval of the technology. Their clinical Xe hyperpolarizer model 9820 (Figure 6), has completed Phase III clinical trials and is currently under FDA review as a drug–device combination product. Polarean devices are currently in use in more than a dozen research institutes located across North America and Europe.

Figure 6.

a) Schematic of the SEOP cell used within a Polarean 9820-A hyperpolarizer, featuring a gas flow path through a pre-saturated Rb chamber. b) Diagram of the overall device. c) Borosilicate cryotrap used with this polarizer. Reprinted with permission from Ref. ^[41]. Figure courtesy of Joseph Plummer.

The 9820 Xe polarizer is by design a CF hyperpolarizer. It runs at a moderate pressure of ~2 bar, and has a relatively large diameter cell, ~85 mm, which is fully illuminated by circularly polarized light from a ~170 W laser. The device delivers a polarization level exceeding 55% for 300 mL batches produced at a rate of 2 sL/h, equivalent to a DE production rate of ~1 sL/h (85% isotopically enriched ¹²⁹Xe). For the units operating in the field, research groups have reported polarization levels up to 40% under similar conditions.^{[41, 42]}

3.3. Other Key Developments in CF-SEOP

Meersmann and co-workers^[43] showed that the product of P_Xe and Xe concentration increases with an increasing Xe partial pressure and as a result, it is possible to operate even at near 100% Xe mole fraction. This approach produces HP Xe gas with in-cell SEOP Xe pressure of over 1 atm obviating the requirement for cryo-collection of HP Xe and allowing truly continuous delivery of dense Xe gas to a sample.^[44] In this method, removing the molecular nitrogen from the gas mixture initially reduces achievable signal intensities due to radiation trapping, however at high ¹²⁹Xe densities, P_Xe is less affected by radiation trapping because the dominating source of Rb depolarization is Xe itself, due to the increased significance of spin relaxation by Rb-¹²⁹Xe collisions at higher xenon densities.

Other methods of CF delivery have been investigated including operating without ⁴He as a buffer gas^[45] and using very low ¹²⁹Xe concentrations.^[46] Fujiwara et al.^{[47, 48]} showed that possible alternatives to N₂ can successfully be used as quenching gases this way, including isobutene, furan, and butane, and Meersmann et al.^{[49, 50]} demonstrated a method of removing N₂ without the need for cryo-collection. However, due to the complexities (and potential toxicities) that such alternative methods may bring to the SEOP process, they may not be suitable for implementation in a clinical setting.

Continuous, real-time delivery of HP ¹²⁹Xe for gas imaging has been successfully demonstrated using ‘traditional’ SEOP methods. Driehuys and co-workers^[51] showed real-time delivery of HP ¹²⁹Xe to rodents, using dilute concentrations (1%) of ¹²⁹Xe produced on demand. Instead of performing cryo-accumulation, the xenon pressure was reduced to physiological levels and the gas mixture, with P_Xe=25%, could be directly delivered to the subject to enable continuous imaging within a clinical setting.

4. Stopped Flow SEOP Production Method

4.1. SF SEOP Fundamentals

The Stopped Flow (SF) modality, alternatively called “Batch-Mode”, is a technique of producing a defined quantity of HP ¹²⁹Xe with a desired gas mixture. In this process, the gas mixture of Xe and (often) N₂ gas is loaded into the SEOP cell with Rb metal and sealed; the SEOP cell is then heated and illuminated with circularly-polarized laser light. Once Xe gas in hyperpolarized, the SEOP cell is opened and the gas is expanded from the SEOP cell into a storage container allowing its subsequent use. The SEOP cell is then reloaded.

Particularly in the 1980s and 1990s, a large number of groups around the world developed small-scale single-batch setups for fundamental studies of SEOP, as well as applications in chemistry/biochemistry, materials science, and physics (for review, see Refs. ^{[19, 52, 53]}). The first ¹²⁹Xe HP lung MRI was performed by Albert et al. in 1994 using a small-scale SEOP device, which was employed for the imaging of a mouse lung.^[7] In their single-batch SEOP polarization experiment, 1–2 Watts of circularly polarized light produced by a Ti:sapphire laser was applied to polarize the Rb atoms’ electron spins. Following 5–20 minutes of optical pumping, P_Xe ~25% was achieved, and the contrast agent was administered into the mouse’s lung through the breaking of the SEOP cell to expel the HP gas. This pioneering report allowed for a high concentration HP ¹²⁹Xe mix to be delivered into the lungs, permitting ultrafast (0.3 s) ¹²⁹Xe images to be taken, and demonstrating the viability of the imaging technique.

Following the demonstration of very high P_Xe in small quantities by Jänsch and co-workers,^[55] the batch/SF design was improved in 1999 by Rosen et al.^[54] through the creation of a SEOP polarization and collection system to increase the capacity of HP ¹²⁹Xe production and to enable SEOP cell re-loading and re-use without physical glassware destruction—making the device a true stopped-flow system. The so-called Rosen SEOP cell (Figures 7 and 8) at the center of the device comprised an inner cylinder (75 mL, 1-inch diameter) that was contained within a larger (2-inch diameter) glass cylinder that acted as a forced-air oven, enabling the inner cell to be heated to a desired temperature. The cell had two stopcocks allowing the inner cell’s volume to be reproducibly loaded and unloaded as desired. The hyperpolarizer also included a cryo-collection apparatus, allowing the HP ¹²⁹Xe produced over multiple batches to be accumulated over time and enabling a greater overall volume to be collected for in-vivo studies. The SEOP system also employed a 30 W LDA system (comprised of two 15 W FAPs with FWHM of 2–3 nm) for irradiation of the SEOP cell resulting in the increased HP ¹²⁹Xe gas production rate and quantity in a compact and cost-effective manner.^[26]

Figure 7.

Rosen stopped-flow polarizer design. Once Xe is polarized through the SEOP cell, Xe is expanded into the remainder of the system for accumulation, storage and delivery. Large quantities of HP Xe are accumulated and then frozen in the glass cryo-vessel which is held in a 500 G magnetic field provided by a permanent magnet (not shown). Once enough Xe has been accumulated in the cryo-vessel it is thawed and flows into the storage cylinder where it has a gas polarization lifetime of 18 minutes.^[54]

Figure 8.

A Rosen SEOP cell, during illumination by a high-power IR laser (coming in from the left). The bright violet light emanating from the inner cylinder (containing the gas and Rb vapor) is caused by the lack of N₂ gas, intentionally absent to allow study of radiation trapping processes. Re-emission of resonant photons gives rise to energy pooling, resulting in the population of higher-lying Rb excited states.^{[30, 57]}

The cryo-storage vessel was placed within a 500 G magnetic field while kept at a temperature of 77 K (by immersion in a liquid N₂ reservoir), resulting in a ¹²⁹Xe T₁ on the order of 1 hour. Thus, the ¹²⁹Xe is frozen in storage while any N₂ within the gas mixture can be pumped out, leaving the concentrated HP ¹²⁹Xe gas mix. Once the hyperpolarization process is completed, the SEOP cell is opened to expand the HP ¹²⁹Xe gas into the transportation manifold for later use. Following the HP ¹²⁹Xe expansion, the HP ¹²⁹Xe gas has a T₁ of 18 minutes.^{[24, 56]} The Rosen device marked a useful demonstration of the ability to perform SEOP using an LDA configuration, and again increase the capacity of HP ¹²⁹Xe production: This hyperpolarizer performed SEOP of a gas mixture containing 1700 Torr of ¹²⁹Xe and 150 Torr of N₂ producing 0.16 sL of HP ¹²⁹Xe gas with P_Xe of ~7.5% every 5 minutes.^[54]

4.2. Enabling Advances in Laser Technology

Regardless of overall design, the performance of ¹²⁹Xe hyperpolarizers was clearly limited by the low quality of the light provided by the high-power LDA sources. Luckily, new methods to spectrally narrow LDA outputs were developed to mitigate the primary weakness of these light sources for ¹²⁹Xe SEOP. To better appreciate the technical challenge that had to be overcome, it is instructive to consider the work of Levron et al.,^[58] who investigated cesium (Cs) optical pumping using a single-mode LD with a narrow spectral width of only ~0.12 nm FWHM. While this LD had a suitably narrow emission bandwidth (i.e., approaching the D1 linewidth) for optical pumping, it had very limited optical power of just ~0.3 W. To increase the power, more LD elements would need to be added, but such a LDA would necessarily have a broader spectral output because the output of each element has its own independent spectral profile—producing a wide distribution of frequency outputs when all the elements were used together. To reduce the frequency range and narrow the output, Walker et al.^[59] employed a dispersive ruled grating (DRG) within an external cavity diode laser (ECDL) in 2000 to improve the optical pumping efficiency. The LDA used an external cavity with a diffraction grating (1800 lines/mm) to narrow the laser emissions. The diffraction grating reflects back a narrow spectral range to the LDA. This process in turn seeds the LDA’s gain medium (of each element) to “lock onto” the reflected spectral wavelength resulting in a narrowed spectral width.^[60] As a result, the bandwidth of the overall LDA emission (tuned to Rb D₁ frequency) was narrowed from ~2 to 0.1 nm. While the DRG also decreased the power output from 4 W to 2.5 W, this reduction in total photon flux was significantly outweighed by the large increase in the fraction of light that was resonant with the Rb vapor—as demonstrated by the fact that the narrowed 2.5 W laser resulted in 1.40-fold greater P_Xe compared to a 15 W unnarrowed LDA.^[59]

While the use of the DRG increased the proportion of laser photons that were tuned to the Rb D₁ absorption frequency, the efficiency (manifested by the decreased power output) remained a limitation. Moreover, the effectiveness of the DRG relied sensitively on the optical alignment of the system to ensure optimal efficiency and spectral quality, creating a complex arrangement that is difficult to align and maintain. However, instead of using a dispersive 2D diffraction grating as the feedback element, the narrowing can also be performed using volume holographic gratings (VHG, also referred to as a VBG – Volume Bragg Grating) as proposed by Barlow et al. for use in SEOP.^[61] By using VHGs, a simpler cavity feedback arrangement for spectral narrowing of the laser emission was achieved, with efficiencies up to 90%^[62] being realized (compared to ~40–66% efficiency with external cavity narrowing^[61]), thereby providing more resonant light for SEOP.

VHGs are composed of a frequency-selective grating that can be formed in a slab of photosensitive glass. As in the ECDL design, the VHG feedback wavelength is defined by the repeat spacing of the Bragg grating to lock the diode emission and restrict the spectral width^[63]; the spacing, in turn, can vary slightly with the temperature of the optic—providing a narrow tuning range (see below). This approach results in the LDA being made to lase at the required wavelength, narrowing the emissions and increasing the proportion of photons tuned to the Rb D₁ frequency.

The use of the VHGs for ¹²⁹Xe hyperpolarization was first performed by Nikolaou et al.^[61] in 2009 using a Rosen SEOP cell setup, with binary (Xe/N₂) gas mixtures totaling 2000 Torr. Using the setup, the power output from the LDA was varied to tune the VHG and maximize polarization. When first used for ¹²⁹Xe SEOP, it was found that the narrowed LDA spectral output led to a large increase in absorption by Rb atoms at the front of the cell. Thus, reduced photon transmission to the back of the SEOP cell occurs due to the high optical density of the Rb vapor, resulting in “dark” regions where Rb polarization tends to zero. To correct this limitation, the laser emission was offset ~0.1 nm from the center of the D1 resonance to increase the transmission through the optically thick Rb vapor and ensure good illumination throughout the entire length of the SEOP cell. However, such attempts to tune “fixed” VHGs can present an additional problem: a dependence of the laser’s wavelength on power. This issue arises from the fact that as the power is increased, the VHG heats up, which causes the spacing of the VHG’s grating to increase—thereby red-shifting the wavelength at which the VHG “locks” the LDA’s output. To address this issue, both the laser’s power and external temperature had to be adjusted in concert to tune the laser with respect to the D1 line, albeit at the expense of reduced LDA power (here 27 W, generating a FWHM of 0.27 nm). By performing SEOP in this configuration, a three-fold increase in P_Xe was achieved compared to a conventional LDA of the same power.

For most applications, it is not P_Xe, but the magnetization “payload” (proportional to the product of P_Xe and the number of spins), that is the most important parameter for determining the clinical quality of the resulting images and spectra. However, while it can often be relatively easy to achieve high P_Xe with low in-cell Xe densities [¹²⁹Xe], P_Xe can often drop dramatically as Xe density is increased because of the inverse relation between P_Xe and [¹²⁹Xe]^[43] due to the non-spin conserving collisions that arise between the ¹²⁹Xe and the Rb atoms^{[27, 64]} (in addition to reduced contributions from more-efficient three-body collisions to the Rb/Xe spin-exchange rate^[32], as discussed above). Because of this problem, increasing [¹²⁹Xe] often resulted in lower (or at best, similar) overall magnetization payload. However, it was found that the decrease in polarization as a result of increasing [¹²⁹Xe] can mitigated by changing the temperature at which SEOP is performed, particularly when applying high resonant photon flux (e.g., with a high-wattage VHG-narrowed LDA): Under these conditions, an inverse relationship was found between Xe density and the optimal SEOP temperature (i.e., T_opt, the temperature at which the highest P_Xe was achieved)—a result that was well-reproduced by simulations (see below).^[65] Under such conditions where the in-cell ¹²⁹Xe T₁ was very long (i.e., so that 1/T₁ was much smaller than γ_SE), higher Xe densities require better cell illumination and a greater “photon-to-Rb ratio” (because of the increased Rb spin-destruction rate). Thus, lowering the cell temperature can help ensure higher P_Rb throughout the cell—and hence, higher overall P_Xe at high Xe densities—albeit built up over a longer time.^[61] Indeed, temperature optimization resulted in a levelling off of polarization with increasing density.^[66] Although this work was performed using single batches in smaller volumes (Rosen cells, 75 cc), it showed that that it was possible to use SF SEOP and high in-cell densities to create HP Xe with much higher magnetization “payload” than previously anticipated.

4.3. Scaling SF-SEOP to Clinical Production

As a result, the hyperpolarizer developed by Nikolaou et al.^{[67, 68]}in 2012, dubbed “XeNA”, aimed to increase P_Xe at higher densities while maintaining a high production rate. Moreover, the hyperpolarizer strived to increase accessibility of the HP ¹²⁹Xe production method, via open-source use of “off-the-shelf” components while introducing an automated method of production to improve ease of use. The focus of the hyperpolarizer was to maximize the magnetization at clinically useful volumes of the delivered HP product (~0.5–1 L at 1 atm in a Tedlar bag) at increased ¹²⁹Xe concentration (up to 75% content). This device utilized a ~180 W VHG-narrowed LDA (FWHM ~ 0.27 nm^[69]) to compensate for the increased Rb spin destruction rate that arises due to higher [¹²⁹Xe]. A design advantage of using higher Xe densities is that the cryo-collection step may be entirely obviated—simplifying the HP ¹²⁹Xe production process and making the device easier to automate. With the XeNA hyperpolarizer, the 0.5-L SEOP cell operated at a total pressure of 2000 Torr^[69] resulting in the production rate of approximately 1 sL of HP gas mixture per hour, with a ~0.8 sL batches able to made each run. To improve ease of use, a graphical user interface (GUI) controlled the majority of the system components allowing for easier adoption into clinical settings. With this system, a P_Xe of up to 90% was achieved with an in-cell ¹²⁹Xe partial pressure of 300 Torr. However, at higher ¹²⁹Xe partial pressure of 1570 Torr, P_Xe was reduced to 30% but generated a greater magnetization, demonstrating the benefits of the higher density polarization. Korchak et al. created a portable polarizer with a design that has many similar features, except that it is also capable of running in both CF and SF modes. This device produced P_Xe up to ~40 % in small SF batches, or values of P_Xe=25% and 13% in CF mode with flow rates (and Xe partial pressures) of 6.5×10⁻³ sL/min (0.1 bar) and 26×10⁻³ sL/min (0.4 bar).^[70]

An improvement to the XeNA system was introduced by Nikolaou et al. through the introduction of the second-generation (GEN-2) “XeUS” system.^[71] XeUS offered some notable advantages over the previous hyperpolarizer designs that improved the automation and performance of the system. One of the most important changes was the introduction of the highly-integrated 3D printed oven featuring a thermo-electric cooling (TEC) module for temperature control of the SEOP cell (Figure 9).^[71]

Figure 9.

a) Annotated photograph of the second-generation (GEN-2) XeUS hyperpolarizer device and chassis, outlining the key components (e.g., power supply units, PSUs) and their orientation; b) 3D rendered schematic showing the GEN-2 hyperpolarizer SEOP cell and 3D-printed forced air oven interface; c) corresponding photograph. Display (a) reproduced with permission from Figure 1 from Ref. ^[72]; Displays (b,c) reproduced with permission from Figure 3 from Ref. ^[73].

The redesigned SEOP cell uses a premixed Xe-containing gas mixture, resulting in a substantially simplified gas-handling manifold design^{[73, 74]}; operating at a constant positive pressure mitigates the potential for Rb in-cell oxidation otherwise caused by slow leaks between gas-reloading stages for the SEOP cell. Furthermore, a 2”-diameter laser beam expander has been integrated in the LDA design, which when coupled with the mechanical aligning legs of the oven assembly, allowed for convenient alignment of the laser beam with the SEOP cell. The originally introduced LDA with FWHM of 0.20–0.30 nm was later upgraded to the LDA with FWHM of 0.154 nm.^[73] Through these modifications, a maximum P_Xe of 83.9%±2.7% was achieved with a ¹²⁹Xe density of 1000 Torr, while the highest DE values was achieved at a higher ¹²⁹Xe density of 1330 Torr at P_Xe of 72.6±1.4%, Figure 10.^[75–77] This polarizer was also used as a testing platform for temperature-ramped SEOP, where faster initial build-up rates (γ_SEOP) could be achieved by starting with oven temperatures that would otherwise be too high to yield stable SEOP for the whole duration of the Xe polarization run; after an initial period (but before the onset of “Rb runaway”^{[31, 75, 78]}), the oven temperature would then be ramped down to achieve more stable SEOP and yield a high final polarization near what would be achieved without the ramping process—but in half the time.

Figure 10.

a) and b) Temperature maps of steady-state P_Xe, γ_SEOP, and P_Rb values following SEOP at different temperatures in a single cell using a generation-2 “XeUS” hyperpolarizer with forced air cooling/heating. Note two different color-coded y-axes for P_Xe, P_Rb and γ_SEOP. c) Results comparing P_Xe, T_b, P_Rb and T₁ from a single SEOP cell across a quality assurance study. Displays (a,b) are reproduced with permission from Ref. ^[76] (Figures 5a,c); Display (c) is reproduced with permission from Figure 5c from Ref. ^[72].

The presence of two nearly identical GEN-2 “XeUS” hyperpolarizers on two continents allowed the first-in-kind pilot quality assurance (QA) study of clinical-scale hyperpolarizers,^[72] Figure 10c. In this study, the repeatability of SEOP in a 1000 Torr Xe/900 Torr N₂/100 Torr ⁴He (2000 Torr total pressure) gas mixture was investigated over several hundred gas loading cycles, with very little decrease in performance during the first ~200 cycles: P_Xe = 71.7±1.5%, γ_SEOP = 0.019±0.003 min⁻¹, ¹²⁹Xe T₁ = 90.5±10.3 mins. Although the SEOP cell in this study exhibited a detectable performance decrease after 400 cycles, the cell continued to produce potentially useable P_Xe = 42.3±0.6% even after nearly 700 refill cycles.^[72] The ability to reproducibly achieve and maintain performance over so many refill cycles without having to replace the SEOP cell bodes well for the ultimate translation of the device to biomedical and clinical imaging applications.

A third-generation (GEN-3) SF-SEOP hyperpolarizer was introduced in 2020 by Birchall et al.^[74] One of several improvements embodied by the design was replacement of the forced air oven with an aluminum jacket in direct thermal contact with the SEOP cell, offering rapid heat transfer during the temperature cycles, Figure 11. This new design resulted in a sevenfold decrease in the heating and cooling times to ~4 minutes while also ensuring that the process is more thermally stable with the γ_SEOP build-up rate of up to 0.2 min⁻¹ (Figure 12^[74]); as a consequence, the design also allows far more nimble temperature ramping during SEOP. The GEN-3 device also offers improved in situ NMR polarimetry: Figure 12a shows a reference ¹H NMR spectrum from a thermally-polarized water phantom (used for absolute calibration of ¹²⁹Xe polarization) acquired with 1024 scans in only 5 minutes. Moreover, the GEN-3 device has a small footprint, 1.2 m (width) x 1.5 m (height) x 0.6 m (depth)—enabled in part by use of a compact solenoid magnet—and provides a true push-button operation with advanced automation and integration over a WiFi interface.

Figure 11.

(a) Schematic drawing of the third-generation (GEN-3) SF hyperpolarizer. The SEOP cell (blue cylinder) is encased within an aluminum heating jacket (red cylinder) to achieve the Rb vapor density necessary for efficient SEOP. The cell and heating jacket are contained within a magnetic solenoid coil (orange cylinder), which provides the homogeneous magnetic B₀ field created by a solenoid. (b) 3D rendering of the GEN-3 hyperpolarizer SEOP cell and aluminum heating jacket design. (c) Thermal photo taken during hyperpolarizer operation. (d) 3D rendering of the upper chassis with solenoid magnet coil shown as a cut-away to depict SEOP cell and other internal components. The images are adopted with permission from Ref.^[74] (copyright 2020, American Chemical Society).

Figure 12.

(a) ¹H (blue) and ¹²⁹Xe (red) NMR measurements of a thermally polarized water (doped with 10 mM CuSO₄, 1024 scans over 5 minutes, and a HP Xe gas mixture, 1 scan, respectively to determine ¹²⁹Xe nuclear spin polarization (P_Xe); (b) IR spectroscopic measurements of the SEOP pump laser to estimate Rb electron polarization (P_Rb); (c) Typical SEOP ¹²⁹Xe polarization build-up (red, P_Xe(max) = 34.3±0.9%) and decay (blue) to determine polarization build-up rate (γ_SEOP) and ¹²⁹Xe relaxation decay constant (T₁); (d) Temperature-dependent SEOP map of steady-state P_Xe, γ_SEOP and P_Rb, P_Xe(max) = 37.9±6.7% at a jacket temperature of 60 C. All data acquired from various SEOP cells and Xe-containing gas mixtures using the GEN-3 hyperpolarizer.^[74]

5. Simulations of SEOP Processes

5.1. CF-SEOP Simulations

Broadly speaking, the theoretical mechanisms important in CF production of HP ¹²⁹Xe are well-understood, and this insight allows quantitative modeling of CF SEOP processes. Such computational models generally belong to one of two categories: finite difference models (FDMs)^{[79, 80]} or finite element models (FEMs).^[81–86] Although these descriptions refer to the computational methods that are used, the important distinction between these types is the complexity of the physical model employed: For modeling SEOP, FDMs generally use one spatial dimension to approximate laser-light propagation through the SEOP cell, and one- or two-dimensional (2D) uniform or laminar-flow approximations. Nuances of SEOP cell geometry are often ignored, along with effects such as laser heating, alkali density inhomogeneities, and convection. In contrast, FEMs typically use 2D- or three-dimensional (3D) geometries; moreover, as both commercial and open-source “multiphysics” packages using FEMs are relatively common, thermal, diffusion, and convective-flow effects can easily be incorporated. Additionally (and in contrast to most FDMs), FEMs typically include the frequency distribution of the laser light, often approximated as Gaussian.

For CF systems that use large SEOP cells (such as the Hersman design^{[15, 35]}) simple 2D laminar-flow FDMs predict the observed P_Xe relatively well.^[79] In addition to good agreement with P_Xe, Schrank et al. also observed relatively good agreement between such models and the in situ Rb polarization.^[35] Further, Fink et al.^[82] used an FEM to simulate a geometry similar to that used by Ruset et al.,^[15] and found good agreement between their model’s predictions and the available experimental results at the time. Fink et al. made several recommendations for CF polarizer construction based on the results of their model. They determined that it was best practice to pre-saturate the Rb vapor in the gas stream and to pre-heat the gas stream before flowing it to the SEOP region. They also recommended flowing the gas anti-parallel to the laser beam. Finally, they recommended orienting the SEOP cell vertically—supporting many of the design features present in the Hersman design.

However, for CF systems that use smaller SEOP cells, there is an apparent discrepancy between the modelled and observed P_Xe values—especially when using high-power pumping lasers. For example, Freeman et al. characterized this disagreement by examining P_Xe in 100-cc, 200-cc, and 300-cc SEOP cells and compared those observations with predictions of a 2D FDM model using uniform flow.^[87] The observed P_Xe differed from the model by as much as a factor of 4. A few approaches have been attempted to resolve this large disagreement. For example, Plummer and co-workers simulated^[41] the entire CF-SEOP process (including ¹²⁹Xe cryo-collection) and found that when Rb density and solid-state ¹²⁹Xe T₁ were allowed to be freely floating parameters fit to experimental data, both values tended to be significantly less than literature expectations---potentially explaining much of the perceived underperformance of certain CF-SEOP designs. Earlier, Freeman et al. suggested that the discrepancy was a result of a novel form of contamination in the SEOP cell manifested by Rb cluster formation. This hypothesis originated from the fact that small SEOP cells coupled with high-powered pump lasers give rise to very high power densities. Researchers who study alkali molecular clusters frequently use heated tubes containing bulk alkali metal with coupled plasma to form the clusters; thus, the apparatus used to produce the molecular clusters share a number of similarities with small SEOP setups. Freeman et al. added terms associated with theoretical clusters to their simulation, affecting it in three ways: (1) a decreased optical pumping rate, owing to reduced photon flux caused by light scattering off of clusters; (2) an increased Rb spin-destruction rate, under the expectation that cluster-Rb collisions would randomize the Rb electron spin-polarization; and (3) an increased ¹²⁹Xe spin-relaxation rate, for similar reasons. By including these additional terms for the clusters, Freeman et al. were able to fit an FDM with the available data to find estimates for the scattering cross-section, spin-destruction cross-section, and spin-relaxation cross-section due to the clusters of 1×10⁻¹² cm², 4×10⁻⁷ cm³/s, and 4×10⁻¹³ cm³/s, respectively. The estimates assumed a metal cluster density that was 1/1000 of the Rb atomic vapor density. Using those fit parameters and assumptions, the FDM had much better agreement with the observed experimental data.

Flower et al. later disassembled used SEOP cells from a CF apparatus and examined them using scanning electron microscopy and transmission electron microscopy.^[88] This study found hemi-spherical particles on the interior surface of the optical pumping cells ranging in diameter from 200 nm to 10 μm. Additionally, the study found pits on the interior glass surface of the cells. Researchers ascribed existence of the particles and the glass pitting to be related to Rb cluster formation. Nevertheless, to date Rb metal clusters have not been detected in situ in active SEOP cells.^[80]

FEM models also have had success reproducing the behavior of CF SEOP hyperpolarizers. Fink et al.^[82] examined small-cell systems and reported relatively good agreement with data from Shah et al.^[89], who used a SEOP cell with a cross-section of 13 cm² and a modest laser power of 50 W. More recent FEM models of CF SEOP created by Burant and Schrank^[83–85] are currently under development with the goal of better capturing all phenomena of CF SEOP hyperpolarizers—including complexities of mass- and thermal transport within different regions of CF-SEOP cells.^[65]

5.2. SF-SEOP Simulations

As with CF-SEOP, simulations of SF-SEOP can also be useful to both gain insight into underlying phenomena and to help devise practical guidance for hyperpolarizer design and operation. SF-SEOP modeling requires several simplifying modifications, such as the omission of the gas flow terms and Rb clusters (which, if relevant, are thought to occur only in certain CF regimes).^[90] Two such simulations of SF-SEOP were implemented by Skinner et al.^[65] and validated against two experimental datasets^{[75, 91]} from the high-Xe-density, high resonant-photon-flux regime. Both simulations provided excellent qualitative and quantitative reproduction of the Xe-rich SF SEOP datasets, including the inverse relationship between the optimal cell temperature (T_opt) and [Xe]. Furthermore, under certain conditions the simulations trended towards a ‘universally optimal’ laser absorption when P_Xe is maximized—which may help provide i) an explanation for the experimentally observed interplay of laser linewidth, T_opt, and [Xe], in the context of optimizing the photon-to-[Rb] ratio (mentioned above), and ii) a new way to optimize SF hyperpolarizer operation. The simulations also studied the design and operation of such hyperpolarizers, considering cell temperature, cell geometry, laser linewidth, laser power, gas mixture, and gas pressure, to help further improve performance—including the prediction that reduction of laser linewidth improves P_Xe disproportionately at higher [Xe], and the potential utility of ternary gas mixtures containing ⁴He (for improved temperature stability owing to higher thermal conductivity)—recently realized in a 2^nd-generation clinical-scale hyperpolarizer.^[65]

Finally, very recently Branca and co-workers^{[92, 93]} published simulation work that aimed to improve over previous efforts by going against assumptions that may lead to underestimation of Rb/Xe spin-exchange cross-sections, but significant overestimatation of Rb vapor densities (particularly in CF polarizers) and Xe cell residence times (exclusively in CF polarizers). The resulting simulations showed good quantitative reproduction of experimental polarization values and dynamics in both CF-SEOP and SF-SEOP, without having to consider contributions to spin exchange or relaxation beyond those considered in established theory.^{[19, 29][32]} Future experimental observations (including careful measurements of spin exchange contributions under relevant conditions, position-dependent Rb polarization and density, light and heat transport, presence or absence of Rb clusters, etc.) may help shed light on differences among the various approaches.

6. Other emerging HP ¹²⁹Xe production technologies

6.1. Alternative SEOP-based Xe Hyperpolarization Methods

While CF and SF SEOP based designs have dominated the production of HP ¹²⁹Xe over the years, a number of alternative approaches have been also developed. One such approach stems from an otherwise “traditional” CF SEOP design where the flow of the produced HP ¹²⁹Xe gas mixture is re-directed back into the SEOP cell.^{[94, 95]} Raftery and co-workers investigated the use of such a recirculating SEOP setup for long-term acquisition of enhanced NMR spectra of materials’ surfaces. The ¹²⁹Xe is circulated through the hyperpolarizer multiple times, allowing higher P_Xe to be achieved compared to a single pass through system. This system proved to be surprisingly effective, producing P_Xe as high as 69% after only 5 minutes of recirculation with two 40 W coupled lasers.^[96]

Saam and co-workers introduced a different SEOP hyperpolarizer design with the goal of creating large quantities of liquid HP ¹²⁹Xe^[97] from a relatively small device. In their method, ¹²⁹Xe is still hyperpolarized in the gas phase via SEOP; however, once polarized, the gas undergoes phase exchange (facilitated by convection) with a column of liquid xenon, allowing polarization to accumulate in the liquid phase. Using just an 8-mL spherical SEOP cell and 15 W LDA, the apparatus provided 8% P_Xe in 0.1 mL of liquid Xe in 15 min.

To dramatically decrease the size of the hyperpolarization equipment, a microdevice can be employed to perform ¹²⁹Xe SEOP. In the device by Jimenez-Martınez et al.^[98], a gas mixture of ¹²⁹Xe and N₂ flows from an external gas manifold into the chip’s pump chamber. Rb is deposited in the micro-device chamber (via a chemical reaction of barium azide and RbCl) and is illuminated by circularly polarized light in the “pump” chamber, where the ¹²⁹Xe entering the chamber is hyperpolarized by SEOP. The ¹²⁹Xe polarization is then measured optically in the adjacent (downstream) “probe” chamber via an independently controlled laser beam using the response from the polarized ⁸⁷Rb to perform optical magnetometry.^{[99, 100]} The pump and probe chambers, embedded within a 1-mm thick chip measuring 1×3 cm², are only 5×5 mm² and 3×3 mm², respectively. In principle, the HP ¹²⁹Xe gas could then be allowed to flow out of the chip for use in investigations. This first-generation on-chip SEOP device demonstrated the viability of the overall approach; however, ¹²⁹Xe polarization was limited (~0.5%). Collaboration with the Pines Lab and optimization of the chip design led to a significantly improved device (Figure 13). Kennedy et al. reported an on-chip SEOP device with lateral dimensions of 4×4 mm² for pump and probe chambers; the device could employ a gas flow rate of ~5 μL·s⁻¹ while also achieving P_Xe levels of 7% with a ¹²⁹Xe polarization lifetime of ~6 s.^[101] While the HP Xe production capacity of such on-chip devices may be limited, the small device size (39×19 mm²) allows direct integration into other microfluidic devices.^[102]

Figure 13.

The optimized “on-chip” ¹²⁹Xe hyperpolarizer.^[101] a) A mixture of Xe and N₂ gas flows from an external gas manifold into the chip, where laser light (795 nm) optically pumps the electron spins of Rb vapor atoms; ¹²⁹Xe that flows through this pump chamber is hyperpolarized via SEOP. Additional 795 nm laser beams are used to perform optical magnetometry in both pump and probe chambers. b) photo of the device. Courtesy of V. Bajaj and A. Pines, reprinted with permission.

6.2. Non-SEOP-based Approaches for Hyperpolarizing ¹²⁹Xe

Non-SEOP hyperpolarization techniques for production of HP ¹²⁹Xe have also been developed. So-called “Brute force” polarization is one such technique; it instead relies on the dependence that the equilibrium P_Xe has on temperature and applied magnetic field (Eq. 1). To induce large P_Xe, the temperature is reduced to milli-kelvin values, while subjecting the sample to a large magnetic field.^[103–106] However, when such low temperatures are applied to the sample, the ¹²⁹Xe T₁ increases dramatically, meaning that the sample needs to be kept in the brute force conditions for a long period of time—an issue which can be mitigated with the addition of paramagnetic substances. For example, molecular oxygen can be mixed into the sample to further reduce the ¹²⁹Xe T₁, and thus reduce the time that brute force polarization needs to be applied from 65 to 9.5 hours.^[107] Although P_Xe levels of up to 40% have been achieved,^[108] the extraction of the HP ¹²⁹Xe gas can also be challenging without losing a substantial fraction of polarization during the phase-transition process.

Finally, in Dynamic Nuclear Polarization (DNP),^{[109, 110]} a sample is mixed with a free radical and then cooled to ≤1 K while subjected to a high magnetic field (between 3–7 T).^{[111, 112]} Under these conditions, the electron spins gain near-unity polarization,^[113] and microwaves tuned to the electron spin resonance are applied to drive electron polarization to nearby nuclei in the lattice. When the hyperpolarization process is completed, the sample is rapidly heated with a solvent, allowing the dissolved HP substance to be removed and used for HP NMR or MRI; this variant is known as dissolution DNP (d-DNP).^[113] Although d-DNP is most commonly applied for hyperpolarization of liquid ¹³C-labelled molecules, there has also been success in producing HP ¹²⁹Xe via DNP^[114][115] including HP ¹²⁹Xe gas via sublimation (dubbed s-DNP),^[114] with clinically-useful P_Xe levels as high as 30% achieved,^[114] Figure 14. Currently, s-DNP is limited by a number of factors. First, the effectiveness of the s-DNP process relies upon the homogeneity of the sample, which can be hard to achieve for solid-state ¹²⁹Xe mixtures. Second, P_Xe can be limited by a “spin bottleneck”, in which the electrons cannot efficiently transfer polarization to ¹²⁹Xe (the permanent magnetic moment of the paramagnetic dopants can shift the frequencies of ¹²⁹Xe spins near the dopants from those farther away, slowing the spread of polarization into the bulk lattice). Third, to date s-DNP provides limited quantities of HP ¹²⁹Xe (ca. 0.015 sL), which is not yet sufficient for clinical applications. Current s-DNP efforts aim to overcome these limitations.

Figure 14.

a) Schematic of thermal equilibrium of Xe nuclei in the solid state; b) application of high-power microwaves hyperpolarizes electron spins at cryogenic temperature to near unity polarization followed by polarization transfer to nearby Xe nuclear spins; c) Bulk HP ¹²⁹Xe states is created over time; d) sublimation DNP hyperpolarization of ¹²⁹Xe: ¹²⁹Xe polarization build-up curve measured at 5 T and 1.15±0.05 K in a 1.5 M xenon sample dissolved in d₆-ethanol containing 50 mM TEMPO. Inset: DNP-enhanced (red) and thermal equilibrium (black) ¹²⁹Xe NMR spectra measured to determine P_Xe. Display d) Reprinted from Ref. ^[114] (copyright ACS, 2015)–courtesy of Dr. Arnaud Comment.

7. Biomedical Outlook

The inherent properties of Xe such as diffusivity, blood solubility, and frequency shifts in different environments,^[116] can be collectively exploited to directly probe Xe transport and distribution through various biological tissues and organs. In the case of lungs, in addition to the straightforward imaging of HP ¹²⁹Xe gas in pulmonary airways, referred to as “ventilation imaging”, it provides the unique capability of essentially imaging the gas exchange pathway followed by inhaled O₂.^[117] More specifically, once ¹²⁹Xe is inhaled and enters into the alveolar space, it diffuses across the alveolar-capillary barrier membrane, where it exhibits a 197.8 ppm shift in its resonant frequency^[118] relative to ¹²⁹Xe atoms remaining in the airspaces, corresponding to a shift of ~3.5 kHz at 1.5 T. Xe then passes into the local capillary network and enters into the red blood cells (RBCs)^[119] where it exhibits an even larger frequency shift of 217.6 ppm.^[120] These two compartments, barrier and RBC, are collectively referred to as the “dissolved-phase”; see Figure 15.

Figure 15.

¹²⁹Xe MRS showing signal intensity in human lungs at 1.5 T. A) NMR spectrum obtained when exciting the gas- and dissolved-phase ¹²⁹Xe equally, showing that the dissolved phase ¹²⁹Xe signal is only 1–2% as large as gas; B) When the dissolved phase is selectively excited the barrier and RBC spectral peaks are better appreciated. Reprinted with permission from Ref. ^[121].

The inherent challenge of imaging dissolved phase ¹²⁹Xe, with more than 50-fold smaller signal than the gas phase (because the absorbed faction is substantially lower than that of gas phase), is overcome by continuous replenishment of HP ¹²⁹Xe diffusing in from the airspaces—allowing one to use a much larger flip angle excitation than that typically used for ventilation imaging.^[121] This property permits imaging of dissolved and gas-phase ¹²⁹Xe during the same breath-hold,^[122] which is critical for quantitative analysis of gas transfer. ¹²⁹Xe images are typically processed into quantitative binning maps of all three compartments using thresholds based on the mean and standard deviations of distributions derived from the healthy reference cohort, subsequently analyzed to derive quantitative measures of ventilation, barrier uptake, and RBC transfer. In the following sections, we briefly review HP ¹²⁹Xe biomedical applications in various organs and diseases, where it provides a unique competitive advantage over other diagnostic imaging methods.

7.1. Idiopathic Pulmonary Fibrosis (IPF)

Idiopathic Pulmonary Fibrosis (IPF) is a progressive scarring disease of the lungs of unknown cause, and this disease is ultimately fatal. IPF affects 13 to 20 people per 100,000 population worldwide. While the IPF diagnosis has been revolutionized by high resolution computed tomography (HRCT), this imaging modality still exhibits significant limitations, particularly in assessing disease progression and therapy response. The need for non-invasive regional lung function assessment has become more acute in light of recently introduced and in-the-pipeline novel therapies. Thus, it will likely be valuable to complement 3D imaging of lung structure with 3D regional assessment of function. This challenge is well addressed by HP ¹²⁹Xe MRI, exploiting the unique properties of this inert gas to image its distribution, not only the airspaces, but also in the interstitial barrier tissues and red blood cells (RBCs). This single-breath imaging exam could ultimately become the ideal, non-invasive tool to assess pulmonary gas-exchange impairment in IPF. Here, we detail the evolution of HP ¹²⁹Xe MRI from its early development to its current state as a clinical research platform. The key imaging biomarkers generated from the ¹²⁹Xe MRI exam for potential utility in IPF diagnosis, prognosis, and assessment of therapeutic response. We conclude by discussing the types of studies that must be performed for HP ¹²⁹Xe MRI to be incorporated into the IPF clinical algorithm and begin to positively impact IPF disease diagnosis and management.

HP ¹²⁹Xe MRI permits the evaluation of heterogeneous, fibrotic thickening of the interstitial barrier tissue and alveolar collapse that collectively impair gas exchange.^[123] A reliable method for concurrent mapping of all three Xe compartments utilizes thoracic ¹H/¹²⁹Xe MRI with one-point Dixon decomposition of multiple tissue resonances to obtain separate images of ¹²⁹Xe in airspaces, barrier, and RBCs.^[120] In a recent study, HP ¹²⁹Xe MRI depicted functional impairment in patients with IPF, whose mean barrier uptake increased by 188% compared with the healthy reference population.^{[118] 129}Xe MRI metrics correlated poorly and insignificantly with CT fibrosis scores but RBC:barrier ratio correlated strongly with diffusing capacity of the lungs for carbon monoxide (DLCO)—indicating a potentially better imaging metric of the disease progression.

In a study investigating the spatial effects of antifibrotic drugs an IPF patient who started on anti-fibrotic treatment one month before baseline MRI, returned five months later for a follow-up scan.^[124] At baseline, the patient presented with 49% of their lung volume exhibiting high barrier uptake, while focal RBC transfer defects at the lung bases accounted for low RBC transfer in 35% of the lung. Upon return five months later, the percentage of lung exhibiting high barrier uptake had improved to encompass only 30% of the lung, while the RBC transfer defects remained stable at 35% of lung volume, Figure 16. Thus, the HP ¹²⁹Xe MRI provided feedback on drug efficacy.

Figure 16.

Observed improvement of ¹²⁹Xe gas exchange metrics for an IPF patient on current therapy. This patient started anti-fibrotic therapy one month prior to baseline MRI and presented with 49% high barrier uptake, and focal RBC transfer defects at the lung bases resulting in 35% low RBC transfer. Upon return 5 months later, the percentage of lung exhibiting high barrier uptake had decreased to 30%, while the RBC transfer defects remained stable at 35% of lung volume. Reprinted with permission from Ref. ^[124].

7.2. Cystic Fibrosis

Cystic Fibrosis (CF) is a genetic disease which carries high morbidity and mortality from lung-function decline. Monitoring disease progression and treatment response in young patients is desirable, but serial imaging via CT is often considered prohibitive. HP ¹²⁹Xe MRI has demonstrated regional ventilation defects in mild CF lung disease with normal FEV₁ (i.e., the amount of air one can force from the lungs in 1 s) and more effectively discriminated CF from controls than FEV₁.^[125] The feasibility, safety, and tolerability of HP ¹²⁹Xe MRI has been demonstrated in children as young as 6 years old.^[126] O₂ saturation (SpO₂) changes in these young patients are consistent with the expected physiological effects of a short anoxic breath-hold and known anesthetic properties of xenon. In addition to single breath HP ¹²⁹Xe MRI scans, multi-breath washout protocols have been reported to be feasible in pediatric CF patients allowing to measure fractional ventilation in this population.^[127] In a more recent study the size and extent of regional ventilation defects in pediatric CF were quantified by HP ¹²⁹Xe MRI and associated with the presence of specific structural abnormalities identified by ultra-short echo time (UTE) ¹H MRI: bronchiectasis (a lung condition where the bronchial tubes are permanently damaged, widened, and thickened), bronchial wall thickening, mucus plugs, and ground glass opacities.^[128]

7.3. Pulmonary Arterial Hypertension and Cardiopulmonary Diseases

A significant subset of patients exhibit concurrent cardiac and pulmonary diseases who are most adversely affected by standard diagnosis and management methods. HP ¹²⁹Xe MRI and MRS methods can provide valuable information to help with diagnostic challenges in cardiopulmonary disease such as left heart failure (LHF) and pulmonary arterial hypertension (PAH), and increase understanding of regional lung function and hemodynamics at the alveolar-capillary level. In addition to spectral parameters of HP ¹²⁹Xe described above, dynamic ¹²⁹Xe MRS provides additional insight, with particular focus on quantifying cardiogenic oscillations in the RBC resonance. Decoupling the spectral parameters and temporal dynamics of the airspace, interstitial barrier, and RBCs confirms that in addition to altered RBC:barrier ratio, all three ¹²⁹Xe resonances are affected by the breathing maneuver, with several RBC spectral parameters exhibiting prominent oscillations at the cardiac frequency.^[129] A study utilizing this methodology demonstrated that COPD, IPF, LHF, and PAH each exhibit unique ¹²⁹Xe MRI and dynamic spectroscopy signatures.^[130] COPD patients demonstrated the most ventilation and barrier defects, whereas IPF patients demonstrated elevated barrier uptake, increased RBC amplitude, and shift oscillations compared to healthy controls. Patients with COPD and PAH both exhibited decreased RBC amplitude oscillations, and LHF was distinguishable from PAH by enhanced RBC amplitude oscillations; see Figure 17.

Figure 17.

Radar plots to display the primary ¹²⁹Xe MRI and spectroscopic signatures associated with (a) healthy volunteers, (b) COPD, (c) IPF, (d) LHF, and (e) PAH. The mean cohort values of the key markers are plotted on one of the six radials: ventilation defect, barrier defect, barrier high and RBC defect percentages derived from imaging, and RBC shift oscillation (SO) and amplitude oscillation (AO) from MRS. Reprinted with permission from Ref. ^[130].

A more recent work extended this methodology by 3D mapping of the magnitude of cardiogenic dynamics using radial MRI to assess regional blood volume fluctuations within the pulmonary microvasculature throughout the cardiac cycle.^[131] The regional heterogeneity of oscillations was found to be sensitive to disease state in PAH and IPF patients, with greater percentages of lungs exhibiting low-amplitude oscillations in PAH patients, and high-amplitude oscillations increase significantly over time in IPF patients. Other emerging pulmonary applications of HP ¹²⁹Xe MRI include localized measurement of lung function impairment caused by covid-19^[132]---currently the subject of clinical trials in Europe and North America (see for example, Refs. ^{[133, 134]}).

7.4. Brain

After inhalation, HP ¹²⁹Xe dissolves in pulmonary blood and is transferred to the brain through systemic circulation. Uptake of inhaled Xe gas into extravascular brain tissue compartment across the blood-brain barrier provides an opportunity to probe regional cerebrovascular physiology. The contrast mechanism of HP ¹²⁹Xe in brain, beyond regional microvascular perfusion, is a function of gas exchange, T₁ relaxation and RF depolarization history of the dissolved signal on route to the brain tissue.

Imaging dissolved HP ¹²⁹Xe in the brain was first demonstrated in healthy humans by optimizing the imaging RF coil and pulse sequences to achieve 20–40 SNR after inhaling 1 sL of HP ¹²⁹Xe.^[136] Similar results were subsequently reported in subjects with established stroke.^[137] HP ¹²⁹Xe washout rate has been explored as a biomarker for Alzheimer’s Disease (AD).^[138] Xe washout parameter is influenced by cerebral perfusion, T₁ relaxation of ¹²⁹Xe and the associated partition coefficient, which all are in turn influenced by AD. Xe washout parameter was extracted by fitting the HP ¹²⁹Xe data to a pharmacokinetic model, showing that Xe washout rate in white and grey matters in AD were approximately half of that in healthy controls. In a recent study, quantitative maps of regional perfusion in healthy human brains were acquired by using time-resolved depolarization of HP ¹²⁹Xe (SNR ~10) and subsequently evaluated to detect changes in cerebral blood flow (CBF) due to a hemodynamic response in response to brain stimuli, Figure 18.^[135]

Figure 18.

Example of ¹²⁹Xe perfusion map acquisition in human brain: (a,f) High-resolution, T₂-weighted ¹H scans for brain localization. (b–d) Three dynamic HP ¹²⁹Xe images acquired 2.5 s, 6.8 s, and 7.1 s after the application of a depolarization RF pulse in the axial projection. (e) The perfusion map in mL of blood per mL of tissue per min. (g–i) Three dynamic images acquired after 1s, 6.5 s, and 8 s in the sagittal view. (j) Perfusion map in the sagittal view. Reprinted with permission from Ref. ^[135].

7.5. Other applications in tissues

The exquisite sensitivity of ¹²⁹Xe’s chemical shift^[53] can be exploited in other applications, e.g. brown fat HP MRI, as demonstrated by Branca and co-workers.^[139] However, in other cases resonances of ¹²⁹Xe dissolved in tissues may be too weak or non-specific to provide desired biomolecular insights. Pines and co-workers^[140] showed that the specificity of Xe could be enhanced by “functionalizing” it with analyte-binding molecular cages (e.g., cryptophanes,^[141] cucurbiturils,^[142] gas-binding protein nanostructures,^[143] etc.^{[144, 145]})—giving rise to unique analyte-bound ¹²⁹Xe resonances.^{[144, 145]} Implementing chemical exchange saturation transfer (Hyper-CEST)^[146] allows the strong “bulk” ¹²⁹Xe signal to be encoded with the much-weaker response of the analyte-bound Xe biosensor, greatly improving the detection sensitivity. In vivo HyperCEST using Xe-binding molecular cages has now been demonstrated, Figures 19 and 20.^[147] The approach offers great promise for extending HP ¹²⁹Xe MRI to molecular imaging of a wide range of biomarkers.

Figure 19.

a) HP ¹²⁹Xe is mixed with O₂; b) HyperCEST agent is administered via tail-vein injection; c) HP ¹²⁹Xe is administered via inhalation; d) model “Analyte-bound” ¹²⁹Xe resonance is selectively irradiated; e) corresponding MRI image is encoded. Image courtesy of Prof. M. Albert, reproduced with permission from Ref. ^[147]

Figure 20.

a) HyperCEST saturation map of a rat abdomen zoomed in and overlaid on a ¹H MRI showing the location of the CB6 cage contrast agent. b) Same as (a) but of the rat brain. Image courtesy of Prof. M. Albert, reproduced with permission from Ref. ^[147].

8. Conclusions

Hyperpolarization of ¹²⁹Xe via SEOP on a clinical scale has been achieved by CF and SF approaches, which have matured and transitioned to commercialization of the enabling instrumentation technologies. A number of clinical studies and trials have already demonstrated the efficacy of HP ¹²⁹Xe MRI and MRS for diagnosis, prognosis, and monitoring response to treatment of several diseases. We envision that the regulatory approval for diagnostic use of HP ¹²⁹Xe MRI will be in place in the US, EU, UK and other countries in the near future for this game-changing diagnostic modality. While much work will remain to be done for this technology to translate from the leading pioneering sites to the clinical realm to make it available for the general population, the production technologies are certainly ready to support production of HP ¹²⁹Xe at clinical scale.

Biography

Ali Khan completed his MSci in Physics at the University of Nottingham in 2020. He is starting a PhD at the University of Cambridge investigating MR hyperpolarisation technology to profile the metabolic flux of glioblastoma supervised by Prof. Ferdia Gallagher and Dr. Mary McLean.

Rebecca Harvey studied Physics at the University of Nottingham, obtaining her MSc in 2020.

Prof. Eduard Y. Chekmenev, PhD in Physical Chemistry (supervisor Prof. Richard J. Wittebort) 2003, University of Louisville, KY, USA. Postdoctoral Fellow at NHMFL in Tallahassee, FL (Prof. Timothy Cross), Caltech (Prof. Daniel P. Weitekamp) and HMRI (Dr. Brian D. Ross). In 2009, Dr. Chekmenev started his hyperpolarization program at Vanderbilt University, and he was tenured in 2015. In 2018, he moved to Wayne State University (Detroit, MI) to continue his research on MR hyperpolarization. Research interests include development of methods of hyperpolarization and their Biomedical and industrial use.