Hyperpolarized gas MRI in pulmonology

Abstract

Lung diseases have a high prevalence amongst the world population and their early diagnosis has been pointed out to be key for successful treatment. However, there is still a lack of non-invasive examination methods with sensitivity to early, local deterioration of lung function. Proton-based lung MRI is particularly challenging due to short T₂^* times and low proton density within the lung tissue. Hyperpolarized gas MRI is aan emerging technology providing a richness of methodologies which overcome the aforementioned problems. Unlike proton-based MRI, lung MRI of hyperpolarized gases may rely on imaging of spins in the lung’s gas spaces or inside the lung tissue and thereby add substantial value and diagnostic potential to lung MRI. This review article gives an introduction to the MR physics of hyperpolarized media and presents the current state of hyperpolarized gas MRI of ³Headvasd and ¹²⁹Xe in pulmonology. Key applications, ranging from static and dynamic ventilation imaging as well as oxygen-pressure mapping to ¹²⁹Xe dissolved-phase imaging and spectroscopy are presented. Hyperpolarized gas MRI is compared to alternative examination methods based on MRI and future directions of hyperpolarized gas MRI are discussed.

Introduction

Lung diseases cause a significant burden to the world population, constituting 4 out of the 10 leading causes of death in the world. Chronic obstructive pulmonary disease (COPD) has been estimated by the World Health Organization to be the fourth leading cause of death in the year 2015, likely rising to third leading cause of death by 2030.^{1, 2} While early diagnosis was recognized as key in treatment of COPD,^{3, 4} the diagnosis of early disease proves to be difficult due to an absence of sensitive, non-invasive tests. Spirometry as a global measurement lacks sensitivity for early, mostly local deterioration of lung function. CT and scintigraphy inevitably bring about exposure to ionizing radiation and are therefore not ideal for application in children and in longitudinal clinical follow-up or longitudinal research studies with multiple measurements. Further, while CT is well-suited to delineate fine anatomical structures, functional information is usually not derived.

MRI of the lungs is an emerging technology; however, conventional proton-based MRI suffers from low proton density in the lungs and rapid signal decay due to a large number of jumps in magnetic susceptibility within the lung tissue, which leads to low signal-to-noise ratio (SNR). In addition, lung and heart motion, may contribute to significant artefacts in the lung parenchyma. MRI of hyperpolarized gases overcomes most of these limitations and provides a richness of methodologies for assessing lung function locally and non-invasively. Besides static spin density ventilation imaging, hyperpolarized gas MRI provides information on ventilation dynamics, lung microstructure as well as perfusion and alveolar-capillary diffusion. It is applicable in a wide range of diseases including, but not limited to COPD, asthma, cystic fibrosis (CF) and idiopathic pulmonary fibrosis (IPF).

Hyperpolarization refers to the alignment of nuclear spins inside a sample, enabling imaging of various gases despite their intrinsically low spin density. Most of the research in the field of hyperpolarized gas MRI has focused so far on the two spin-1/2 isotopes ³He and ¹²⁹Xe, although the quadrupolar nucleus ⁸³Kr shows surface-sensitive nuclear magnetic resonance properties, which could make it valuable for application in pulmonology.^5–8

When human hyperpolarized gas MRI studies are carried out, the subject typically inhales the gas from a small plastic bag while positioned inside the bore, and the measurement starts immediately at breath-hold. Helium remains almost entirely within the air spaces after inhalation and does not cause any serious side effects after a single breath. Unfortunately, the scarcity and high cost of ³He has prevented a widespread clinical use so far. Xenon, on the contrary, is relatively well-soluble in aqueous tissue and blood and, once inhaled, follows a path similar to that of oxygen in the body. While the inhalation of xenon at certain doses causes well-known anaesthetic side effects, this property in combination with the broad range of chemical shift in ¹²⁹Xe Larmor frequency makes ¹²⁹Xe dissolved-phase imaging a unique tool for probing gas uptake locally and non-invasively. Both helium and xenon are tolerated well when inhaled in doses as used for MRI by healthy volunteers and patients suffering from COPD.^9–11

This review article will first give an introduction to the physics of hyperpolarization of gases followed by current and potential clinical applications in pulmonology, which are summarized in Table 1. Lastly, hyperpolarized gas MRI is compared with some possible alternative MRI techniques.

Table 1.

Common pulse sequences/imaging methods for hyperpolarized gas MRI along with a suggestion of clinical applications

Pulse sequence	Nucleus	Possible applications
Static ventilation imaging	3He, 129Xe	COPD, Asthma, CF
Multibreath imaging	3He, 129Xe	COPD, Asthma, CF
pO2 mapping	3He, 129Xe	COPD
ADC mapping	3He, 129Xe	COPD, Asthma
Dissolved-phase imaging	129Xe	COPD, Asthma, IPF, RILI, PH
CSSR	129Xe	COPD, RILI
XTC	129Xe	COPD

ADC, apparent diffusion coefficient; CF, cystic fibrosis; COPD, chronic obstructive pulmonary disease; CSSR, chemical shift saturation recovery; IPF, idiopathic pulmonary fibrosis; PH, pulmonary hypertension; RILI, radiation-induced lung injury; XTC, xenon transfer contrast.

Physics of hyperpolarized media

Polarization and SNR

One of the fundamental limitations of clinical MRI is its low SNR  due to the large thermal noise in vivo compared to the MR signal. It is instructive to consider the fraction of the pool of nuclear spins inside a sample, which contributes to the MR signal. If we consider a sample of water containing protons inside a homogeneous magnetic field B₀ along the z-axis and we measure the orientation of a large amount of protons along that axis, in thermal equilibrium we will find a Boltzmann distribution of spin states,

$${\displaystyle \frac{n_{+}-n_{-}}{n_{+}+n_{-}}=\tanh \frac{\mu B_0}{k_{B}T}\approx \frac{\mu B_0}{k_{B}T}}$$

where n₊ denotes the number of spins found with orientation parallel to B₀, n_– the number of spins antiparallel to B₀, k_B the Boltzmann constant, T the temperature of the sample and μ the nuclear magnetic dipole moment.¹² After application of a radio frequency (RF) pulse, only the fraction

$${\displaystyle \frac{n_{+}-n_{-}}{n_{+}+n_{-}}\equiv P\propto M_{0,z}}$$

of the total amount of spins, which is called the polarization P of the sample, will contribute to the net transverse magnetization and hence MR signal. Increasing the degree of polarization by four to five orders of magnitude will hence greatly increase SNR and more than compensate for the low spin density and small nuclear magnetic dipole moment of gases like ³He and ¹²⁹Xe when compared to ¹H in water, Figure 1(a).

Figure 1.

(a) Parallel and antiparallel orientation of spins in a magnetic field corresponds to two different energy levels. Nonetheless, due to the small energy difference, the populations are almost equal in thermal equilibrium, which can be seen in the left diagram. Hyperpolarization can be achieved by SEOP (see main text) leading to the distribution in the right diagram but will be lost by relaxation to thermal polarization over time. (b) Hyperpolarized media show a fundamentally different behaviour during a pulse sequence (right diagram) when compared to thermally polarized media (left diagram). In the case of hyperpolarization, spin-lattice relaxation will decrease longitudinal magnetization to thermal equilibrium M_eq. Longitudinal magnetization diminished by irradiation of RF pulses (circles at top of diagram) will not be recovered assuming complete spoiling. Whereas thermally polarized media reach a finite steady-state magnetization M_ss, both equilibrium and steady-state magnetization of hyperpolarized media can be seen as practically zero. RF, radio frequency; SEOP, spin-exchange optical pumping.

Spin-exchange optical pumping

The most obvious way of achieving hyperpolarization of a sample—according to Equation 1—is putting it into a very strong magnetic field at very low temperature for some time and warming it up quickly (brute force polarisation) and this approach works indeed.¹³ Usually, however, hyperpolarization of the noble gases ³He and ¹²⁹Xe is achieved by a technique called spin-exchange optical pumping (SEOP).¹² An optical cell is employed which contains a small amount of the noble gas, buffer gases and an alkali metal—typically rubidium. The rubidium gets vapourized once the optical cell is heated and irradiated by a high power laser. The laser light is circularly polarized and may excite and thereby align the Rubidium’s 5 s¹ electrons in a magnetic field. Spin is then exchanged between a rubidium’s electron and the noble gas nucleus via a hyperfine interaction, mostly during collisions of rubidium atoms and noble gas atoms, Figure 2.

Figure 2.

The SEOP process is a two-step process, consisting of optical pumping and subsequent spin exchange. Circularly polarized laser light excites the outer electrons of rubidium atoms contained in an optical cell, which leads to an alignment of electron spins in the magnetic background field penetrating the cell (optical pumping). This electron can exchange its spin with a noble gas nucleus during a collision of both atoms. This leads to a depolarization of rubidium electron spins and the process can start from the beginning (adapted from Mugler and Altes¹⁴). SEOP, spin-exchange optical pumping.

This exchange is much more likely for ¹²⁹Xe than for ³He, making the polarization process of ¹²⁹Xe much faster. A technical difficulty of ¹²⁹Xe polarization is the requirement of a continuous gas flow through the optical cell and subsequent freeze-out of ¹²⁹Xe unlike in the case of ³He, where the gas resides inside the cell for many hours.

Pulse sequences design considerations for hyperpolarized media

Assuming an initial magnetization M₀ of a sample parallel to B₀, directly after irradiation of an RF pulse with flip angle α the longitudinal magnetization is diminished according to the equation

$${\displaystyle M_{1,z}=M_{0,z}\cos \alpha }$$

whereas the transverse magnetization

$${\displaystyle M_{1,\mathrm{\perp }}=M_{0,z}\sin \alpha }$$

is created. Assuming no change in longitudinal magnetization and complete spoiling, the transverse magnetization after the second RF pulse is given by

$${\displaystyle M_{2,\mathrm{\perp }}=M_{1,z}\sin \alpha =M_{0,z}\cos \alpha \sin \alpha }$$

which is reduced in comparison to M₁,_⊥ by a factor cos α. After irradiation of a train of N RF pulses, the transverse magnetization M_N,_⊥ will have been diminished compared to M₁,_⊥ by a factor cos^N α. This results in a correspondingly diminished SNR for data sampled after this pulse, Figure 1(b).

Proton-based MRI typically makes use of spin-lattice relaxation back to a steady-state longitudinal magnetization in order to maintain high SNR throughout the entire acquisition by introducing waiting periods in the pulse sequence. By contrast, spin-lattice relaxation does not increase longitudinal magnetization in the case of hyperpolarized media, but rather decreases it to thermal equilibrium, making any waiting periods detrimental. For ³He and ¹²⁹Xe, the longitudinal magnetization in thermal equilibrium is approximately zero when compared to the hyperpolarized state, so an efficient handling of the available reservoir of initial magnetization is most important. This and the previously described type of signal decay may lead to unwanted k-space filtering and hence image artefacts if not properly accounted for during data acquisition and reconstruction.^{15, 16}

While hyperpolarized gas can be stored for a period ranging from minutes to hours or even days depending on the type of gas, buffer gas admixture, isotopic composition, magnetic environment and state of matter without losing substantial polarization,^17–19 longitudinal magnetization is always rapidly diminished with a time constant on the order of seconds¹⁴ once the gas is inhaled by a subject due to interactions with paramagnetic oxygen.

Pulse sequences for imaging of hyperpolarized gases are typically based on gradient-recalled echoes—spin-echoes are generally not appropriate. Fully balanced pulse sequences making use of transverse magnetization from previous excitations provide particularly high SNR (approximately three-fold compared to spoiled gradient echo) in hyperpolarized gas imaging but may suffer from image blurring at high-resolution due to diffusion-weighting introduced by spatial encoding gradients and from banding artefacts mostly near the diaphragm in lung imaging,^20–22 Figure 3. TrueFISP sequences can be expected to be better suited for ¹²⁹Xe than for ³He imaging due to its lower diffusivity and gyromagnetic ratio.

Figure 3.

Representative slices for hyperpolarized ¹²⁹Xe 3D ventilation imaging in a healthy volunteer (39 years, male) using a TrueFISP sequence. Acquisition parameters included TR/TE 3.73/1.8 ms, matrix size 144 × 144 × 96, isotropic resolution 2.4 mm. No clear signs of banding are visible with the stated parameters at 1.5 T. Slight blurring is visible in anteroposterior direction, partly attributable to signal decay during acquisition and slow encoding in that direction. The ventilation pattern is mainly homogeneous as expected. Yet, small ventilation defects can be seen in the basal and posterior regions of the lung (arrows), possibly due to airway closure at low lung inflation level, from which the inspiration started.

Clinical applications of hyperpolarized gas MRI

Ventilation imaging

Both hyperpolarized ³He and ¹²⁹Xe are suited extraordinarily well for high-resolution ventilation imaging due to the very high SNR in the hyperpolarized state.^{23, 24} Thus, hyperpolarized gas ventilation MRI is a sensitive imaging technique that reveals the distribution of ventilation within the lung in exquisite detail, allowing regional ventilation heterogeneity to be assessed. It is common to all presented methods that they can provide a quantification of lung function, making them highly relevant for precision medicine.

Static ventilation imaging

The idea behind static ventilation imaging is that a patient inhales hyperpolarized noble gas and holds his/her breath for the time it takes to acquire a high-resolution image of the gas, which is roughly on the order of 5–20 s. B₁Ifinhomogeneities are corrected for, the signal intensity usually represents in good approximation the hyperpolarized gas spin density and hence the amount of gas having reached a particular region inside the lung after inhalation. Healthy subjects typically show a homogeneous gas distribution pattern, although aging effects may introduce small subsegmental ventilation defects (Figure 4).^{25, 26} Minor ventilation defects can be seen also in normal subjects and these tend to increase in number during immobilization in the scanner bore.²⁷

Figure 4.

Hyperpolarized ¹²⁹Xe ventilation imaging (spoiled gradient echo) in healthy volunteer (69 years, female). Several small but prominent ventilation defects are visible in the periphery of the lungs (arrows), likely due to normal aging. Spirometric results are normal (FVC 121.6% pred., FEV₁ 119.5% pred.). FVC, forced vital capacity; FEV₁, forced expiratory volume in one second.

In COPD patients the ventilation pattern shows a high degree of inhomogeneity, depending upon disease severity, Figure 5. Patients suffering from asthma also show ventilation defects and an improvement in some ventilation defects can usually be observed after inhalation of a bronchodilator.^28–31 There is evidence based on hyperpolarized gas MRI that in asthmatics ventilation defects often persist for years in the very same regions of the lung.²⁵

Figure 5.

(a) Hyperpolarized ¹²⁹Xe ventilation imaging (overlay in thoracic cavity) in a patient suffering from COPD (76 years, male; FVC: 84.0% pred., FEV₁: 51.3% pred.) and ¹H anatomical imaging in the same breath-hold. (b) CT imaging from the same subject with emphysematous regions inside the lung (radio density less than–950 Hounsfield units) marked by  overlay. Some emphysematous regions in CT correspond to ventilation defects (arrows). Other emphysematous regions are ventilated, but show a high degree of ventilation heterogeneity (asterisks). COPD, chronic obstructive pulmonary disease.

The most common quantitative readout parameter of static ventilation imaging is ventilation defect percentage (VDP), which is the fraction of voxels in the lung showing signal below a certain threshold, Figure 6. Calculating VDP in practice requires co-registered ¹H and hyperpolarized gas images normalized for B₁ effects.^32–34 VDP is a highly reproducible metric in hyperpolarized gas imaging^{35, 36} and it is predictive for exacerbations in COPD³⁷ and for outcomes in asthma.³⁸ Figure 7 shows the effects of montelukast treatment in exercise-induced bronchoconstriction using ³He MRI.³⁹

Figure 6.

Ventilation imaging as in Figure (5a) but with binned ¹²⁹Xe signal. The binning only applies inside the thoracic cavity as determined from ¹H data by a region-growing algorithm. Bins denote signal 0–20% (ventilation defect), 20–40% (low ventilation), 40–80% (normal ventilation) and over 80% of the 99th percentile of ¹²⁹Xe signal inside the thoracic cavity as in He et al³². Ventilation defect percentage in this subject with COPD is 27.9%. COPD, chronic obstructive pulmonary disease.

Figure 7.

Study of the effects of montelukast treatment in a subject with exercise-induced bronchoconstriction using hyperpolarized ³He MRI. VDP is strongly elevated after exercise challenge. The effects of treatment are clearly visible by reduced VDP after challenge compared to placebo. Figure adapted from Kruger et al³⁹ with permission from John Wiley & Sons. FEV₁, forced expiratory volume in one second; VDP, ventilation defect percentage.

It is important to note, however, that since the diameter of airways in the lungs depends on the inspiratory level, the gas flow kinetics and final distribution of gas depend on the breathing manoeuvre and timing of hyperpolarized gas bolus delivery.^40–42 A standardization across clinical sites should aid in clinical translation of the technique.

In children with mild cystic fibrosis lung disease hyperpolarized gas ventilation MRI detected ventilation defects when abnormality was often not detectable on either CT or lung clearance index.^{33, 43} Also, this technique appears to identify longitudinal changes in early cystic fibrosis lung disease prior to other physiological methods of early lung disease detection.⁴⁴ The functional information obtained using hyperpolarized gas ventilation MRI may help understand the pathophysiology of a variety of lung diseases and may help to guide therapy in the future. Hyperpolarized ³He MRI has been used to study the effects of ivacaftor treatment in children with cystic fibrosis, Figure 8.⁴⁵

Figure 8.

Hyperpolarized ³He ventilation imaging in a child with cystic fibrosis. (a) Baseline imaging. (b) Imaging after 4 weeks of ivacaftor treatment. (c) Imaging after additional 2 weeks of washout with placebo. Figure adapted from Altes et al⁴⁵ with permission.

Dynamic ventilation imaging within a breath-hold

Static ventilation imaging records a single gas distribution pattern for one breath-hold, neglecting a possible movement of gas inside the lungs after inhalation. While the approximation of a static gas distribution pattern might be good in healthy subjects, it has been shown that there is a considerable amount of delayed ventilation in patients suffering from COPD. There is evidence that this delayed ventilation is based upon collateral ventilation and one might expect differences between ³He and ¹²⁹Xe in this regard due to the wildly different diffusivities. Studying ventilation dynamics within a single breath-hold may provide deeper insights into lung pathophysiology and also increase the accuracy of a measurement of VDP, which does not take delayed ventilation into account.^{46, 47}

Mapping of oxygen partial pressure

The longitudinal relaxation rate R₁ of hyperpolarized ³He gas after inhalation is inversely proportional to the partial pressure of oxygen as a paramagnetic substance,^{48, 49}

$${\displaystyle \frac{R_1}{1/\mathrm{s}}=\frac{\mathrm{p}{\mathrm{O}}_2/\mathrm{m}\mathrm{b}\mathrm{a}\mathrm{r}}{0.75{\left(T/\mathrm{K}\right)}^{1.42}}\approx \frac{\mathrm{p}{\mathrm{O}}_2/\mathrm{m}\mathrm{b}\mathrm{a}\mathrm{r}}{2587}}$$

with the approximation valid at body temperature. Acquiring multiple ventilation images in a single breath-hold with certain delay times hence enables a mapping of T₁ and thus _PO₂ by measuring the signal decay per pixel. Using the principles of _PO₂ mapping, also a V/Q measurement can be performed, yielding valuable, clinically relevant information for pneumologists.⁵⁰ Signal decay by relaxation must be differentiated from signal decay by irradiation of RF pulses as in Equation 3. Care must be taken that the measured _PO₂ values are not skewed by cardiac motion⁵¹ or by gas flowing between pixels or between slices, especially in the case of patients suffering from COPD, since there is a non-negligible amount of delayed ventilation. ¹²⁹Xe might be superior to ³He for mapping _PO₂ due to its lower diffusivity although xenon uptake by the lungs complicates the signal dynamics.^{47 129}Xe shows an additional field dependence of oxygen relaxation rate and it is not entirely clear if this dependence also exists in the case of ³He.⁵²

Dynamic ventilation imaging during multiple breaths

Relatively slowly filling ventilation defects can be challenging to assess by time-resolved ventilation imaging within one breath-hold. Several multibreath imaging techniques have been employed, allowing for quantitative ventilation measurement. During each breath, a certain volume fraction of gas is replaced by fresh gas at a given voxel location. This volume fraction, the fractional ventilation, is a quantitative measure of regional ventilation, which can be obtained by the gas signal reduction or build-up after every other breath (Figure 9).^53–55 Fractional ventilation mapping enables the assessment of air trapping and delayed ventilation in obstructive lung disease. Apart from lung movement, which leads to localization errors over multiple breathing cycles, signal decay due to relaxation may be a challenge in quantitative multibreath imaging using hyperpolarized gases.

Figure 9.

FV maps obtained in a volunteer (33 years, female) using hyperpolarized ¹²⁹Xe wash-out imaging with prescribed breathing manoeuvres and breath-hold during imaging. Signal decay was compensated for using an additional calibration scan. FV exhibits a clearly visible gravitational gradient due to the subject’s supine position. FV, fractional ventilation.

Diffusion-weighted imaging

One of the fundamental differences between gas MRI and standard ¹H MRI is the vastly higher diffusivity of the gases, with ³He having a diffusion coefficient of 0.86 cm² s^–1 when dilute in air and ¹²⁹Xe still having a diffusion coefficient of 0.14 cm² s^–1.⁵⁶ This must be compared to the self-diffusivity of water of ~3.0⋅10^–5 cm² s^–1 at body temperature.⁵⁷ The MR pulse sequence can be sensitized for diffusion by introducing gradient pulses with vanishing gradient moment of zeroth order before signals are acquired. The gradients have the effect that phase coherence and consequently the MR signal magnitude is reduced when spins have diffused through them and thereby changed their Larmor frequency.

Apparent diffusion coefficient mapping

Simple diffusion-weighted imaging assumes a mono-exponential decay of the MR signal with increasing diffusion-weighting b and increasing apparent diffusion coefficient (ADC). Lower ADCs generally indicate a smaller average diameter of the alveoli inside a voxel and higher ADCs indicate greater diameters or possibly emphysematous destruction of the lung parenchyma. While the microscopic diffusion coefficient of gases depends primarily on gas composition, the ADC also includes contributions from diffusion restriction at alveolar walls. Measurements of ADC using ³He diffusion-weighted MRI show sensitivity to subclinical disease in healthy smokers^{58, 59} and to more severe emphysematous tissue destruction.⁶⁰ They further enabled measurement of human lung growth during adolescence by the detection of an increase of ADC with age, which indicates a reduction of alveolar surface-to-volume ratio.⁶¹ Through the lower microscopic diffusion coefficient, ¹²⁹Xe diffusion-weighted MRI probes shorter length scales of diffusion restriction compared to ³He MRI assuming equal pulse sequence timing, Figure 10.⁶²

Figure 10.

ADC mapping in (a) a healthy volunteer, (b) an age-matched healthy control and (c) a COPD patient with emphysema. The elevation of ADC in COPD due to emphysematous tissue destruction is clearly seen. Figure adapted from Kaushik et al⁶² with permission from John Wiley & Sons. ADC, apparent diffusion coefficient; COPD, chronic obstructive pulmonary disease.

Lung morphometry using multi-b diffusion-weighted imaging

While conventional ADC mapping of hyperpolarized gases shows high sensitivity for various lung diseases, there is usually no obvious link to widely-used lung microstructure parameters such as those of the Weibel model or other standard parameters as obtained by invasive lung biopsy specimens. When the MR signal is measured at multiple different b values, deviations of the MR signal decay from the classical mono-exponential model can be observed,^{63, 64} providing information on lung microstructure, e.g. the Weibel parameters (acinar airways radii and alveolar depth) and standard metrics (mean linear intercept, surface-to-volume ratio and alveolar density) that are widely used by lung researchers but were previously available only from invasive lung biopsy. Very high SNR in the underlying images is required for an accurate determination of microstructure parameters⁶⁵ and multi-b measurements can make multiple inhalations of ¹²⁹Xe necessary.⁶⁶ Compressed sensing and undersampling of k-space can be helpful in making the breath-hold times needed for multi-b measurements clinically feasible.⁶⁷

Dissolved-phase imaging and spectroscopy

Dissolved-phase imaging is unique to hyperpolarized ¹²⁹Xe, as unlike helium, xenon diffuses through the alveolar walls and can be dissolved in aqueous tissues and blood and thereby shifts its Larmor frequency by ~200 ppm, Figure 11. It should be noted that at every instant of time only a small fraction of xenon is dissolved and the vast majority remains in the gas-phase. Increasing levels of polarization have enabled a direct imaging of this dissolved-phase after a single inhalation of hyperpolarized ¹²⁹Xe in humans. Spectroscopic measurements like chemical shift saturation recovery (CSSR) have been there for a longer time⁶⁸ and also show differences between healthy and diseased lung.

Figure 11.

Spectrum of inhaled hyperpolarized ¹²⁹Xe in a healthy volunteer (27 years, female). Three distinct resonances are clearly seen, corresponding to gaseous ¹²⁹Xe, dissolved in tissue and blood plasma (barrier) and bound to haemoglobin (RBC), which together form the dissolved-phase. Spectrum was acquired after highly frequency-selective excitation, reducing the apparent gas-phase peak by several orders of magnitude.

Chemical shift saturation recovery

The CSSR pulse sequence saturates the dissolved compartment and probes the dissolved-phase magnetization build-up by irradiating an excitation pulse after a certain delay time. The magnetization of spins coming from the pool of gaseous xenon can be thought of as a label, flagging them as having just diffused into the lung tissue or blood. In this way, the process of gas uptake can be directly investigated. Several different models have been suggested for obtaining physiological parameters describing lung microstructure and blood flow from the xenon uptake as a function of CSSR delay time,^69–71 Figure 12.

Figure 12.

Hyperpolarized ¹²⁹Xe CSSR data showing gas uptake in the lung in barrier and RBC-phase. Solid lines show Patz model fits to the barrier and RBC data, respectively. CSSR, chemical shift saturation recovery.

Experimental results for these model parameters in healthy volunteers agree well with values obtained by histology ex vivo, although some constants going into the model are not known precisely and more validation of the models is needed.^{70, 72} Some model parameters can be assessed with very good reproducibility, however, and the sensitivity of the method for lung diseases even in the subclinical range and for normal aging has been shown in several studies.^73–76

The lacking localization poses a limitation of standard CSSR spectroscopy, which can be overcome using a multichannel receive coil, coil sensitivity mapping and dedicated algorithms.⁷⁵

Dissolved-phase imaging

A simultaneous imaging of hyperpolarized ¹²⁹Xe remaining in the airspaces and dissolved in lung tissue/blood provides a local measurement of gas uptake inside the lung. Early work used the chemical shift artefact in Cartesian k-space sampling for separation of gaseous ¹²⁹Xe and dissolved ¹²⁹Xe in the images.⁷⁷ While being highly robust against off-resonance, this technique limits selectable sequence parameters considerably and does not separate ¹²⁹Xe dissolved in lung tissue and blood plasma (barrier) from ¹²⁹Xe bound to haemoglobin (RBC) with an additional ~20 ppm shift. Today most implementations use a highly frequency-selective excitation to separate gas-phase signal and the far smaller dissolved-phase signal.⁷⁸ The chemical shifts of ¹²⁹Xe in other species are in general different from those in humans, which has an impact on dissolved-phase imaging in these species.⁷⁹

A significant improvement in diagnostic potential of ¹²⁹Xe dissolved-phase imaging could be made by introducing Dixon-method techniques for separating the dissolved compartment further.^{80, 81} A combined acquisition of ¹²⁹Xe in gas, tissues and RBCs then allows assessment of ventilation, diffusion and perfusion at the same instant of time, Figure 13. Dissolved-phase imaging has the potential to distinguish normal and abnormal regions of the lung in many diseases like COPD, asthma and idiopathic pulmonary fibrosis.^76,82–84

Figure 13.

Functional ratio maps as obtained by hyperpolarized ¹²⁹Xe dissolved-phase imaging in a healthy volunteer (26 years, female). The individual subfigures show the ratio of ¹²⁹Xe in (a) barrier and gas-phase, (b) RBC and gas-phase and (c) barrier and RBC-phase. Gravitational gradients are visible in all three maps.

Robertson et al have recently suggested that the barrier resonance in the dissolved-phase ¹²⁹Xe spectrum actually consists of two resonances 4.7 ppm apart.⁸⁵ This would pose a limitation of present implementations of ¹²⁹Xe dissolved-phase imaging sequences but more research on this topic seems necessary. Other limitations are the relatively low achievable spatial resolution, partly due to the very short T₂^* times, and the fact that often hypo-ventilated regions in the lung–mostly regions with pronounced lung pathology–cannot be assessed.

Imaging of other organs than the lung

Increasing levels of ¹²⁹Xe polarization have made it possible to perform imaging of ¹²⁹Xe dissolved in organs other than the lung after inhalation of hyperpolarized ¹²⁹Xe. Well-perfused organs like brain^86–88 and kidneys⁸⁹ (Figure 14) are particularly suitable candidates for an extension of the diagnostic potential of hyperpolarized ¹²⁹Xe MRI and MR spectroscopy. Preliminary work has suggested existence of quite different chemical shifts of ¹²⁹Xe in the kidney compared to the lung.⁹⁰ A different approach for imaging of hyperpolarized ¹²⁹Xe outside the lung is binding it to cryptophane cages.⁹¹ Recently, Zheng and co-workers have even proposed an entirely new imaging modality based on the anisotropic emission of Y-rays from hyperpolarized quadrupolar nuclei like ^131mXe, which might be suitable for this purpose.⁹² Further work is necessary to investigate the diagnostic potential of hyperpolarized gas MRI in organs other than the lung.

Figure 14.

Hyperpolarized ¹²⁹Xe dissolved-phase imaging in the upper abdomen of a healthy volunteer (31 years, male) as overlay with ¹H imaging for anatomical reference in background. The kidneys are clearly depicted along with other organs.

Comparison with possible alternatives

MRI of fluorinated gases

Imaging of fluorinated gases has been performed already in the early days of MRI⁹³ and a number of different fluorinated gases have been used for ¹⁹F lung MRI, e.g. SF₆,⁹⁴ C₃F₈,⁹⁵ C₂F₆⁹⁶ and CF₄.⁹⁷ Unlike in the case of ¹²⁹Xe and ³He, in vivo lung MRI of ¹⁹F is possible also without hyperpolariation of the gas due to its high gyromagnetic ratio, its high natural abundance, the fact that multiple ¹⁹F nuclei per molecule contribute to the MR signal and due to its short T₁ times of 1–30 ms, depending on the molecule used and environmental conditions.⁹⁸ The relatively strong signal obtained with thermal polarization makes ¹⁹F MRI particularly well-suited for quantification of gas wash-in and wash-out kinetics over multiple breathing cycles, which enables regional ventilation quantification also in (partially) obstructed lung segments in COPD patients.^{99, 100}

The achievable resolution in ¹⁹F static ventilation imaging is considerably lower than with hyperpolarized ³He or ¹²⁹Xe and the solubility in lung tissue of the aforementioned fluorinated gases is negligible compared to xenon.^{101, 102} Nonetheless, Adolphi and Kuethe have described an interesting method for V/Q measurement by performing ¹⁹F T₁ mapping.¹⁰³

Oxygen-enhanced MRI

Molecular oxygen is paramagnetic and serves as a contrast agent when inhaled, similar to gadolinium. The primary mechanism for signal intensity change after inhalation of pure oxygen is the dissolution of oxygen within arterial blood, which can shorten T₁ by approximately 9–15% in healthy individuals.¹⁰⁴ A time series of images of the lung provides the wash-in time of oxygen or the maximum signal enhancement. These quantities have been shown to be sensitive for airway inflammation¹⁰⁵ and narrowing of airways in obstructive lung disease.^{106, 107} Oxygen-enhanced MRI is also sensitive for detection of chronic lung allograft dysfunction already in an early stage.¹⁰⁸

The oxygen induced signal changes are likely due to a mixture of ventilation, perfusion and oxygen diffusion into the blood,¹⁰⁹ which can be challenging and further research is necessary for clinical translation of this technique.

Fourier decomposition MRI

Fourier decomposition (FD) lung MRI is a method which does not require any contrast agent and which relies on continuous imaging during free breathing. Provided that this imaging process is fast enough, application of a Fourier transformation on the data in the time dimension can extract pulsations of the lung parenchyma due to respiration and perfusion. A crucial step in extracting physiological quantities is the registration of the images after reconstruction using a non-rigid deformation.^{110, 111} Ventilation can be quantified by means of Fourier decomposition MRI using a suitable normalization taking breathing tidal volume into account.¹¹² Other techniques measure the proton density for quantification of ventilation independent of respiration.¹¹³ Dynamic ventilation and perfusion information can be obtained by sorting images into full cardiac and respiratory cycles.¹¹⁴ It could be shown that perfusion-weighted FD MRI is feasible for diagnosing chronic pulmonary emboli¹¹⁵ and ventilation-weighted FD MRI for detection of chronic lung allograft dysfunction.¹¹⁶ Ventilation-weighted FD MRI could also be validated using ¹⁹F MRI¹¹⁷ and single photon emission CT.¹¹⁸

FD MRI is very cost-efficient and provides a high patient comfort. Unlike hyperpolarized gas MRI, FD can only measure ventilation indirectly by MR signal variations in the lung parenchyma, however. The high frame rate necessary for FD MRI makes 3D acquisitions¹¹⁹ difficult whereas 2D acquisitions suffer from problems of through plane motion during image acquisition.

Conclusion and outlook

Numerous possible applications in respirology and quantitative biomarkers obtained by hyperpolarized gas MRI have been described in recent years with a variable degree of maturity. Some of these biomarkers have the potential to become of crucial importance for personalized medicine and many clinical research applications in pulmonology. It has also been shown that hyperpolarized gas MRI can outperform the diagnostic potential of other routinely applied examination methods.

Due to its higher cost and scarcity, a widespread clinical use of ³He MRI appears to be less likely although there is the possibility of recovering exhaled ³He with high efficiency.¹²⁰ In recent years, the advent of improved polarizer technology for ¹²⁹Xe accelerated research in the field and has enabled an even greater diversity of applications in hyperpolarized gas MRI. Improved commercial availability of hyperpolarizers and higher polarization levels with reduced gas doses and isotopic ratio, make a translation of ¹²⁹Xe ventilation imaging to the clinic economically feasible. Further work has to be carried out by the polarizer manufacturers, having currently completed phase-II trials for hyperpolarized ¹²⁹Xe administration,^{121, 122} in order to obtain regulatory approval for administration of hyperpolarized gases in routine clinical imaging.