Diverse cardiopulmonary diseases are associated with distinct xenon magnetic resonance imaging signatures

Abstract

Background

As an increasing number of patients exhibit concomitant cardiac and pulmonary disease, limitations of standard diagnostic criteria are more frequently encountered. Here, we apply noninvasive ¹²⁹Xe magnetic resonance imaging (MRI) and spectroscopy to identify patterns of regional gas transfer impairment and haemodynamics that are uniquely associated with chronic obstructive pulmonary disease (COPD), idiopathic pulmonary fibrosis (IPF), left heart failure (LHF) and pulmonary arterial hypertension (PAH).

Methods

Healthy volunteers (n=23) and patients with COPD (n=8), IPF (n=12), LHF (n=6) and PAH (n=10) underwent ¹²⁹Xe gas transfer imaging and dynamic spectroscopy. For each patient, three-dimensional maps were generated to depict ventilation, barrier uptake (¹²⁹Xe dissolved in interstitial tissue) and red blood cell (RBC) transfer (¹²⁹Xe dissolved in RBCs). Dynamic ¹²⁹Xe spectroscopy was used to quantify cardiogenic oscillations in the RBC signal amplitude and frequency shift.

Results

Compared with healthy volunteers, all patient groups exhibited decreased ventilation and RBC transfer (both p≤0.01). Patients with COPD demonstrated more ventilation and barrier defects compared with all other groups (both p≤0.02). In contrast, IPF patients demonstrated elevated barrier uptake compared with all other groups (p≤0.007), and increased RBC amplitude and shift oscillations compared with healthy volunteers (p=0.007 and p≤0.01, respectively). Patients with COPD and PAH both exhibited decreased RBC amplitude oscillations (p=0.02 and p=0.005, respectively) compared with healthy volunteers. LHF was distinguishable from PAH by enhanced RBC amplitude oscillations (p=0.01).

Conclusion

COPD, IPF, LHF and PAH each exhibit unique ¹²⁹Xe MRI and dynamic spectroscopy signatures. These metrics may help with diagnostic challenges in cardiopulmonary disease and increase understanding of regional lung function and haemodynamics at the alveolar–capillary level.

Introduction

Over the past 30 years, mortality related to both chronic obstructive pulmonary disease (COPD) and interstitial lung disease (ILD) has increased in the USA [1]. Among the factors associated with increased mortality in both COPD and ILD is the frequent presence of comorbid cardiovascular disease such as heart failure [2, 3]. Acute respiratory symptoms in patients with comorbid heart failure and lung disease can often have mixed cardiac and pulmonary origin, which complicates the interpretation of standard diagnostic testing [4]. For example, in patients with concomitant COPD and heart failure, interpretation of spirometry may be challenging [5] and echocardiographic acoustic windows may be impeded by gas trapping [6]. Similarly, pulmonary oedema can mimic certain types of ILD on computed tomography (CT) scans [7].

Magnetic resonance imaging (MRI) has become a valuable tool for the evaluation of pulmonary hypertension and heart failure, owing to its noninvasive nature, high reproducibility, and sensitivity to change in morphological, functional and flow-related parameters [8]. Multiparametric cardiac MRI models have high diagnostic accuracy in patients suspected of having pulmonary hypertension [9]. Dynamic contrast-enhanced lung perfusion MRI has high sensitivity compared with scintigraphy in screening for chronic thromboembolic pulmonary hypertension [10] and MRI has been used to predict mean pulmonary arterial pressure (mPAP) in patients with COPD [11]. However, these methods do not directly address ventilation and pulmonary gas exchange nor do they address impairment occurring within the microvasculature.

To address these challenges, MRI using hyperpolarised ¹²⁹Xe is emerging as a promising means for noninvasive, spatially resolved, assessment of pulmonary function [12]. This method, which is rapid and well tolerated, permits direct three-dimensional (3D) imaging, not only of ventilation distribution, but also pulmonary gas transfer. This is enabled by the free diffusion of xenon from the alveoli, through the interstitial barrier tissues (comprised of alveolar epithelial cells, interstitial tissues, capillary endothelial cells and plasma) and into red blood cells (RBCs) (figure 1); in each of these three compartments ¹²⁹Xe exhibits distinct frequency shifts [12]. Thus, with two breath-holds of ¹²⁹Xe, quantitative maps can now be generated to depict the distribution of ¹²⁹Xe in airspaces, its uptake in barrier tissues and its transfer to RBCs [13, 14]. This imaging approach is now capable of spatially resolving the heterogeneous disease burden for a wide array of conditions, including COPD, idiopathic pulmonary fibrosis (IPF) and pulmonary vascular disease [15, 16]. Furthermore, while ¹²⁹Xe imaging provides useful quantification of regional functional burden, a more detailed characterisation of whole-lung ¹²⁹Xe spectroscopic indices provides additional metrics that may help to further discriminate the underlying pathologies. For example, in IPF the spectroscopic ratio of RBC to barrier signal provides a global metric of gas transfer impairment that has been shown to strongly correlate with the diffusing capacity of the lung for carbon monoxide (D_LCO) [17]. Moreover, the signal frequency of ¹²⁹Xe interacting with capillary RBCs is uniquely sensitive to their level of oxygenation [18]. Such ¹²⁹Xe spectra can be acquired dynamically every 20 ms, revealing temporal oscillations in pulmonary gas exchange that reflect cardiopulmonary haemodynamics [19].

FIGURE 1.

a) Alveolar–capillary model and ¹²⁹Xe spectrum. ¹²⁹Xe freely diffuses from airspace to interstitial barrier tissue to red blood cells (RBCs). In these compartments, the ¹²⁹Xe atom exhibits distinct frequency shifts of 0, 198 and 217 ppm, respectively. b) Three-dimensional (3D) imaging and quantification. These properties are exploited to enable 3D imaging and quantification of ¹²⁹Xe distribution in airspace (ventilation), its barrier uptake and RBC transfer. These maps are colour coded to represent different signal intensity levels, with the central green bins representing voxels in the normal reference range. c) Dynamic spectroscopy and quantification. ¹²⁹Xe spectra are acquired dynamically every 20 ms, revealing cardiogenic oscillations of RBC amplitude (%) and frequency shift (ppm).

The array of noninvasive imaging and spectroscopic markers of pulmonary gas transfer and haemodynamics that can be derived from hyperpolarised ¹²⁹Xe presents an appealing approach for comprehensive and noninvasive phenotyping of cardiopulmonary physiology in individual patients. However, to date most studies have focused on only one form of ¹²⁹Xe contrast or investigated them in only a single disease. Here, we sought to apply a comprehensive panel of noninvasive ¹²⁹Xe MRI and spectroscopy to a cohort of patients with known heart and lung disease in order to identify features that could differentiate signatures of COPD, IPF, left heart failure (LHF) or pulmonary arterial hypertension (PAH).

Methods

Subject recruitment

The protocol was approved by the Institutional Review Board of Duke University Medical Center (Durham, NC, USA). Healthy volunteers and patients with a clinical diagnosis of COPD, IPF, LHF or PAH were recruited, and all provided written, informed consent. All healthy volunteers had no smoking history or known respiratory conditions. COPD was diagnosed using spirometric measures of post-bronchodilator forced expiratory volume in 1 s (FEV₁)/forced vital capacity (FVC) ≤70% predicted [20]. The diagnosis of IPF was established according to American Thoracic Society/European Respiratory Society criteria, either from a confirmed pattern of usual interstitial pneumonia on CT or from surgical lung biopsy [21]. LHF consisted of patients with both heart failure with preserved ejection fraction (n=4) or heart failure with reduced ejection fraction (n=2), which was confirmed by echocardiogram [22]. The patients with heart failure with preserved ejection fraction had a left ventricular ejection fraction ≥55% and either evidence of an elevated pulmonary capillary wedge pressure (PCWP) >15 mmHg at right heart catheterisation (n=2: PCWP 19 and 38 mmHg) and/or clinical evidence of LHF without echocardiographic or catheterisation evidence of pre-capillary pulmonary hypertension or right heart failure (n=4). PAH was defined according to the World Health Organization criteria and diagnosed by right heart catheterisation with a resting mPAP ≥25 mmHg and a PCWP ≤15 mmHg [23]. Haemodynamics of the PAH group at the time of diagnosis were (median (interquartile range)): mPAP 46 (39–59) mmHg, PCWP 12 (10–14) mmHg, pulmonary vascular resistance 6 (4–8) Wood Units and cardiac index 2.9 (2.4–3.5) L·min⁻¹·m⁻² . All clinical tests were performed as a part of routine care. Pulmonary function tests were performed on all patients and 83% of healthy volunteers to assess baseline pulmonary function.

¹²⁹Xe gas hyperpolarisation and MRI acquisition

¹²⁹Xe gas was hyperpolarised via continuous flow spin exchange optical pumping and cryogenic accumulation [24], using commercially available systems (Model 9820 and 9810; Polarean, Durham, NC, USA). ¹²⁹Xe spectroscopy and imaging were acquired on either a 1.5 T (15M4 Excite; GE, Chicago, IL, USA) or a 3 T (Magnetom Trio; Siemens, Erlangen, Germany) scanner during two separate breath-holds of ¹²⁹Xe. Subjects first underwent dynamic spectroscopy during which ¹²⁹Xe free induction decays (FIDs) were collected every 20 ms (echo time 0.932 ms, flip angle □20°, dwell time 32 μs, 512/1024 points) during a breath-hold [19]. 3D images were then acquired using an interleaved radial acquisition of gas- and dissolved-phase signal during a 15 s breath-hold. The signal was acquired at an echo time that allowed the two dissolved-phase compartments to be decomposed using the 1-point Dixon method [12]. This generated 3D images of the gas, barrier and RBC components with 2.8 mm isotropic voxels.

Quantitative processing and analysis

3D images of each compartment were rendered into quantitative maps and cast into colour clusters using thresholds derived from healthy reference cohorts established for both 1.5 and 3 T acquisitions [13, 25]. Maps of ¹²⁹Xe in the barrier tissues and RBCs were normalised on a voxel-by-voxel basis by dividing by the local ventilation signal. The resulting binning maps depict ¹²⁹Xe ventilation, barrier tissue uptake (¹²⁹Xe dissolved in interstitial tissue) and RBC transfer (¹²⁹Xe dissolved in RBCs). Each of these maps was quantified by calculating the percentage of the lung exhibiting signal defects and high signal [13]. The dynamically acquired FIDs were fit in the time domain to determine the gas, barrier and RBC spectral parameters [19]. The time-dependent cardiogenic oscillations in the RBC amplitude and frequency shift were quantified by their peak-to-peak value relative to the mean [19]. Imaging and spectroscopic findings were compared across all cohorts.

Statistical methods

Imaging and spectroscopic features were compared between cohorts. All computations were performed using JMP version 14 (SAS Institute, Cary, NC, USA). First, ANOVA was performed using the nonparametric Kruskal–Wallis test. When a significant difference was detected, the Mann–Whitney U-test was further used for pair-wise analysis. Statistical significance was concluded when p<0.05.

Results

Study cohort

This study included 23 healthy volunteers, 8 patients with COPD, 12 patients with IPF, 6 patients with LHF and 10 patients with PAH. Subject demographics and pulmonary function test results are summarised in table 1. 3D isotropic images of ¹²⁹Xe in the gas, barrier and RBC compartments were acquired on 19 healthy volunteers and all patients. Dynamic spectroscopy was acquired on 13 healthy volunteers, 6 patients with COPD, 8 patients with IPF, 5 patients with LHF and 10 patients with PAH.

TABLE 1.

Demographic and clinical characteristics stratified by condition

	Healthy volunteers#	COPD#	IPF#	LHF#	PAH#,¶,+
Subjects	23	8	12	6	10
Age years	26 (22–32)	61.5 (57.0–71.8)	68 (63.3–71.5)	63.5 (58.0–69.0)	50 (46.5–54.0)
Female	5 (22)	3 (38)	2 (17)	2 (33)	5 (50)
Non-White race	7 (30)	0 (0)	0 (0)	1 (17)	2 (20)
BMI kg·m−2	24.0 (23.3–27.4)	24.3 (22.5–26.5)	28.8 (24.7–33.0)	29.4 (25.1–32.3)	29.6 (24.9,33.9)
Current or prior tobacco use	0 (0)	6 (75)	8 (67)	4 (67)	3 (30)
Supplemental oxygen at rest	0 (0)	5 (63)	4 (33)	1 (17)	2 (20)
6MWD m		395 (324–429), n=4	461 (369–515), n=10	451 (374–528), n=2	500 (426–575)
PFTs§
FEV1 % pred	92 (78–102), n=19	34 (27–55)	70 (57–80)	72 (58–85), n=4	83 (63–94), n=9
FVC % pred	96 (83–103), n=19	79 (63–91)	63 (50–76)	75 (57–95), n=4	96 (68–99), n=9
FEV1/FVC % pred	81 (78–88), n=19	49 (40–54), n=7	84 (80–89)	84 (81–89), n=3	82 (74–87), n=4
TLC % pred	102 (92–113), n=18	100 (93–107), n=2	48 (43–63), n=3	94 (88–100), n=2	95 (86–107), n=9
DLCO mL·min−1·mmHg−1	28 (22–31), n=17	11 (9–15)	11.5 (9–14)	21 (15–28)	16 (13–21)
Kco mL·min−1·mmHg−1·L−1	5.2 (4.7–5.8), n=17	2.7 (1.7–4.0), n=8	3.3 (2.9–3.9)	4.7 (2.7–4.8)	3.9 (3.4–5.2)

Data are presented as n, median (interquartile range) or n (%). COPD: chronic obstructive pulmonary disease; IPF: idiopathic pulmonary fibrosis; LHF: left heart failure; PAH: pulmonary arterial hypertension; BMI: body mass index; 6MWD: 6-min walk distance; PFT: pulmonary function test; FEV₁ : forced expiratory volume in 1 s; FVC: forced vital capacity; TLC: total lung capacity; D_LCO diffusing capacity of the lung for carbon monoxide; K_CO : transfer coefficient of the lung for carbon : monoxide; MRI: magnetic resonance imaging.

^#:13 healthy subjects were scanned on a 1.5 T MRI system while 10 were scanned at 3 T; all IPF patients were scanned at 1.5 T, and all COPD, LHF and PAH patients were scanned at 3 T;

^¶:PAH patients were on therapy with a combination of prostacyclin analogues, phosphodiesterase-5 inhibitors or endothelin receptor antagonists;

^+:the PAH cohort consisted of seven patients with idiopathic PAH, one patient associated with HIV, one patients associated with drugs/toxins and one patients associated with connective tissue disease;

^§:most recent available clinical values presented (PFTs were performed within 6 months of MRI for all subjects; right heart catheterisations were historical and performed at the tim of initial diagnosis).

Identifying disease-specific imaging-derived metrics

Representative ventilation and gas transfer maps from subjects in each group are depicted in figure 2. For each map, the percentages of voxels falling in the defect, low and high bins are reported. In the healthy volunteer, the majority of the ¹²⁹Xe signal in all three compartments fell within ±1SD from the mean of the reference distribution and thus in the “normal” green colour bins. By contrast, the COPD subject exhibited significant defects in all three compartments (ventilation, barrier and RBC), indicated by the red colour bins. The IPF subject exhibited relatively normal ventilation, but significant areas of high barrier uptake (purple), accompanied by defects in RBC transfer in the lower lobes. Both the LHF and PAH subjects exhibited slight ventilation defects, relatively normal barrier, but more significant deficits in RBC transfer.

FIGURE 2.

Ventilation, normalised barrier uptake and RBC transfer maps of representative subjects from each cohort. COPD: chronic obstructive pulmonary disease; IPF: idiopathic pulmonary fibrosis; LHF: left heart failure; PAH: pulmonary arterial hypertension. The colour bins represent signal intensity, with red for the lowest, blue/purple for the highest and green representing voxels in the healthy reference range. Each map is quantified by the percentage of defect (D), low (L) and high (H), calculated as the voxel fraction of the lowest, second lowest and the highest two bins for each map, respectively. The voxels with ventilation defect were excluded from the analysis of barrier uptake and RBC transfer maps.

Figure 3 evaluates these imaging features quantitatively across the cohorts, comparing the percentages of ventilation defects, RBC defects, barrier defects and high barrier. Compared with healthy subjects, all patient groups exhibited a larger percentage of defects in ventilation (p≤0.01 for all comparisons) and RBC transfer (p≤0.01 for all comparisons). The COPD cohort stood out for exhibiting the largest percentages of ventilation defects (41.5±22.6%; p≤0.02 for all comparisons) and was the only one to show defects in barrier uptake (10.4±7.1%; p≤0.02 for all comparisons). By contrast, IPF patients were distinguished from the other groups by the largest percentage of voxels with high ¹²⁹Xe uptake in the barrier tissue (39.8%; p≤0.007 for all comparisons). IPF subjects exhibited only modest ventilation defects (11.5±6.7%; p=0.0003 versus healthy) but substantial RBC defects (11.3±6.7%; p=0.0001 versus healthy). LHF and PAH patients presented with similar imaging characteristics with mildly elevated ventilation defects (LHF: 11.7±6.2%; p=0.01 versus healthy and PAH: 8.4±4.7%; p=0.01 versus healthy) and increased RBC transfer defects (LHF: 13.3±10.2%; p=0.01 versus healthy and PAH: 14.5±9.3%; p=0.002 versus healthy).

FIGURE 3.

a) Ventilation defect, b) red blood cell (RBC) defect, c) barrier defect and d) barrier high percentage comparisons for all cohorts. COPD: chronic obstructive pulmonary disease; IPF: idiopathic pulmonary fibrosis; LHF: left heart failure; PAH: pulmonary arterial hypertension. Asterisks indicate a significantly increased (red) or decreased (green) value compared with all other cohorts. Compared with healthy subjects, all disease cohorts showed increased ventilation defect (p≤0.01) and RBC defect (p≤0.01). COPD was characterised by significantly elevated percentages of ventilation defects (p≤0.02) and barrier defects (p≤0.02). IPF uniquely exhibited reduced barrier defects (p≤0.02), but elevated percentages of high barrier (p≤0.007).

PAH and LHF exhibited slightly elevated ventilation defects and modestly elevated RBC defects.

Disease-specific spectroscopy-derived metrics

Figure 4 shows the RBC signal amplitude and shift oscillations for representative subjects from each group. Notably, the RBC signal amplitudes for each patient oscillate at a frequency identical to the heart rate. The IPF patient also exhibits such cardiogenic oscillations prominently in the RBC frequency shift.

FIGURE 4.

Red blood cell (RBC) signal amplitude and frequency shift demonstrate cardiogenic oscillations in representative patients. IPF: idiopathic pulmonary fibrosis; LHF: left heart failure; PAH: pulmonary arterial hypertension; COPD: chronic obstructive pulmonary disease. The IPF and LHF patients exhibit enhanced RBC amplitude oscillations. By contrast, RBC signal oscillations are diminished in the PAH and COPD patients. Only the IPF patient exhibits oscillations in the RBC shift.

Figure 5 shows the group-wise comparison of the cardiogenic RBC amplitude and shift metrics. In healthy subjects, the RBC amplitudes oscillated at a height of 10.0±2.6% peak-to-peak with very little RBC shift oscillation (0.07±0.05 ppm). The RBC shift only oscillates significantly in the IPF cohort (0.46±0.33 ppm; p≤0.01 for all comparisons). IPF patients also exhibited larger RBC amplitude oscillations (16.7±5.5%; p=0.007) compared with healthy volunteers. RBC amplitude oscillations were diminished in both COPD and PAH compared with healthy volunteers (COPD: 5.5±4.7%; p=0.02 and PAH: 6.0±3.6%; p=0.005). In subjects with LHF, the RBC amplitude oscillations were larger than in healthy volunteers, but this did not reach statistical significance (13.0±5.1%; p=0.2). However, these oscillations were significantly higher compared with subjects with PAH (p=0.01).

FIGURE 5.

Red blood cell (RBC) a) amplitude and b) frequency shift oscillations compared across cohorts. COPD: chronic obstructive pulmonary disease; IPF: idiopathic pulmonary fibrosis; LHF: left heart failure; PAH: pulmonary arterial hypertension. Black asterisks indicate a significant difference between cohorts and the red asterisk indicates an increased value compared with all other cohorts. Compared with healthy subjects, COPD (p=0.02) and PAH (p=0.005) exhibited decreased RBC amplitude oscillations, while they were increased in IPF (p=0.007). Moreover, in LHF the RBC amplitude oscillations were significantly increased compared with PAH (p=0.01). IPF patients exhibited significantly increased RBC shift oscillations compared with all other cohorts (p≤0.01).

Discussion

In this study, we identified unique ¹²⁹Xe MRI and spectroscopy signatures for patients with COPD, IPF, PAH and LHF. COPD was characterised by significantly elevated ventilation and barrier defect percentages compared with all other disorders, as well as diminished RBC amplitude oscillations. However, in the COPD cohort the ventilation defect percentage varied widely in both scale and distribution, consistent with the heterogeneity of the disease [26]. In contrast, IPF was characterised primarily by elevated barrier uptake, virtually absent barrier defect percentage, elevated RBC amplitude oscillations and prominent oscillations in RBC shift. PAH and LHF presented with similar imaging characteristics (slight elevations in ventilation, barrier and RBC defect percentages compared with healthy volunteers). However, PAH was distinguished from LHF by RBC amplitude oscillations that were lower than in healthy subjects, whereas in LHF such oscillations were enhanced. All four disease cohorts exhibited increased RBC transfer defects compared with healthy subjects.

The observation that ventilation defects are prominent in patients with COPD is consistent with numerous previous studies [14, 27, 28]. We now add the observation that, in COPD, barrier uptake is also diminished, likely reflecting emphysematous lung destruction and loss of surface area for gas exchange. This loss would further lead to diminished RBC transfer. In IPF we confirm that the disease is characterised by increased barrier uptake with defects in RBC transfer primarily in the lung bases [14]. Furthermore, we provide important context for our prior work showing that cardiogenic oscillations in ¹²⁹Xe RBC amplitude and shift are significantly enhanced in patients with IPF relative to healthy controls [19]. Having now acquired such data in this broader cohort, this suggests that the RBC shift oscillations are, thus far, unique to IPF, and are not observed in COPD, LHF and PAH. Moreover, the enhanced RBC amplitude oscillations seen in IPF are only additionally seen in LHF, suggesting that perhaps this is a marker of post-capillary pulmonary hypertension.

Connecting ¹²⁹Xe MRI to conventional metrics of gas exchange

The ability of ¹²⁹Xe MRI to produce 3D quantitative maps of ventilation, barrier uptake and RBC transfer invites an effort to relate these imaging features to diffusing capacity (D_LCO) and its underlying rate constant, the transfer coefficient of the lung for carbon monoxide K_CO (D_LCO divided by accessible alveolar volume V_A). This appeals to the framework of Roughton and Forster [29], which treats the pulmonary diffusing capacity D_L as being comprised of two serial conductances:

$$\frac{1}{D_{\text{L}}}=\frac{1}{D_{\text{M}}}+\frac{1}{{\text{θ}\mathit{V}}_{\text{c}}}$$

where D_M is the membrane conductance and θV_c is attributable to the capillary blood volume V_c and its reaction rate θ with CO. The membrane term D_M is directly proportional to the alveolar surface area available for gas exchange, but inversely proportional to barrier thickness. Both these aspects are inherently seen in the ¹²⁹Xe barrier/gas signal distribution. When mean barrier/gas (Bar) decreases, it indicates a loss of surface area, and when it increases, it indicates barrier thickening. This suggests that a patient’s Bar compared with a healthy reference value, $\widetilde{Bar}=Bar/Ba{r}_{\text{ref}}$ (or its inverse when Bar>Bar_ref), can serve as a surrogate for D_M. $\widetilde{Bar}$, constructed in this fashion, is akin to a relative rate constant, and we thus multiply by V_A and a proportionality constant α to arrive at the membrane conductance $D_{\text{M}}=\alpha \widetilde{Bar}{V}_{\text{A}}$. In a similar manner, mean RBC/gas signal relative to reference reflects the capillary blood volume conductance according to ${\text{θ}\mathit{V}}_{\text{c}}=\beta \widetilde{RBC}{\text{V}}_{\mathrm{A}}$. Thus, with knowledge of α and β the overall diffusing capacity can be calculated from these xenon MRI metrics as:

$$\frac{{\text{V}}_{\mathrm{A}}}{D_{\text{LCO}}}=\frac{\text{1}}{K_{\text{CO}}}=\frac{\text{1}}{\alpha \widetilde{Bar}}+\frac{\text{1}}{\text{β}\widetilde{RBC}}$$

Notably, the left-hand side of the equation is simply the inverse of the transfer coefficient K_CO. Combined with alveolar volume V_A, it is also possible to estimate D_LCO. Using the ¹²⁹Xe MRI metrics from this full patient cohort, the relevant reference population values [13, 30] and K_CO data, we found α=11.0 and β=12.8 mL·min⁻¹·mmHg⁻¹ L⁻¹ provided a best fit. This is represented in figure 6a, showing a 3D surface that reflects the way in which relative barrier and RBC contribute to K_CO. Similarly, the correlations of the image-estimated K_CO and D_LCO values with the measured values are shown in figure 6b and c. Within this framework we see that imaging predictions of K_CO are relatively strong and when multiplied by V_A generate an image-predicted D_LCO that is remarkably well correlated with the measured value. This approach, which will need to be confirmed prospectively, provides a strong foundation for using ¹²⁹Xe MRI to interpret D_LCO across a range of patient populations with all combinations of low and high barrier and as well as capillary blood volumes.

FIGURE 6.

a) Three-dimensional surface plot reflects the relative contribution of barrier and red blood cells (RBCs) to the transfer coefficient of the lung for carbon monoxide (K_CO). COPD: chronic obstructive pulmonary disease; IPF: idiopathic pulmonary fibrosis; LHF: left heart failure; PAH: pulmonary arterial hypertension. b) The image-estimated K_CO shows strong correlation with the measured K_CO. c) This estimated K_CO, when combined with measured alveolar volume (V_A), produces an estimated diffusing capacity of the lung for carbon monoxide (D_CO) that shows an even stronger correlation with the measured D_LCO values

Alveolar–capillary interface models depicting disease phenotypes

To aid in interpreting the patterns of ¹²⁹Xe MRI and spectroscopic signatures of each disease in the context of gas transfer physiology, we propose the following conceptual alveolar–capillary interface architecture (figure 7). In a healthy subject, ¹²⁹Xe atoms freely diffuse into the alveoli and into the alveolar–capillary interface, translating into images reflecting a normal range of ventilation, barrier uptake and RBC transfer. In COPD, chronic airway inflammation and small airway obstruction [31] create ventilation defects, while the loss of alveolar surface area associated with emphysema results in diminished uptake of ¹²⁹Xe in the interstitial barrier tissues. This drives a concomitant decrease in RBC transfer, consistent with the low D_LCO values observed in this cohort. However, we note that many patients exhibit disproportionately worse RBC transfer that may reflect an additional loss of vasculature [32]. By contrast, in IPF, interstitial fibrosis creates a larger reservoir for ¹²⁹Xe uptake, thus enhancing the barrier signal [33], and in conjunction with regions of RBC transfer defects, reduces D_LCO [15]. It is less clear what causes the defects in RBC transfer in LHF, but it is known that these patients can develop gas exchange abnormalities including a reduction in D_LCO that is thought to be secondary to chronic damage resulting from pulmonary venous congestion [34, 35]. Finally, PAH is characterised by increased pre-capillary impedance, resulting from remodelling and obliteration of the pulmonary arterioles. This results in a loss of alveolar membrane diffusing capacity and pulmonary capillary blood volume as reflected in figure 7.

FIGURE 7.

Conceptual architectures of the alveolar–capillary interface to aid in interpreting the ¹²⁹Xe imaging and spectroscopic biomarkers across disease state: a) healthy volunteers, b) chronic obstructive pulmonary disease (COPD), c) idiopathic pulmonary fibrosis (IPF), d) left heart failure (LHF) and e) pulmonary arterial hypertension (PAH). RBC: red blood cell. The diagrams illustrate the alveoli, capillary blood vessel, interstitial barrier tissues, RBCs and ¹²⁹Xe atoms (see figure 1). For each disease state the anticipated effect on the ¹²⁹Xe biomarker is shown.

RBC amplitude oscillations reflect changes in ¹²⁹Xe transfer driven by differences in capillary blood volume during the cardiac cycle. These, in turn, depend on the right ventricular stroke volume, driving flow though the pre- and post-capillary impedances. In IPF, for example, the stroke volume is delivered to a partially destroyed capillary bed [15]. If stroke volume is preserved, it will produce larger relative capillary blood volume oscillations between systole and diastole, thereby provoking larger RBC amplitude oscillations. Such enhancement is also seen in pulmonary venous hypertension caused by left-sided heart failure. In these patients, enhanced cardiogenic RBC oscillations are consistent with diastolic reserve limitation during the cardiac cycle [36], caused by a high post-capillary impedance. This results in a relative increase in pulmonary capillary blood volume, as right ventricular output transiently exceeds left ventricular output that leads to blood pooling in the pulmonary circulation [34]. In PAH, while capillary blood volume is also reduced [37], the larger impedance to flow in the arterioles causes RBC oscillations to be reduced. It is these amplitude oscillations that appear to be the feature that most strongly differentiates pre-capillary from post-capillary PH. It is possible that this could be even more clearly demonstrated by correcting the RBC oscillation amplitude for patient-specific stroke volume and capillary bed volume. Thus, future studies could benefit from the addition of cardiac MRI to determine stroke volume and allow for such refinement.

Unique to the IPF patients in this cohort is the observation that the RBC resonant frequency also exhibits cardiogenic oscillations. As the ¹²⁹Xe RBC frequency is strongly dependent on blood oxygenation level [18], we hypothesise that this indicates variability in the blood oxygenation. In IPF subjects, diffusion impairment likely causes large excursions in blood oxygenation as the capillary blood volume is replaced over the course of the cardiac cycle. This likely explains why the RBC frequency shift oscillates cardiogenically in conditions where blood oxygenation is diffusively impaired [19]. Such oscillations are not observed in COPD where gas exchange impairment results from loss of membrane surface area rather than membrane conductance. We should note that all the aforementioned interpretations and proposed models will require confirmation in both in vitro systems and in vivo animal models where each condition can be carefully controlled.

Potential for differentiating cardiopulmonary diseases in the clinical setting

Taken together, this combination of noninvasive ¹²⁹Xe MRI and spectroscopic parameters enables interrogation of gas transfer at the alveolar–capillary level that appears useful, not only to characterise disease burden, but also to identify signatures that may ultimately help differentiate cardiopulmonary disorders. A potential approach to such differentiation is seen in figure 8, which shows radar plots of the four key imaging features and two key spectroscopic features: ventilation defect, barrier defect, high barrier uptake, RBC defects, and RBC amplitude and shift oscillations. Integrating these features for each disease group provides an initial means of displaying these phenotypes in a visually distinct way. Generating such plots for individual patients could provide a powerful means to identify the primary phenotypes that should be considered.

FIGURE 8.

Radar plots to display the primary ¹²⁹Xe magnetic resonance imaging and spectroscopic signatures associated with a) healthy volunteers, b) chronic obstructive pulmonary disease (COPD), c) idiopathic pulmonary fibrosis (IPF), d) left heart failure (LHF) and e) pulmonary arterial hypertension (PAH). RBC: red blood cell. Here 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 spectroscopy.

While our study benefited from using well-characterised patient cohorts, work in patients with mixed cardiopulmonary disease may demonstrate the utility of ¹²⁹Xe MRI to determine the underlying cause of dyspnoea in patients with concomitant diseases. Furthermore, as early diagnosis is increasingly emphasised in disorders such as ILD [38] and PAH [39], ¹²⁹Xe spectroscopic indices may provide a sensitive probe for this as well as allowing for noninvasive tracking of disease progression. Importantly, the RBC transfer signal depicts the ultimate disease burden for gas transfer function, and therefore might be used to evaluate disease progression and therapeutic response [40]. Given the limitations of current diagnostic testing, the information provided by ¹²⁹Xe gas transfer imaging and dynamic spectroscopy has the potential to improve patient care.

Study limitations and future directions

We must acknowledge several limitations to this first study comparing ¹²⁹Xe MRI and spectroscopic signatures across cardiopulmonary conditions. First, the heterogeneity and possible comorbidities of patients in each disease cohort may have limited our ability to identify patterns in ¹²⁹Xe imaging and spectroscopy. For example, all PAH patients were undergoing clinically indicated, targeted treatment, which may have limited the severity of their PAH at the time of the ¹²⁹Xe study. Because many did not have a recent right heart catheterisation available, the severity of their disease could not be assessed at the time of their study. Furthermore, while our study aimed to recruit patients with isolated LHF as a model for post-capillary impedance, several may have also had right heart failure given the common pathogenic evolution from left heart dysfunction to right heart dysfunction over time [41]. In fact, this phenotypic evolution may partly explain the large variation in RBC amplitude oscillation exhibited by our LHF cohort (maximum 21.5%, minimum 8.0%, _SD 5.1%). Additionally, IPF and COPD subjects were not specifically evaluated to rule out pulmonary hypertension. Another limitation is that our subject scans were conducted on different platforms with two field strengths [13, 30]. Our quantification method, using a healthy reference group constructed under the same acquisition protocol, was designed to incorporate the potential factors such as T₁ and T₂* decay, which may affect the gas transfer measurements. However, these and other factors constrained the size and sex ratio of the healthy reference cohorts, which were also significantly younger than the typical patients in our cohorts and predominantly male. Since the aging lung is reported to undergo physiological changes that could impact gas transfer functions [42], future studies will benefit from constructing a larger and age-controlled healthy population.

Conclusions

In this study, we applied ¹²⁹Xe gas transfer imaging and spectroscopy on healthy subjects and patients with COPD, IPF, LHF and PAH. As a noninvasive and nonionising tool, hyperpolarised ¹²⁹Xe gas transfer MRI provides a fundamentally new approach to directly image regional function while also capturing haemodynamics at the alveolar–capillary level. We identified unique imaging and spectroscopic signatures for each of these diseases that may help overcome some of the diagnostic challenges faced by clinicians treating patients with cardiopulmonary disease. ¹²⁹Xe gas transfer imaging and spectroscopy is a promising technology in characterising cardiopulmonary disease pathophysiology and, with further validation in larger studies, can contribute to a comprehensive understanding of the multifactorial pathogenesis of dyspnoea and developing personalised treatment approaches.