Correlating Pulmonary Hemodynamics With HP ¹²⁹Xe Dissolved‐Phase MRSI in a Controlled Porcine Intervention Study

ABSTRACT

Purpose

Pulmonary hypertension is related to a 1‐year mortality ranging from < 5% to > 20% in low‐ and high‐risk groups. Right heart catheterization is the diagnostic gold standard, associated with a 1.1% rate of severe adverse events, posing the need for a non‐invasive diagnostic tool. This study aims to delineate the effects of controlled changes on pulmonary hemodynamics by quantifying corresponding changes in hyperpolarized ¹²⁹Xe MRSI dissolved‐phase parameters, including RBC oscillations.

Methods

Seven Danish Landrace pigs underwent hyperpolarized ¹²⁹Xe dissolved‐phase 3D MRSI with interleaved MRS under baseline conditions and three vasoactive interventions, monitored with invasive catheters or phase‐contrast MRI. Results were evaluated using one‐way ANOVA or mixed‐effects analysis followed by Dunnett's multiple comparisons test.

Results

Hypoxia increased mPAP (p = 0.038), reduced RBC:M (−31%, p = 0.007) and RBC:G (−34%, p = 0.009), while increasing RBC chemical shift (0.57%, p = 0.020). Adenosine and dobutamine increased RBC:M (22%, p = 0.003) and RBC:G (30%, p = 0.009), while decreasing RBC chemical shift (−0.38%, p = 0.11). Linear regression revealed an inverse correlation between mPAP and RBC oscillation amplitude (R ² = 0.158, p = 0.04), RBC:M (R ² = 0.25, p = 0.05), and RBC:G (R ² = 0.22, p = 0.01).

Conclusion

This study demonstrates that dissolved‐phase hyperpolarized ¹²⁹Xe MRSI parameters are sensitive to hemodynamic changes, exhibiting consistent inverse relationships with invasively measured mPAP. By leveraging concurrent invasive reference measurements, these findings provide physiological validation that dissolved‐phase ¹²⁹Xe metrics reflect pulmonary vascular pressure, supporting their further development as a non‐invasive adjunct to right heart catheterization.

1. Introduction

Pulmonary hypertension (PH) is defined by the European Society of Cardiology and European Respiratory Society guidelines as a mean pulmonary arterial pressure (mPAP) > 20 mmHg [1]. It is a frequent comorbidity in cardiopulmonary disease, affecting 65%–80% of patients with left heart disease, and carries a 1‐year mortality of < 5% in low‐risk groups, and reaching > 20% in high‐risk groups [1]. The current golden standard for diagnosing PH is right heart catheterization (RHC), associated with a 1.1% rate of severe adverse events, and 0.055% rate of procedure‐related mortality [1]. Because PH can only be confirmed with RHC, diagnosis is often delayed; patients typically present more than 2 years after symptom onset, resulting in patients presenting with advanced disease [1]. Suspicion is based on symptoms like dyspnea, findings such as dilation of the right atrium, signs of left heart disease, and elevated atrial biomarkers [1]. The invasive nature of RHC makes it unfit for fragile patients, leaving a gap in diagnostic tools for precise assessment of cardiopulmonary hemodynamics [1, 2, 3].

Dissolved‐phase hyperpolarized (HP) ¹²⁹Xe magnetic resonance spectroscopy imaging (MRSI) has emerged as a promising non‐invasive approach for probing pulmonary vasculature and cardiopulmonary physiology [4, 5, 6, 7, 8, 9, 10]. Following alveolar diffusion, xenon dissolves into the pulmonary parenchyma and bloodstream, where it binds to hemoglobin. Dissolved‐phase ¹²⁹Xe signals therefore reflect gas transfer across the alveolar‐capillary barrier and serve as indirect markers of pulmonary perfusion, capillary blood volume, and red blood cell (RBC) oxygenation [11, 12, 13, 14, 15, 16, 17, 18, 19]. Prior studies have reported associations between dissolved‐phase ¹²⁹Xe metrics and invasive hemodynamic parameters, including mPAP, cardiac output (CO), and capillary blood volume [5, 20, 21, 22]. In addition, RBC signal is sensitive to changes in capillary blood volume expressed as cardiogenic RBC oscillations aligning with the heart rate (HR) [15] and its amplitude further influenced by blood flow along with systolic and diastolic blood pressure [15, 22]. RBC oscillation is influenced by cardiopulmonary pathology: post‐capillary PH typically increases RBC oscillation amplitude, whereas pre‐capillary PH reduces it [21].

However, human studies linking dissolved‐phase ¹²⁹Xe metrics to invasive hemodynamics have largely relied on heterogeneous case–control cohorts, often differing in age, burden of comorbidity, disease etiology, invasive hemodynamic measurements and timing of reference measurements [4, 20, 21]. These factors complicate interpretation and limit the ability to establish robust, causal relationships between ¹²⁹Xe‐derived signals and pulmonary hemodynamics.

As a result, it remains unclear to what extent dissolved‐phase HP ¹²⁹Xe measurements can reliably track controlled changes in mPAP and related cardiopulmonary parameters under well‐defined physiological conditions. Addressing this gap is essential for evaluating the potential of HP ¹²⁹Xe MRSI as a quantitative, non‐invasive tool for PH assessment.

In this study, we investigate how 3D dissolved‐phase HP ¹²⁹Xe MRSI reflects controlled experimentally modulated cardiopulmonary hemodynamics using a porcine model. By acquiring simultaneous invasive reference measurements, we assess the ability of dissolved‐phase ¹²⁹Xe metrics to track mPAP and related hemodynamic changes under reproducible and physiologically controlled conditions.

2. Methods

2.1. Study Design

2.1.1. Animal Setup

The study was approved by the Danish Animal Research Inspectorate 2023‐15‐0201‐01553. Experiments were conducted on seven healthy female Danish Landrace pigs, weighing approximately 40 kg, equivalent to a child. Intravenous access was established for anesthesia, and intubation was done for mechanical ventilation. Anesthesia was secured by infusing fentanyl (0.009 mg/kg/h), propofol (2.29 mg/kg/h), and midazolam (up to 1.14 mg/kg/h).

During preparations and catheterization, the pigs were ventilated with a fraction of inhaled oxygen (F_iO₂) of 40%. During interventions, the pigs were ventilated with F_iO₂ 21% to avoid unwanted vasodilation [23]. The ventilator was set to Pressure Controlled Ventilation—Volume Guaranteed (PCV‐VG) mode, with set‐off values: positive end‐expiratory pressure (PEEP) of 4.0 cmH₂O, tidal volume (TV) of 450 mL and respiration rate (RR) of 14 per minute. TV and RR were adjusted to maintain stable end tidal CO₂ (etCO₂) and sO₂ for each pig during baseline and interventions. Mean arterial pressure (MAP) was measured through a catheter in the femoral artery. mPAP and pulmonary arterial occlusion pressure (PAOP) were measured in the pulmonary artery using fluid‐filled standard equipment (Swan‐Ganz catheter 111F7, Edwards Lifesciences). For catheter overview, see Figure 1A–D. A cardiopulmonary hemodynamic assessment including arterial blood gas, phase‐contrast scan and invasive measurements was done momentarily before each scan. Pulmonary vascular resistance (PVR) was calculated as $\mathit{PVR}=\frac{\text{mPAP}-\text{PAOP}}{\mathrm{CO}}$. mPAP measurements were recorded seconds before, during, and after each scan.

FIGURE 1.

Animal setup and the effects of vasodilation and ‐constriction on RBC oscillations. (A, B) arterial and venous femoral catheterization. (C, D) Swan Ganz catheters are placed in the atrium and pulmonary artery. (E, F) Increased 129Xe uptake and RBC oscillation amplitude during vasodilation (G, H) Reduced 129Xe uptake and RBC oscillation amplitude during vasoconstriction.

2.1.2. Vasoactive Interventions

To mimic hemodynamics commonly found in diseases such as chronic obstructive pulmonary disease and heart failure, three interventions were included in all pigs after the initial baseline scan: Hypoxia, adenosine with dobutamine, or adenosine diphosphate (ADP). For an overview of expected cardiopulmonary hemodynamic effects of each intervention, see Table 1. The order of interventions was randomized by drawing lots and can be seen in Table S1.

TABLE 1.

Expected cardiopulmonary hemodynamic response to vasoactive interventions.

	Hypoxia	Adenosine and dobutamine	ADP
Physiological response	Hypoxic vasoconstriction	Vasodilation and cardiac stress	Normoxic vasoconstriction
CO	$\uparrow$	$\uparrow \uparrow$	—
mPAP	$\uparrow \uparrow$	$\downarrow$	$\uparrow \uparrow$
PVR	$\uparrow \uparrow$	$\downarrow$	$\uparrow \uparrow$
MAP	$\downarrow$	$\downarrow$	$\downarrow$
PAOP	$\uparrow$	$\downarrow$	$\uparrow$
sO2	$\downarrow \downarrow$	—	—

Abbreviations: ADP, adenosine diphosphate; CO, cardiac output; MAP, mean arterial pressure; MPAP, mean pulmonary arterial pressure; PAOP, pulmonary arterial occlusion pressure; PVR, pulmonary vascular resistance; sO2, oxygen saturation.

Hypoxia induces physiological vasoconstriction in the pulmonary vasculature, expected to increase PVR and mPAP [24]. To induce hypoxia, the pigs were ventilated with F_iO₂ 15% for 15 min.

Adenosine monophosphate (a vasodilator [25]) and dobutamine (positive inotrope with vasodilatory effects [26, 27]) were administered to increase cardiac contractility and heart rate, and to induce pulmonary vasodilation. This intervention was expected to increase CO and reduce PVR and mPAP. Adenosine monophosphate (0.083 mg/kg/h) and dobutamine (5.25 μg/kg/h) were infused into the femoral vein for 10 min.

ADP causes pulmonary vasoconstriction, expected to increase mPAP and PVR [28, 29]. ADP (0.05 mg/kg/min) was infused into the right atrium for 4 to 7 min.

All interventions induce a response in the capillaries, along with the pre‐ and post‐capillary bed. Figure 1 shows catheter placement (A–D), and a visualization of the expected RBC oscillation amplitude during vasoconstriction and ‐dilation (E–H). Figure 2 shows a flowchart of the experiment.

FIGURE 2.

Experiment protocol. The pigs undergo an HP 129‐Xe MRSI with interleaved unlocalized MRS at baseline, followed by three interventions in a random order. During the experiments, hemodynamic measurements and blood samples are drawn.

2.1.3. Arterial Blood Gas Measures

Arterial blood was drawn from the femoral artery and analyzed within 30 min of sampling to monitor acid–base homeostasis and metabolic status of the pigs between each intervention. A clinical blood gas analyzer (Radiometer, ABL80, UK) was used to analyze samples. Values of O₂ saturation (sO₂), O₂ partial pressure (PaO2), CO₂ partial pressure (PaCO₂), hemoglobin (HgB), hematocrit (HcT), and pH were measured.

2.2. Image Acquisition

2.2.1. Xenon Preparation and Procedures

Polarization of ¹²⁹Xe was done by a spin exchange optical pumping (SEOP) polarizer (POLARIS, University of Sheffield, UK) [30]. Hyperpolarized gas was dispensed as a fixed 800 mL dose (450 mL xenon, 350 mL nitrogen) in a Tedlar bag, chosen to provide a controlled inspiratory maneuver approximately 1.7–2.0 times expected tidal volume of the pigs, administered after a 1–2 min walk to the MRI scanner. Prior to xenon gas administration, the ventilation tube was unplugged for 4 s to allow the lungs to deflate and reach functional residual capacity. Gas was infused by attaching the Tedlar bag to the ventilation tube and applying gentle pressure to it, forcing a steady gas flow into the lungs over approximately 4 s.

2.2.2. Xenon Imaging

Xenon images were acquired on a 3 T MRI scanner (MR750, GE Healthcare, Waukesha, WI, USA) using a ¹²⁹Xe transmit‐receive quadrature vest coil (Clinical MR Solutions, Brookfield, WI, USA) tuned to 35.3 MHz. As anatomical reference, coronal 3D T₁‐weighted (T₁w) proton images were acquired using a 3D SSFP imaging sequence with FOV = 40 × 40 × 20 cm³, matrix = 256 × 256 × 128, FA = 3°, TR = 3.4 ms, and TE = 2.1 ms on the scanner's body coil within a breath‐hold after administration of a 800 mL Tedlar bag of ambient air corresponding to the volume of administrated xenon gas. Dissolved‐phase imaging was acquired using a 3D Cartesian CSI with a matrix size of 20 × 20 × 6 covering a FOV of 40 × 40 × 20 cm³. A spectrally tailored RF pulse with a duration of 1.2 ms and partial self‐refocusing was designed to excite the dissolved and gas phases with flip angles of 10° and 0.1° and passbands of 500 Hz and 200 Hz, respectively [18]. TR was 10.6 ms resulting in a total acquisition time of 15 s. Signal from the cardiac blood pool and large pulmonary vessels were suppressed using 20 non‐phase encoded dummy RF excitations with a 30° flip angle prior to image acquisition. This preparation train was used to deplete longitudinal magnetization of dissolved‐phase xenon within the imaging volume, resulting in residual magnetization of less than 6% prior to data collection. MRSI and MRS were spectrally encoded to 160 points at a bandwidth of 20 kHz. The interleaved MRS was acquired every 10th RF pulse in 144 dynamic samples.

2.2.3. Phase Contrast Proton Imaging

CO was measured using a retrospectively cardiac‐triggered phase‐contrast sequence. A 2D phase contrast imaging slice was positioned perpendicular to the ascending aorta between the aortic valve and the coronary ostia. Imaging parameters were as follows: TR/TE = 5/3.1 ms, flip angle = 15°, FOV = 360 × 252 mm², matrix size = 192 × 128, slice thickness = 6 mm. Through‐plane velocity encoding was set to 160 cm/s. Analysis was done using Siswin software (Steffen Ringgaard, Aarhus, Denmark). Region‐of‐interests were positioned manually around the aorta, and CO was calculated in L/min.

2.3. Data Analysis

RBC oscillations were quantified as described by Bier et al. [31] by first localizing the RBC resonance peak in each spectrum. Signal points preceding the maximum signal intensity in the time resolved dissolved‐phase curves were excluded prior to fitting. This initial phase reflects non steady state inflow and exchange dynamics and was omitted to allow mono exponential modeling of the subsequent decay phase, consistent with prior literature [31]. RBC oscillations were analyzed using spectral analysis and quantified via fast Fourier transform of the time‐series, with RBC oscillation amplitude defined as the mean difference between peaks and lows (Figure 3). Xenon gas transfer ratios: RBC to membrane (RBC:M), RBC to gas (RBC:G) and membrane to gas (M:G), and chemical shifts were fitted using Advanced Method for Accurate, Robust and Efficient Spectral fitting (AMARES) in OXSA [32] as described in a previous publication [33]. RBC:M and RBC:G are reported in arbitrary units consistent with prior literature, reflecting normalized signal intensity ratios rather than absolute calibrated concentrations.

FIGURE 3.

Quantification of RBC oscillation amplitude. (A) Dynamic MRS showing RBC and gas phases. (B) Temporal spectra summary; red dot marks RBC phase. (C) Dynamic RBC signal clipped at peak (build‐up excluded). (D) RBC oscillations after exponential fit; red dots mark detected amplitude peaks and lows.

2.4. Statistics

Statistical analysis was performed using GraphPad Prism version 10.5.0 for Windows (GraphPad Software, Boston, Massachusetts, USA). Hemodynamic metrics during each intervention were compared with dissolved‐phase parameters using simple linear regression analysis.

Intervention effects were assessed using one‐way repeated measures ANOVA, followed by Dunnett's multiple comparisons test versus baseline. This approach accounts for within subject pairing and controls for multiple comparisons against a single control condition. In the presence of missing values, a linear mixed effects model with intervention as a fixed effect and pig as a random effect was used, followed by Dunnett adjusted comparisons versus baseline. Linear regression analyses were performed on pooled data across all interventions. These analyses did not account for within subject clustering and should therefore be interpreted as exploratory. Linear regression results are reported as slope with confidence interval, coefficient of determination (R ²), and p value. Results from ANOVA and mixed effects analyses are reported as p values and considered statistically significant at p < 0.05. All one‐way ANOVA and mixed effects analyses were done as repeated measurements of baseline and interventions.

Model assumptions for linear regression were evaluated by visual inspection of residuals versus predicted values. Homoscedasticity of residuals from ANOVA and mixed effects analyses was assessed using residuals versus predicted values, and normality of residuals was evaluated using QQ‐plots. All statistical tests were two sided.

To quantify the regional heterogeneity and spatial distribution shifts of the xenon metrics, voxel‐wise distributions were analyzed within the gas‐phase mask for each subject and intervention. In addition to the mean values, the following distribution parameters were calculated: (i) Coefficient of Variation (CoV), defined as the ratio of the standard deviation to the mean (σ/μ × 100), to assess relative spatial dispersion; (ii) Skewness, to measure the asymmetry of the distribution; and (iii) Kurtosis, to assess the relative concentration of values around the mean and the presence of outliers in the distribution.

3. Results

3.1. Intervention Effects on RBC Dissolved‐Phase Metrics

Interventions significantly affected dissolved‐phase ¹²⁹Xe metrics, predominantly within the RBC compartment.

The RBC:G ratio showed significant overall variation (p = 0.015), significantly decreasing during hypoxia and increasing during adenosine and dobutamine infusion (Figure 4B, Table S2). The RBC:M ratio followed a similar pattern (p < 0.001), significantly decreasing during hypoxia and increasing during adenosine and dobutamine infusion (Figure 4C, Table S2).

FIGURE 4.

HP ¹²⁹Xe dissolved‐phase parameters during interventions (A) Membrane to gas ratio (M:G). (B) RBC to gas ratio (RBC:G). (C) RBC to membrane ratio (RBC:M), (D) RBC chemical shift and (E) RBC oscillation amplitude. Asterisks denote statistical significance (p < 0.05 (*); p < 0.01 (**); ns, not significant). Mean and SD using multiple comparisons can be found in Table S2.

The RBC chemical shift varied significantly between interventions (p = 0.030), showing an increase in chemical shift during hypoxia and values during adenosine and dobutamine infusion that were comparable to baseline conditions (Figure 4D, Table S2).

The omnibus ANOVA detected an overall effect of intervention on RBC oscillation amplitude (p = 0.018). However, planned post hoc comparisons relative to baseline were not significant (Figure 4E, Table S2), suggesting that the overall effect was driven by differences among intervention conditions rather than baseline‐associated changes. Representative dissolved‐phase images from all interventions can be seen in Figure 5.

FIGURE 5.

Dissolved‐phase images quantified by Advanced Method for Accurate, Robust and Efficient Spectral fitting (AMARES) of all interventions on pig 2, representative of overall outcome of all pigs. A center slice showing membrane to gas ratio (M:G), red blood cell to gas ratio (RBC:G), red blood cell to membrane ratio (RBC:M) and Chemical shift (CS) of red blood cell (RBC) during each intervention. Ratios shown in arbitrary units (a.u.) and chemical shift in parts per million (ppm).

3.2. Hemodynamic Responses to Vasoactive Interventions

mPAP increased significantly during hypoxia before and during the breath hold. No significant changes were observed during ADP or adenosine and dobutamine infusion (Figure 6A,B, Table S3). MAP remained stable before the breath hold across all interventions but decreased during breath holds, with the largest reductions observed during hypoxia and during adenosine and dobutamine infusion (Figure 6C,D, Table S3).

FIGURE 6.

Physiological responses to interventions with standard deviation (SD) and mean. (A–D) shows changes in mean peripheral arterial pressure (MAP) and mean pulmonary arterial pressure (PAP) before and during breath hold during MRI. E–H shows cardiac output (CO), oxygen saturation (sO₂), pulmonary vascular resistance (PVR) and pulmonary arterial occlusion pressure (PAOP) momentarily before MRI. Statistical analysis using mixed‐effects analysis and pairwise comparison. Asterisks denote statistical significance (p < 0.05 (*); p < 0.01 (**); p < 0.0001 (***); ns, not significant). p values can be seen in Table S3.

CO increased significantly during adenosine and dobutamine infusion (Figure 6E, Table S3). sO₂ elicited a non‐significant decreasing trend during hypoxia and ADP (Figure 6F, Table S3). PaO₂ significantly decreased during all interventions. The remainder of arterial blood gas parameters remained stable (Table S4 and Figure S6). PVR seemed to increase during hypoxia and decrease during adenosine and dobutamine infusion as well as during ADP administration, yet was not statistically significant (Figure 6G, Table S3). PAOP decreased significantly across all interventions, most pronounced during hypoxia (Figure 6H, Table S3).

3.3. Linear Regression of Dissolved‐Phase Metrics and Pulmonary Hemodynamics

No significant associations were observed in linear regressions comparing CO, PVR, or PAOP and RBC oscillation amplitude.

RBC oscillation amplitude showed a weak yet significant association with mPAP before (R ² = 0.158, p = 0.040) and during (R ² = 0.146, p = 0.050) breath hold, with no meaningful relationship after breath hold (R ² = 0.033, p = 0.366) (Figure 7A–C; Table 2).

FIGURE 7.

Linear regression analyses of MRS metrics and mean pulmonary arterial pressure (mPAP) measured before, during and after breath hold. Regression analyses are shown with 95% CI lines, R ² and p values, functions are described as Y = a * X + b, where “Y” is respectively RBC oscillation amplitude, RBC:M and RBC:G, and “x” is mPAP. (A‐C) Linear regression analyses of RBC oscillation amplitude to mPAP. (D–F) Linear regression analyses of RBC:M to mPAP (G–I) Linear regression analyses of RBC:G to mPAP.

TABLE 2.

Heterogeneity and statistical summary of ¹²⁹Xe MRSI metrics across interventions vs. baseline.

Metric	Stat	ANOVA p	vs. Hypoxia p	vs. Adenosine dobutamine p	vs. ADP p
RBC:G	Mean	< 0.01	0.03	0.03	0.22
	CoV (%)	0.01	0.06	0.22	0.56
RBC:M	Mean	< 0.01	0.03	0.03	0.06
	Skewness	< 0.01	0.03	0.03	0.03
	Kurtosis	< 0.01	0.03	0.06	0.03
M:G	Mean	0.44	0.22	0.69	0.44
CSRBC	Mean	< 0.01	0.03	0.06	0.22
	Skewness	< 0.01	0.03	0.84	0.06

Note: Includes mean, coefficient of variation (CoV), skewness, and kurtosis for all subjects across baseline, hypoxia, adenosine‐dobutamine, and ADP interventions. p values are derived from one‐way repeated measures ANOVA with post hoc comparisons vs. baseline. Statistically signifacant results are indicated in bold. For more details see Table S5.

For RBC:M, the association with mPAP was significant before (R ² = 0.257, p = 0.048), during (R ² = 0.343, p = 0.045), and after (R ² = 0.274, p = 0.047) breath hold (Figure 7D–F; Table 2).

RBC:G was also significantly correlated to mPAP before (R ² = 0.222, p = 0.013), during (R ² = 0.270, p = 0.006), and after (R ² = 0.219, p = 0.014) breath hold (Figure 7G–I; Table 2).

3.4. Xenon Metrics Heterogeneity

The spatial distribution of xenon metrics exhibited significant sensitivity to physiological interventions (Table 2; Detailed results in Table S5). A one‐way repeated measures ANOVA revealed that while the M:G remained spatially stable across all conditions (ANOVA p > 0.30), metrics associated with the RBC compartment showed significant shifts in both magnitude and distribution shape.

The heterogeneity of the RBC:G ratio, measured by the CoV, changed significantly across the study (ANOVA p = 0.01), with a trend toward increased dispersion during hypoxia (p = 0.06). The RBC:M ratio demonstrated the most pronounced morphological shifts, with significant global changes in skewness and kurtosis (ANOVA p < 0.01). Specifically, hypoxia significantly altered RBC:M skewness (p = 0.03) and kurtosis (p = 0.03), indicating a transition toward a more heavy‐tailed distribution. Similarly, the chemical shift of xenon in RBCs (CS_RBC) showed significant global changes in mean value and distribution symmetry (skewness ANOVA p < 0.01), further reflecting the regional heterogeneity of the pulmonary physiological response to the investigated stressors.

A detailed summary of the distribution parameters and statistical comparisons for all xenon metrics is provided in Table S5.

4. Discussion

This study examined dissolved‐phase hyperpolarized ¹²⁹Xe metrics under tightly controlled alterations in cardiopulmonary hemodynamics, enabling direct, time‐matched comparison with invasive measurements of pulmonary vascular pressure. By experimentally modulating hemodynamic state, we demonstrate that RBC:M, RBC:G, and RBC oscillation amplitude exhibit consistent inverse relationships with mPAP, supporting the sensitivity of ¹²⁹Xe uptake to changes in capillary recruitment and pulmonary vascular loading. These results extend prior human studies by clarifying how dissolved‐phase metrics depend on both hemodynamic state and measurement timing, factors that are difficult to disentangle in heterogeneous clinical cohorts.

Using a porcine model enabled concurrent invasive measurement of mPAP before and during scanning and allowed repeated hemodynamic manipulations. To our knowledge, this is the first HP ¹²⁹Xe MRI study to report repeated extensive hemodynamic measures from subjects during different hemodynamic states, along with measurements while scanning.

The quantitative analysis of spatial distribution confirms that the interventions investigated in this study—hypoxic and pharmacological stress—exerted a profound and non‐uniform effect on pulmonary gas exchange. To address the shift in spatial distributions observed in our MRSI data, we employed a comprehensive suite of heterogeneity metrics, including the coefficient of variation (CoV), skewness, and kurtosis.

The global shift in the RBC:G ratio (ANOVA p < 0.01) suggests a systemic response to these physiological stressors. Specifically, the significant increase in RBC:G mean during hypoxia (p = 0.03) and adenosine‐dobutamine infusion (p = 0.03) aligns with expected changes in pulmonary hemodynamics [12]. Crucially, the RBC:G CoV showed a significant global change (ANOVA p = 0.01), and a trend toward increased heterogeneity during hypoxia (p = 0.06). This suggests that hypoxia not only alters the magnitude of xenon uptake but also challenges the lung's ability to maintain uniform ventilation‐perfusion matching, likely due to regional variations in hypoxic pulmonary vasoconstriction [34, 35].

Perhaps the most striking finding was the sensitivity of the RBC:M ratio's distribution shape to physiological changes. Unlike the M:G ratio, which remained statistically stable across all metrics (p > 0.30), the RBC:M ratio exhibited significant global changes across all four distribution parameters (p < 0.01). During hypoxia, the significant shifts in skewness (p = 0.03) and kurtosis (p = 0.03) indicate that the intervention does not merely shift the mean efficiency of xenon transfer. Instead, it creates a more “heavy‐tailed” and “peaked” distribution, implying the emergence of regional “hotspots” or “coldspots” of gas transfer efficiency [36]. This shift in distribution is indicative of recruited or de‐recruited capillary beds that behave differently than the baseline pulmonary parenchyma [36].

The chemical shift of ¹²⁹Xe in red blood cells (CS_RBC) is a known surrogate for blood oxygenation and the physical environment of hemoglobin [12]. Our results show significant global alterations in CS_RBC mean and heterogeneity (ANOVA p < 0.01). Specifically, during hypoxia, the CS_RBC skewness and kurtosis were significantly altered (p = 0.03), suggesting that the reduction in oxygen saturation is not spatially uniform. These findings provide evidence of intra‐pulmonary oxygenation gradients [37] that are exacerbated during respiratory stress, which would be masked if only mean values were considered.

Dissolved‐phase parameters were inversely correlated to mPAP, in line with several previous human reports [15, 21, 22, 31, 38]. This relationship was highlighted by Costelle et al. [22], who demonstrated a correlation between RBC oscillations and PVR using a simple model correcting for capillary blood volume. In our study, RBC oscillations correlated with mPAP but did not demonstrate consistent associations with other hemodynamic parameters such as CO, PVR, or PAOP. This divergence suggests that RBC oscillatory dynamics are primarily reflected by fluctuations in capillary filling pressure driven by the cardiac cycle rather than changes in flow.

The loss of association between RBC oscillation amplitude and mPAP following breath hold, unlike RBC:G and RBC:M, may suggest that oscillation amplitude is more sensitive to cardiovascular dynamics, including pulmonary capillary blood volume, whereas RBC:G and RBC:M are likely more sensitive to xenon gas transfer processes. Alternatively, biological differences between pigs, including variations in hemoglobin concentration and individual cardiovascular reserve capacity, may lead to more volatile cardiopulmonary hemodynamic responses during prolonged breath holds (> 15 s). This may weaken correlations across dissolved‐phase metrics, with RBC cardiogenic oscillations appearing particularly susceptible and therefore losing statistical significance.

Prior work has shown that diseases known to affect oxygenation can affect dissolved‐phase metrics; for example, Niedbalski et al. [16] reported an RBC:M ratio of 0.59 in healthy volunteers and reduced it to 0.39 in patients with PAH (34% decrease). In our study, we found a similar reduction, from 0.22 at baseline to 0.16 during hypoxia‐induced PAH (28% decrease).

The ADP intervention was designed to cause vasoconstriction without altering the paramagnetic effects on RBC chemical shift from oxygen found at baseline. Consistent with this design, we observed a modest reduction in sO₂ accompanied by a mild increase in mPAP (Table S3). When comparing dissolved‐phase responses between ADP and hypoxia, the chemical shift closely mirrored changes in sO₂, which decreased mainly during hypoxia, whereas RBC:G and RBC:M appeared to be primarily influenced by mPAP, which was elevated in both interventions (Figure 4). Although lower FiO₂ levels during the hypoxia intervention may have induced stronger pulmonary vasoconstriction and deviated more from the ADP intervention, pilot experiments demonstrated unacceptable physiological instability at 10% FiO₂, and 15% FiO₂ was therefore chosen as a compromise between intervention strength and experimental safety.

The heterogeneity of pulmonary hypertension complicates interpretation of RBC oscillations: pre‐capillary PH, indicated by elevated mPAP, suppresses RBC oscillations, whereas post‐capillary PH, indicated by elevated PAOP, enhances them, leading to overlapping values in mixed disease [21]. In our model, the interventions influenced pre‐, mid‐, and post‐capillary vessels, providing a model mimicking conditions seen in chronic obstructive pulmonary disease.

Our findings indicate that the interventions influenced both mPAP and PAOP, with mPAP exhibiting the largest absolute changes, while PAOP showed greater relative changes and stronger statistical significance (Figure 6). Despite the stronger association between interventions and PAOP, the pattern of diminished RBC oscillations with increased mPAP suggests that the pre‐capillary conditions outweigh the post‐capillary influence. This interpretation is supported by no correlation between PAOP and RBC oscillation amplitude, along with a clear correlation between mPAP and RBC oscillation amplitude (Figure 7).

Dissolved‐phase parameters and RBC oscillation amplitude were lower, and effects of disease models were less pronounced than those reported in previous human studies [15, 21, 22, 31]. Baseline RBC:M values were lower than some reported human values [12]; however, these differences are likely explained by subjects in this experiment being age‐equivalent to a child, the hemoglobin molecule being different in pigs [39], or lower FiO₂ during imaging, as RBC:M is known to increase with higher inspired oxygen fractions [40].

This attenuation may reflect interspecies physiological differences, healthy pulmonary parenchyma in young pigs, the effects of anesthesia and mechanical ventilation [40], or differences in MR acquisition and processing techniques. We chose an F_iO₂ of 21% to mimic atmospheric air, yet previous studies have shown, that using 40% F_iO₂ in a porcine model produces results that more closely resembles those found in humans [33]. Studies with fully anesthetized pigs are typically performed at a FiO₂ of 40% to account for anesthesia related drops in oxygenation, which is also the case in humans during surgery with high levels of anesthesia [41]. Therefore, FiO₂ of 21% is expected to cause a lower saturation when anesthetized as compared to being awake, as shown in Supplementary Table S3 [33] (sO₂˜0.96).

While our interventions were generally successful, the ADP infusion did not reproduce previously reported controllable stable mPAP elevations [28, 29, 42, 43]. We found vast volatility in mPAP which rapidly increased without reaching a plateau, before spontaneously returning to conditions comparable to baseline. Oxygenation during the ADP intervention was found to be inconsistent, with some pigs showing a vast decrease in pO₂, and others unaffected by the intervention. The heterogeneous results from this intervention may mask potential correlations compared to baseline.

Building on this framework, and knowledge of non‐invasive and invasive measurements of CO and MAP being highly correlated [44, 45], our results indicate that integrating non‐invasive hemodynamic measures such as CO and MAP in models estimating mPAP or PVR, rather than relying solely on oxygen‐related parameters, may offer additional predictive value. The relationship between dissolved‐phase xenon metrics and cardiopulmonary hemodynamics is likely governed by a multivariable physiological system involving both pre and post‐capillary pressures. While this study employed univariable linear models for interpretability, future work using larger cohorts may benefit from multivariable regression or dimensionality reduction approaches to better capture the coupled nature of pulmonary vascular physiology.

Several limitations should be considered in the study: The porcine cardiopulmonary system closely resembles that of humans [46], yet species‐specific differences, including a right‐shifted oxygen‐hemoglobin dissociation curve [39], vascular structure, and cardiopulmonary dynamics, may affect dissolved‐phase results. Furthermore, the correlation between oxygenation and RBC chemical shift has previously been proven [33] to be opposite to what is seen in humans [12]. Experiments were performed under general anesthesia, known to cause cardiovascular drift [47], the effects of which are mitigated here by randomized interventions. Also, modeling chronic disease states found in elderly patients by inducing acute pathology in young, healthy pigs with compliant lungs introduces a discrepancy between clinical conditions and our experimental model. Because repeated measurements were obtained from each animal, the linear regression analyses may be influenced by within‐subject correlation, which could affect statistical significance. These results should therefore be interpreted cautiously. Although the overall protocol was successful, the ADP intervention proved particularly challenging, as evidenced by the heterogeneous responses observed after 4–7 min. Finally, the small sample size (n = 7) limits the statistical power of the study.

5. Conclusion

This study demonstrates that dissolved‐phase parameters derived from hyperpolarized ¹²⁹Xe MRSI are sensitive to controlled alterations in pulmonary hemodynamics and exhibit consistent inverse relationships with invasively measured mean pulmonary arterial pressure. By isolating the effects of pulmonary vascular pressure through controlled hemodynamic modulation and time‐matched reference measurements, these results provide direct evidence that dissolved‐phase ¹²⁹Xe metrics reflect pulmonary vascular loading and capillary recruitment. Although variability across interventions and moderate correlation strengths indicate the need for further refinement, these findings establish a robust physiological basis for the continued development and validation of HP ¹²⁹Xe MRSI as a non‐invasive adjunct to right heart catheterization for assessing pulmonary hemodynamics.

Funding

This work was supported by AP Møller Fonden, granted 23/1‐2021. Karen Elise Jensen Fonden, granted 30/6‐2022.

Conflicts of Interest

M.V. and R.F.S. are employees of GE HealthCare. The authors report no conflicts of interest.

Supporting information

Table S1: Order of interventions. Pigs 2, 4 and 7 randomly underwent the same order of interventions. Pig 6, Ambient experiment was done due to technical issues disallowing hypoxia intervention. Adenosine Diphosphate (ADP).

Table S2: MRS and dissolved‐phase parameters during each intervention shown as mean ± SD and p values from mixed effects analysis. Graphs can be seen in Figure 4. Adenosine Diphosphate (ADP).

Table S3: Physiological responses to interventions shown as mean ± SD with p values. Graphs can be seen in Figure 6. Adenosine Diphosphate (ADP).

Table S4: Arterial blood gas samples of partial pressure of O₂ (PaO₂), partial pressure of CO₂ (PaCO₂), oxygen saturation (sO₂), hemoglobin (HgB), Hematocrit (Hct) and potential hydrogenii (pH) shown as mean (SD) with p values. Blood was drawn from the femoral artery and analyzed within 30 min to ensure quality. Arterial blood gas analyses were used to determine real time health, when health was ensured, the next step would commence. Adenosine Diphosphate (ADP).

Table S5: Regional Heterogeneity and Distribution Parameters of 129Xe MRSI Metrics. Includes Mean, Coefficient of Variation (CoV), Skewness, and Kurtosis for all subjects across baseline, hypoxia, adenosine‐dobutamine, and ADP interventions. p values are derived from one‐way repeated measures ANOVA with post hoc comparisons vs. baseline.

Figure S6: Arterial blood gas drawn immediately before each intervention shown as mean and SD. (A) pressure of O₂ (PaO₂), (B) partial pressure og CO₂ (PaCO₂), (C) Oxygen saturation (sO₂), (D) Hemoglobin (HgB), (E) Hematocrit (Hct), (F) potential hydrogenii (pH). Asterisks denote statistical significance (p < 0.05 (*); p < 0.01 (**); p < 0.0001 (***); ns, not significant).

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.