MR Spectroscopic Imaging of Hyperpolarized 129-Xenon in the Dissolved-Phase to Determine Regional Chemical Shifts of Hyperoxia in Healthy Porcine Lungs

Abstract

Lung MRI with hyperpolarized xenon (¹²⁹Xe) gas reveals key characteristics of pulmonary physiology such as ventilation and alveolar-capillary gas transfer. Magnetic resonance spectroscopic imaging (MRSI) offers insights into regional oxygenation saturation (sO₂) through chemical shift changes related to xenon-hemoglobin binding. The similarity between porcine and human anatomy and physiology, particularly in terms of lung volume, airway structure, and alveolar-capillary microstructure, offers the opportunity to investigate physiological effects linked to oxygen supply using ¹²⁹Xe MRSI. We hypothesize that ¹²⁹Xe MRSI can detect regional chemical shift changes related to red blood cell oxygenation and arterial oxygen partial pressure (p_aO₂) in a porcine model. Imaging was performed on a 3-T clinical MRI scanner on four healthy pigs mechanically ventilated at fractional inspired oxygen levels (FiO₂) of 40% and 100%. Dissolved-phase images were acquired using a 3D Cartesian MRSI sequence with a spherical sampling pattern in a matrix size of 28 × 28 × 6. A spectrally tailored RF pulse excited the dissolved and gaseous phases with flip angles of 10° and 0.1°, respectively. Repetition time was 7.4 ms resulting in a total acquisition time of 18 s. In addition, ¹²⁹Xe ventilation, pulmonary anatomical scans, dynamic contrast-enhanced perfusion, and arterial blood gas were measured at each FiO₂. Pair-wise comparisons were performed between inspired oxygen levels, along with linear regression analysis of p_aO₂ and dissolved-phase chemical shift imaging. Porcine lung lobes were segmented, and two-way ANOVA were performed to evaluate regional effects of oxygen concentrations. Arterial blood gas and cardiopulmonary measures showed an increase in p_aO₂ with the increase in FiO₂. Ventilation defect percentage and perfusion metrics did not significantly change with higher oxygen concentration. Dissolved-phase ratios of red blood cells (RBC) to membrane increased with higher oxygen concentration. Increasing inspired oxygen resulted in a lower RBC chemical shift and increased linewidth, indicating RBC measures are sensitive to p_aO₂. Simple linear regression analysis of RBC chemical shift and a multiple linear regression model including linewidth were applied for regional p_aO₂ maps. Regional effects of oxygen were confirmed in the segmented lung lobes. Dissolved-phase ¹²⁹Xe chemical shift of RBC decreased linearly with p_aO₂ in healthy porcine lungs. Regional chemical shift, linewidth, and signal ratio changes were determined in dissolved-phase imaging of RBC at 40% and 100% FiO₂. Our data suggest that regional p_aO₂ prediction is possible with a multiple linear regression model including RBC chemical shift and linewidth as combined effect of oxygen across animal lung lobes affects regions differently.

Keywords: arterial oxygen partial pressure; blood oxygenation saturation; chemical shift imaging; dissolved‐phase imaging; hyperpolarized xenon gas; magnetic resonance imaging; porcine animal model.

FIGURE 1

Scan session overview included a 20‐min wait between interventions to ensure oxygen saturation and pulmonary gadolinium contrast washout. Interventions are illustrated with grey boxes. DCE = dynamic contrast‐enhanced imaging, dissolved = dissolved‐phase imaging, FiO₂ = fractional inspired oxygen level.

FIGURE 2

Illustration of the 3D spherical undersampled Cartesian sampling trajectory in 3D (A), through the center plane at kz = 0 periods/voxel (B), through the center plane at ky = 0 periods/voxel (D), and the point spread function (PSF) (C). k‐space normalized in all three dimensions to ± 0.5 periods/voxel.

FIGURE 3

Fitted MRSI spectra at the two fractional inspired oxygen levels (FiO₂). (A) MRSI spectra of the center slice with indication of selected spectra (B–E) close to the heart (green) and in the lung periphery (orange). Fitted spectra of FiO₂ at 40% (B,D) and 100% (C,E) are illustrated at the right with real values of original spectra (black) and AMARES fitted spectra of the dissolved phases (membrane = orange, RBC = red) and gas (blue). Fitting residual values are shown in green. Zoom inserts at 185–215 ppm are shown to improve visualization of the dissolved‐phase region (dotted blue line). RBC = red blood cells, AMARES = advanced method for accurate, robust, and efficient spectral.

FIGURE 4

Ventilation images at 40% (left) and 100% (right) fractional inspired oxygen levels (FiO₂). Lung volume T₁‐weigthed ¹H images are shown with a corresponding ¹²⁹Xe ventilation image overlay used to calculate ventilation defect percentage.

FIGURE 5

Dissolved‐phase MRSI quantified with AMARES fitting into ratios, chemical shift, and linewidth acquired at 40% (first column) and 100% (second column) inspired oxygen levels (FiO₂). (A) Chemical shift images of gas, membrane, RBC and the difference between RBC and membrane. (B) Ratios of dissolved‐phase gas, membrane, and RBC signal ratios. (C) Linewidth of gas, membrane, and RBC. Images have been masked based on 0.6 times the gas signal mean value. Only the central slice is shown. AMARES = advanced method for accurate, robust, and efficient spectral, a.u. = arbitrary units, G = gas, M = membrane, RBC = red blood cells.

FIGURE 6

Chemical shift values (A), ratios (C), signal‐to‐noise ratios (D) and linewidths (E) of gas and dissolved phases and perfusion (B) at 40% and 100% oxygen levels. Perfusion measures include mean transit time (MTT, s), plasma flow (PF, mL/100 mL/min), volume of distribution (VD, mL/100 mL), and first‐order moment (FOM, ms). *p value < 0.05, **p value < 0.01, ***p value < 0.001, a.u. = arbitrary units, M = membrane, ns = nonsignificant, RBC = red blood cells.

FIGURE 7

Simple linear regression fit between p_aO₂ and RBC chemical shift (A) and RBC spectra linewidth (B). The correlation between predicted and measured p_aO₂ as a multiple linear regression including RBC chemical shift and linewidth (C). Human and porcine oxygen‐hemoglobin dissociation curve (D) based on previously published models [23] with indication of p_aO₂ at FiO₂ = 40% and 100%. FiO₂ = fractional inspired oxygen levels, Hb = hemoglobin, hemoglobin oxygen saturation (sO₂), p_aO₂ = arterial oxygen partial pressure, RBC = red blood cells, RBC_cs = dissolved‐phase RBC chemical shift, RBC_lw = dissolved‐phase RBC linewidth.

FIGURE 8

Arterial oxygen partial pressure (p_aO₂) maps based on simple linear regression fit (SLR) and multiple linear regression fit (MLR) between RBC chemical shift, RBC linewidth and arterial blood gas p_aO₂ at fractional inspired oxygen levels (FiO₂) of 40% (top rows) and 100% (bottom rows). RBC = red blood cells.

FIGURE 9

Perfusion images calculated from the TRICKS T₁w DCE images masked to the lung. Calculated images include mean transit time (MTT), plasma flow (PF), volume of distribution (VD), and first‐order moment (FOM) acquired at fractional inspired oxygen level (FiO₂) of 40% (top row) and 100% (bottom rows). Only the central slice is shown. DCE = dynamic contrast enhanced, T₁w = T₁ weighted, TRICKS = time‐resolved imaging of contrast kinetics.

FIGURE 10

Regional dissolved‐phase ratios, chemical shifts, and linewidth at fractional inspired oxygen levels (FiO₂) of 40% (white circle) and 100% (black circle). Two‐way ANOVA results are shown in Table 2.

FIGURE 11

Regional perfusion at fractional inspired oxygen levels (FiO₂) of 40% (white circle) and 100% (black circle). (A) Mean transit time, (B) plasma flow, (C) volume of distribution, and (D) first‐order moment. Two‐way ANOVA results are shown in Table 2.

FIGURE 12

RBC chemical shift according to oxygenation. (A) Chemical shift of red blood cell (RBC) signal from increased partial pressure of oxygen (p_aO₂) fitted with a logarithmic regression model. (B) Delta chemical shift of membrane and RBC from increasing hemoglobin oxygen saturation (sO₂), from hypoxia to hyperoxia, fitted with an exponential model.