Impact of vasodilation on oxygen-enhanced functional lung MRI at 0.55T

Abstract

Background:

Oxygen enhanced magnetic resonance imaging (OE-MRI) can be used to assess regional lung function without ionizing radiation. Inhaled oxygen acts as a T1-shortening contrast agent to increase signal in T1-weighted (T1w) images. However, increase in proton density from pulmonary hyperoxic vasodilation may also contribute to the measured signal enhancement. Our aim was to quantify the relative contributions of the T1-shortening and vasodilatory effects of oxygen to signal enhancement in OE-MRI in both swine and healthy volunteers.

Methods:

We imaged 14 anesthetized female swine (47 ± 8 kg) using a prototype 0.55T high-performance MRI system while experimentally manipulating oxygenation and blood volume independently through oxygen titration, partial occlusion of the vena cava for volume reduction, and infusion of colloid fluid (6% hydroxyethyl starch) for volume increase. Ten healthy volunteers were imaged before, during and after hyperoxia. Two proton density weighted (PDw) and two T1w ultrashort echo time (UTE) images were acquired per experimental state. The median PDw and T1w percent signal enhancement (PSE), compared to baseline room air, was calculated after image registration and correction for lung volume changes. Differences in median PSE was compared using Wilcoxon signed-rank test.

Results:

The PSE in PDw images following 100% oxygen was similar in swine (1.66 ± 1.41%, P=0.01) and in healthy volunteers (1.99 ± 1.79%, P=0.02), indicating that oxygen-induced pulmonary vasodilation causes ~2% lung proton density increase. The PSE in T1w images following 100% oxygen was also similar (swine, 9.20 ± 1.68%, P<0.001; healthy volunteers, 10.10 ± 3.05%, P<0.001). The PSE in T1w enhancement was oxygen dose-dependent in anesthetized swine, and we measured a dose-dependent PDw image signal increase from infused fluids.

Conclusions:

The contribution of oxygen induced vasodilation to T1w OE-MRI signal was measurable using PDw imaging, and was found to be ~2% in both anesthetized swine and in healthy volunteers. This finding may have implications for patients with regional or global hypoxia or vascular dysfunction undergoing OE-MRI, and suggest that PDw imaging may be useful to account for oxygen-induced vasodilation in OE-MRI.

Oxygen-enhanced (OE) MRI is a tool for assessing regional lung function without the use of ionizing radiation or hyperpolarized gas.^1–6 OE-MRI can be used for diagnosis, risk-stratification and radiation-free longitudinal treatment monitoring of asthma and chronic obstructive pulmonary disease.^2,7–9 Recently, oxygen-enhanced MRI has been developed for a high-performance 0.55T MRI system, which offers reduced susceptibility artifacts in the lung and higher oxygen T1 relaxivity compared to higher field strengths^10–12

OE-MRI uses inhaled 100% oxygen as a T1 shortening contrast agent. The dissolved paramagnetic oxygen in the blood and tissues causes signal enhancement on T1-weighted (T1w) images, and therefore OE-MRI is expected to be sensitive to changes in both ventilation and perfusion. Moreover, oxygen is also a pulmonary vasodilator. It can increase lung proton density due to an oxygen-induced increase in pulmonary blood volume. However, the relative contributions of T1-shortening and pulmonary vasodilation to OE-MRI signal have not been characterized.

Oxygenation can be easily titrated by adjusting the inhaled gas mixture. Pulmonary blood volume may be temporarily decreased through reduction of venous return by partially occluding the inferior vena cava (IVC) or increased through administration of intravenous fluid. This allows for experimental independent manipulation of oxygenation and pulmonary blood volume in anesthetized swine, which can be assessed using T1w and PDw MRI image signal intensity.

We propose to use PDw images, in tandem with T1w images, to evaluate the relative contributions of vasodilation and T1-shortening in OE-MRI using experimental provocations in swine. Then, we use PDw and T1w images to assess the contribution of oxygen induced vasodilation to OE-MRI in healthy volunteers using 0.55T MRI.

Methods

Experimental provocations in swine

To assess the impact of pulmonary vasodilation on OE-MRI, we performed a series of experimental provocations to manipulate oxygenation and pulmonary blood volume in 14 anesthetized female swine (46.7 ± 8 kg). This study was approved by the NHLBI Animal Care and Use Committee. All swine were anesthetized using isoflurane and mechanically ventilated. Six swine underwent an experimental “oxygen titration protocol” and 6 swine underwent a “pulmonary blood volume manipulation protocol”, as summarized in Figure 1. Additionally, 2 swine were imaged using a shorter protocol to replace missing data (see Supplemental Digital Content 1 and 2).

Figure 1:

A) Overview of the magnetic resonance imaging (MRI) and physiological measurements that were acquired in each experimental state. B) Three experimental protocols (two for swine, one for healthy volunteers) containing measurements in several experimental states.

Abbreviations: IVC – inferior Vena Cava. LV – Left ventricle. MPA – Main Pulmonary Artery. mPAP – Mean Pulmonary Arterial Pressure. MRI – Magnetic Resonance Imaging. PAWP – Pulmonary Arterial Wedge Pressure. PDw – Proton Density Weighted. PVR – Pulmonary Vascular Resistance. RV – Right Ventricle. T1w – T1-weighted. UTE – Ultrashort Echo Time.

In the oxygen titration protocol, the inhaled gas mixture was adjusted so that swine were exposed sequentially to room air (baseline, 21% oxygen), 100% oxygen, 50% oxygen, room air (repeat) and hypoxia. The imaging protocol began when fraction of inspired oxygen (FiO₂) was within 5% of the target levels of 100%, 50% and 21%. Hypoxia was induced by titrating oxygen and nitrogen to achieve 88% blood oxygen saturation by pulse oximetry, requiring FiO₂ between 13–17% across individual swine.

In the blood volume manipulation protocol, swine were exposed sequentially to room air, 100% oxygen, partial IVC occlusion at room air, infusion of 5 ml/kg hetastarch (Hespan 6%, B. Braun Medical Inc., Bethlehem, Pennsylvania, USA) at room air, additional infusion of 15 ml/kg hetastarch at room air, and finally 100% oxygen post 20ml/kg hetastarch. We achieved partial IVC occlusion using a double inflated balloon-wedge end-hole catheter, positioned under real-time MRI. Systolic blood pressure measured in the femoral artery was reduced by ~30 mmHg from baseline. This partial occlusion was sustained for ~20–30 minutes during imaging with periodic adjustment of balloon inflation to achieve a stable systemic systolic blood pressure to 50–65 mmHg. A colloid, hetastarch (6% hydroxyethyl starch, B. Braun Medical Inc., Bethlehem, Pennsylvania, USA), was selected instead of crystalloid to induce a sustained increase in blood volume due to its lower propensity to diffuse to the extravascular space.

In addition to imaging, we obtained the following measurements in each experimental state: arterial and venous blood gas samples, invasive systemic systolic and diastolic blood pressure measured in the femoral artery, mean pulmonary arterial pressure (mPAP) and pulmonary arterial wedge pressure measured by MRI-guided right heart catheterization, fraction of inspired oxygen (FiO₂), minimum alveolar concentration of isoflurane and saturation by pulse oximetry.

Healthy volunteers

We imaged 10 healthy volunteers (aged 26.7 ± 5.1, 7 male) with approval from the NHLBI Institutional Review Board. All healthy volunteers provided written informed consent. Healthy volunteers were imaged during inhalation of room air, 100% oxygen and then room air again with a 3-minute adjustment period at the start of each oxygenation level to achieve equilibrium prior to imaging. Oxygen was delivered through a non-rebreather face mask with a flow rate of 15 L/min.

Imaging protocol

Imaging was performed using a high-performance 0.55T MRI system (prototype MAGNETOM Aera, Siemens, Erlangen, Germany)¹². The following images were acquired in each experimental state for swine and healthy volunteers:

1. Two T1w coronal lung images using a 3D free-breathing stack-of-spirals gradient echo ultra-short echo time (UTE) sequence^13,14 (modified from Siemens Healthcare, Erlangen, Germany). Imaging parameters: repetition time (TR): 6.5 ms, echo time (TE): 0.04 ms, flip-angle (FA): 13°, in-plane resolution: 3.5 × 3.5 × 10 mm³, matrix size 128 × 128 × 32, acquisition time ~4.5 minutes.

2. Two PDw coronal lung images using a 3D free-breathing stack-of-spirals gradient echo UTE sequence. Imaging parameters: TR: 6.5 ms, TE: 0.04 ms, FA: 1°, resolution: 3.5 × 3.5 × 10 mm³, matrix size 128 × 128 × 32, acquisition time ~4 minutes. We simulated these sequence parameters to confirm that they generated PDw signal over a physiological range of T1 values (Supplemental Digital Content 3).

3. One 2D main pulmonary arterial flow measurement using a phase-contrast sequence. Imaging parameters: TR: 27.44 ms, TE: 4.19 ms, velocity encoding: typically 100 cm/second, FA: 30°, spatial resolution: 1.9 × 1.9 × 6 mm³, matrix size: 192 × 192, temporal resolution: ~20–30 ms, number of averages: 3, acceleration factor: 2, acquisition time: ~2 minutes.

4. Cardiac 4-chamber 2D T1 mapping (Modified look-locker inversion recovery [MOLLI])¹⁵. Imaging parameters: TR: 3.2 ms, TE: 1.28 ms, resolution: 1.4 mm x 1.4 mm x 8 mm³, matrix size: 256 × 192, FA: 35°, acceleration factor: 2, typical acquisition time: 17 seconds.

T1w and PDw 3D UTE lung images were self-gated using a superior-inferior navigator readout. The navigator signal was used to extract the respiratory signal which was then used to bin 40% of the data acquired during the most stable respiratory phase. These data were reconstructed to produce a static image free from respiratory motion artifacts.¹⁶

The percent signal intensity enhancement, compared to baseline room air, was measured from T1w and PDw images (Figure 2). First, the lungs were manually delineated in the baseline room air reference image. Second, all experimental state images were registered to the reference state image using a non-rigid registration algorithm.¹⁷ Despite respiratory binning, the most-stable respiratory phase in the UTE lung images varied due to differences in breathing patterns between image acquisitions. Third, to compensate for proton density changes related to contraction and expansion of lung parenchymal tissue across respiratory phases, all registered experimental state images were corrected using the Jacobian determinant of the image registration deformation fields ($\mathrm{d}\mathrm{e}\mathrm{t}\left(J_T\right)$), as previously described. ^7,10 Fourth, a pixel-wise percent signal enhancement (PSE) map was generated, and the median lung enhancement was calculated using the manually delineated lung mask. The median instead of mean PSE was used because the PSE in individual pixels may not be normally distributed.⁷

Figure 2:

Summary of the image analysis used to calculate percent signal intensity enhancement (PSE). Each experimental state image was registered to the reference baseline room air image in the same subject. The Jacobian determinant of the deformation fields (det(J_T)) was then used to produce a lung volume corrected registered experimental state image. Finally, a PSE map was produced using the corrected registered experimental state image and the baseline room air reference image, and the median PSE within the lung mask was calculated. The same analysis was applied for both PDw and T1w images.

Abbreviations: PSE – Percent Signal Enhancement. PDw – Proton Density Weighted. T1w – T1 Weighted.

One observer with >8 years of experience in MRI research, blinded to experimental state, performed a quality assessment of all images and registrations. All acquisitions with suboptimal image quality or image registration quality were subsequently reviewed in conference with a second experienced observer. Individual acquisitions were excluded when image registration quality was deemed unsatisfactory. In three swine, image registration parameters were optimized to successfully improve quality. In cases of excessive lung volume change between experimental states, lung volume correction using Jacobian determinant has previously been found to be unreliable. We excluded acquisitions with mean($\mathrm{d}\mathrm{e}\mathrm{t}\left(J_T\right)$) < 0.85 or > 1.15 within the manually delineated lung mask, indicating approximate lung volume change of >15% compared to baseline images.

The median lung PSE was calculated for each experimental state and averaged across the two PDw or two T1w acquisitions.

The T1 of lung tissue, left ventricular blood pool, and right ventricular blood pool were measured using manually drawn regions of interest from MOLLI T1 maps. Cardiac output was calculated from main pulmonary artery flow measured in phase-contrast images. The pulmonary vascular resistance was calculated using transpulmonary pressure gradient (mPAP – pulmonary arterial wedge pressure) divided by the cardiac output.

Repeatability of PSE measurements

The short-term repeatability of lung PSE was assessed in each subject by measuring the difference in median lung PSE within every pair of acquired T1w or PDw images from the same experimental state, and was reported as the mean ± standard deviation (SD) difference in median PSE across all image pairs.

Statistical analysis

Continuous variables were reported as mean ± SD. Differences in T1w and PDw PSE between baseline room air acquisitions, and between 100% and baseline room are were assessed using paired Wilcoxon signed rank test in both swine and healthy volunteers. This clinically-relevant experimental state pair was acquired in all animal protocols and healthy volunteers to allow sufficient power for statistical hypothesis testing. We used Friedman repeated measures analysis of variance by ranks to test for the presence of difference across multiple groups. We refrained from additional statistical testing in other experimental state pairs because our small sample size prohibited appropriate statistical testing.

Results

Repeatability of PSE measurements

In swine experiments, the mean intra-subject difference in lung PSE between the two consecutive baseline room air acquisitions was 0.01 ± 1.00 % (p>0.05) in T1w images and 0.78 ± 0.78% (p>0.05) in PDw images, indicating good repeatability of consecutive measurements. When repeating room air imaging after oxygen titration, we observed a larger difference in lung PSE (T1w PSE difference: 0.92 ± 1.14%, PDw PSE difference: 1.12 ± 2.54%), suggesting that repeatability diminishes with overall scan session duration.

Across all experimental states, the mean ± SD difference of the PSE values in intra-subject, intra-experimental state image pairs was 0.31 ± 0.61% in T1w images and 0.84 ± 0.71% in PDw images. This suggested that PSE variability was low in all consecutive repeated PDw and T1w acquisitions.

A similar pattern was found in the healthy volunteer cohort, where the mean intra-subject difference in lung PSE between the two consecutive baseline room air acquisitions was 0.73 ± 0.56% (p>0.05) in T1w images and 0.03 ± 1.48 in PDw images (p>0.05). Across all healthy volunteer subjects and experimental states, the mean ± SD difference in PSE values in intra-subject, intra-experimental state image pairs was 0.40 ± 1.03% in T1w images and 0.37 ± 1.94% in PDw images, confirming low PSE variability in consecutive repeated acquisitions.

T1w and PDw imaging during experimental provocations in swine

Figure 3 and Figure 4 show lung PSE for T1w and PDw imaging for oxygen titration and for volume manipulation experiments, respectively. Inhalation of 100% oxygen induced a lung PSE of 9.20 ± 1.68% in T1w images (P<0.001), versus a PSE of 1.66 ± 1.41% in PDw images (P=0.01). This finding suggests that lung proton density increased by a mean of 1.66 % from the displacement of air by blood as a consequence of pulmonary vasodilation following inhalation of 100% oxygen. Lung PSE during inhalation of 50% oxygen was 4.59 ± 1.51% in T1w images and 1.89 ± 1.84% in PDw images, suggesting oxygen-related PSE was dose-dependent in T1w images but constant in PDw images above 50% oxygen. Hypoxia resulted in a small positive T1w enhancement (mean: 1.41 ± 0.53%, n = 2) but a small negative PDw enhancement (−1.46 ± 1.11%, n = 4). Only a small number of datapoints acquired during hypoxia remained after excluding images with poor quality due to image misregistration caused by bulk motion and large lung volume changes (all data provided in Supplemental Digital Content 4). Partial occlusion of the IVC caused small but consistent signal decrease in both T1w and PDw images (T1w: −1.65 ± 1.51%, PDw: −2.57 ± 1.83%), indicating that T1w image signal can indeed be influenced by reduced proton density secondary to reduced pulmonary blood volume.

Figure 3:

Oxygen titration protocol results showing median percent signal enhancement (PSE), averaged across two acquisition per experimental state for each subject. Results are shown for proton density weighted (PDw) images (red bars), and T1-weighted (T1w) images (green bars).

Figure 4:

Blood volume manipulation experimental protocol results showing median percent signal enhancement (PSE), averaged across two acquisition per experimental state for each subject. Results are shown for proton density weighted (PDw) images (red bars and T1-weighted (T1w) images (green bars).

Low dose (5 ml/kg) and high dose (20 ml/kg) hetastarch at room air generated a modest signal intensity increase in T1w imaging (mean PSE: 1.71 ± 1.39% and 3.26 ± 3.25%, respectively) but a substantial dose-dependent PSE in PDw imaging (5.20 ± 1.90% and 11.63 ± 2.47%, respectively). The discrepancy between T1w and PDw PSE in room air after hetastarch infusion is compatible with the dose-dependent lengthening of T1 induced by hetastarch seen in T1-mapping images, as shown in Figure 5 and described in more detail below. When imaging was repeated after adding 100% oxygen post high dose hetastarch infusion (20 ml/kg), PSE was similar in T1w (13.5 ± 3.5%) and PDw images (12.6 ± 2.4%). This finding is consistent with the measured lung T1 that returned close to baseline from the balancing effects of hetastarch-related T1 lengthening and oxygen-related T1 shortening.

Figure 5:

The measured T1 percentage change from baseline room air in the lungs (gray bars), left ventricle (LV) blood pool (green bars), and the right ventricle (RV) blood pool (blue bars) for each experimental state.

T1 measurements in swine

Figure 5 shows the percent ΔT1 relative to room air baseline measured in MOLLI T1 mapping images in the left and right ventricular blood pools and in the lung. Mean absolute T1 values per experimental state are reported in Table 3. Oxygen inhalation resulted in a dose-dependent T1 shortening in lungs and the left ventricular blood pool, but not in the right ventricular blood pool. This is consistent with the large change in the arterial partial pressure of oxygen (PaO₂) achieved with oxygenation compared to the much more modest change in venous partial pressure of oxygen (PvO₂) (Table 1). Hetastarch increased T1 in the lungs and both ventricular blood pools in a dose-dependent fashion (ΔT1_lung after 5 ml/kg hetastarch: 4.27 ± 4.59%, after 20 ml/kg hetastarch: 19.44 ± 4.57%). This effect was reversible with 100% oxygen in the lungs (ΔT1_lung = −0.09 ± 2.74%).

Table 3.

Absolute T1 values measured in T1-mapping Modified Look-Locker Inversion Recovery (MOLLI) 4-chamber images.

T1-mapping by region of interest and experimental state
Swine
Oxygen titration protocol	Baseline RA	100% O2	50% O2	Repeat RA	Hypoxia
Lung T1, ms (mean ± SD)	1022 ± 49	890 ± 29	984 ± 58	1058 ± 43	1067 ± 44
Left Ventricle T1, ms (mean ± SD)	1303 ± 80	1067 ± 68	1215 ± 87	1299 ± 89	1277 ± 70
Right Ventricle T1, ms (mean ± SD)	1268 ± 86	1261 ± 80	1262 ± 96	1278 ± 125	1226 ± 84
Volume manipulation protocol	Baseline RA	100% O2	IVC occlusion	5 ml/kg HS RA	20 ml/kg HS RA	20 ml/kg HS O 2
Lung T1, ms (mean ± SD)	1007 ± 53	890 ± 62	1003 ± 68	1074 ± 66	1195 ± 91	998 ± 39
Left Ventricle T1, ms (mean ± SD)	1305 ± 79	1067 ± 77	1278 ± 56	1349 ± 58	1529 ± 120	1266 ± 100
Right Ventricle T1, ms (mean ± SD)	1257 ± 129	1261 ± 148	1306 ± 142	1301 ± 71	1419 ± 176	1390 ± 213
Healthy Volunteers	Baseline RA	100% O 2	Repeat RA
Lung T1, ms (mean ± SD)	992 ± 41	938 ± 64	1010 ± 51
Left Ventricle T1, ms (mean ± SD)	1144 ± 47	1000 ± 55	1127 ± 44
Right Ventricle T1, ms (mean ± SD)	1143 ± 45	1130 ± 71	1118 ± 36

Abbreviations: HS – Hetastarch. IVC – Inferior Vena Cava. RA – Room Air. SD – Standard Deviation.

Table 1.

Physiology parameters in oxygen titration in swine

Parameters	Baseline Room Air	100% O2	50% O2	Repeat Room Air	Hypoxia	P-value
Hemodynamic parameters, mean ± SD	n=6	n=6	n=6	n=6	n=5
Mean Pulmonary Arterial Pressure (mmHg)	17.7 ± 3.4	16.2 ± 3.7	16.2 ± 4.1	18.7 ± 5.5	24.8 ± 9.4	<0.001
Pulmonary Arterial Wedge Pressure (mmHg)	10.5 ± 3.8	10.3 ± 1.9	10.3 ± 2.3	10.3 ± 2.1	10.6 ± 3.8	N.S.
Pulmonary Vascular Resistance (WU)	1.93 ± 0.39	1.56 ± 0.70	1.37 ± 0.49	1.96 ± 0.97	3.41 ± 2.01	0.03
Cardiac Output (L/min)	3.68 ± 0.43	3.86 ± 0.47	4.24 ± 0.42	4.31 ± 0.46	4.60 ± 0.85	0.02
Heart Rate (beats/min)	85.8 ± 12.4	86.8 ± 11.6	91.3 ± 14.2	91.2 ± 13.1	107 ± 16.3	0.001
MAP (mmHg)	56.2 ± 13.6	52.5 ± 11.3	51.3 ± 11.3	51.8 ± 9.20	48.0 ± 10.5	N.S.
SBP (mmHg)	79.0 ± 14.1	74.5 ± 10.3	75.8 ± 10.2	74.8 ± 7.52	68.0 ± 16.4	N.S.
DBP (mmHg)	43.3 ± 12.4	39.8 ± 10.9	41.8 ± 12.3	38.5 ± 9.73	35.8 ± 9.02	N.S.
Oxygenation/anesthesia parameters, mean ± SD
PaO2 (mmHg)	107 ± 14.8	422 ± 37.4	235 ± 22.8	98.4 ± 25.4	49.2 ± 11.0	<0.001
SaO2 (%)	98.5 ± 0.8	100 ± 0	100 ± 0	98.0 ± 1.4	87.3 ± 7.6	<0.001
PvO2 (mmHg)	37.0 ± 4.9	48.8 ± 4.1	40.0 ± 3.6	35.3 ± 6.5	26.8 ± 4.0	<0.001
SvO2 (%)	74.2 ± 7.7	85.2 ± 3.1	78.3 ± 4.1	71.8 ± 7.6	56.0 ± 10.8	<0.001
FiO2 (%)	22.5 ± 1.1	96.3 ± 2.7	50.8 ± 3.3	21.2 ± 2.1	14.5 ± 1.9	<0.001
Isoflurane dose (MAC)	1.80 ± 0.20	1.90 ± 0.13	1.92 ± 0.12	1.87 ± 0.16	1.78 ± 0.16	N.S.

The P-values in the far right column show groupwise comparisons using Friedman test to test for differences across experimental states. Abbreviations: DBP – Diastolic Blood Pressure. FiO₂ – Fraction of Inspired Oxygen. L/min – Liters per Minute. MAP – Mean Arterial Pressure. mmHg – Millimeters Mercury. MAC – Minimum Alveolar Concentration. PaO₂ – Partial pressure of Oxygen in Arterial Blood. SaO₂ – Oxygen Saturation in Arterial Blood. SBP – Systolic Blood Pressure. SD – Standard Deviation. SvO₂ – Oxygen Saturation in Venous Blood. WU – Woods Units.

Physiological parameters during provocations in swine

Hemodynamic measurements including systemic and pulmonary blood pressures, and arterial and venous blood gas measurements are shown in Tables 1 and 2 for the two experimental protocols. The measured physiological parameters reflected the expected effects on hemodynamics and oxygenation. Invasive pressure measurements from the pulmonary and femoral arteries show negligible difference in hyperoxia and repeat room air compared to baseline room air. Hypoxia resulted in increased mPAP and pulmonary vascular resistance indicating hypoxic pulmonary vasoconstriction, as well as an increase in cardiac output to compensate for the decreased arterial blood oxygen saturation. During IVC occlusion, a markedly decreased cardiac output, lower mPAP and higher PVR suggested a reduction of venous return and possible increase in vasoconstriction related to decreased transpulmonary blood flow. In contrast, volume challenge by infusion of hetastarch resulted in a marked increase in cardiac output and mPAP but decrease in PVR, suggesting pulmonary vascular dilation to accommodate the increased blood volume.

Table 2.

Physiology parameters in swine undergoing manipulation of pulmonary blood volume.

Parameters	Baseline Room Air	100% O2	50% O2	Repeat Room Air	Hypoxia	P-value
Hemodynamic parameters, mean ± SD	n=6	n=6	n=6	n=6	n=5
Mean Pulmonary Arterial Pressure (mmHg)	17.7 ± 3.4	16.2 ± 3.7	16.2 ± 4.1	18.7 ± 5.5	24.8 ± 9.4	<0.001
Pulmonary Arterial Wedge Pressure (mmHg)	10.5 ± 3.8	10.3 ± 1.9	10.3 ± 2.3	10.3 ± 2.1	10.6 ± 3.8	N.S.
Pulmonary Vascular Resistance (WU)	1.93 ± 0.39	1.56 ± 0.70	1.37 ± 0.49	1.96 ± 0.97	3.41 ± 2.01	0.03
Cardiac Output (L/min)	3.68 ± 0.43	3.86 ± 0.47	4.24 ± 0.42	4.31 ± 0.46	4.60 ± 0.85	0.02
Heart Rate (beats/min)	85.8 ± 12.4	86.8 ± 11.6	91.3 ± 14.2	91.2 ± 13.1	107 ± 16.3	0.001
MAP (mmHg)	56.2 ± 13.6	52.5 ± 11.3	51.3 ± 11.3	51.8 ± 9.20	48.0 ± 10.5	N.S.
SBP (mmHg)	79.0 ± 14.1	74.5 ± 10.3	75.8 ± 10.2	74.8 ± 7.52	68.0 ± 16.4	N.S.
DBP (mmHg)	43.3 ± 12.4	39.8 ± 10.9	41.8 ± 12.3	38.5 ± 9.73	35.8 ± 9.02	N.S.
Oxygenation/anesthesia parameters, mean ± SD
PaO2 (mmHg)	107 ± 14.8	422 ± 37.4	235 ± 22.8	98.4 ± 25.4	49.2 ± 11.0	<0.001
SaO2 (%)	98.5 ± 0.8	100 ± 0	100 ± 0	98.0 ± 1.4	87.3 ± 7.6	<0.001
PvO2 (mmHg)	37.0 ± 4.9	48.8 ± 4.1	40.0 ± 3.6	35.3 ± 6.5	26.8 ± 4.0	<0.001
SvO2 (%)	74.2 ± 7.7	85.2 ± 3.1	78.3 ± 4.1	71.8 ± 7.6	56.0 ± 10.8	<0.001
FiO2 (%)	22.5 ± 1.1	96.3 ± 2.7	50.8 ± 3.3	21.2 ± 2.1	14.5 ± 1.9	<0.001
Isoflurane dose (MAC)	1.80 ± 0.20	1.90 ± 0.13	1.92 ± 0.12	1.87 ± 0.16	1.78 ± 0.16	N.S.

The P-values in the far right column show groupwise comparisons using Friedman test to test for differences across experimental states. Abbreviations: DBP – Diastolic Blood Pressure. FiO₂ – Fraction of Inspired Oxygen. L/min – Liters per Minute. MAP – Mean Arterial Pressure. ml/kg – Milliliters per Kilogram. mmHg – Millimeters Mercury. MAC – Minimum Alveolar Concentration. PaO₂ – Partial pressure of Oxygen in Arterial Blood. SaO₂ – Oxygen Saturation in Arterial Blood. SBP – Systolic Blood Pressure. SvO₂ – Oxygen Saturation in Venous Blood. WU – Woods Units.

Healthy Volunteer Imaging

Lung PSE during room air and oxygen inhalation in healthy volunteers is shown in Figure 6. Absolute T1 values are provided in Table 3. Inhalation of 100% oxygen caused signal intensity increase of 10.10 ± 3.05% in T1w images (P<0.001). By comparison, there was a smaller signal intensity increase of 1.99 ± 1.79% in PDw images during 100% oxygen inhalation (P=0.02), attributed to pulmonary vasodilation, which was reversed when returning to room air (P>0.05). These findings suggest that inhalation of 100% oxygen caused mean increase in proton density of 2.0% in healthy volunteers, which was reversible upon return to room air. This increase in proton density could be expected to influence T1w image signal intensity linearly given a constant T1. Our measured PSE values were confirmed by simulation over a range of proton density changes (Supplemental Digital Content 3).

Figure 6:

The median percent signal enhancement (PSE) in healthy volunteers, averaged across two acquisitions per experimental state in each subject. Results are provided for proton density weighted (PDw) images (red bars) and T1-weighted (T1w) images (green bars).

Figure 7 shows percent ΔT1 measured in MOLLI T1 mapping images in the lungs, left ventricular blood pool and right ventricular blood pool of healthy volunteers. Again, T1 shortened with oxygen inhalation in the lungs (−5.42 ± 6.47 %) and the left ventricular blood pool (−12.50 ± 4.93 %) but not right ventricular blood pool (−1.14 ± 3.68 %), which was reversed upon return to room air. There was no significant change in mean cardiac output during inhalation of 100% oxygen (change in cardiac output = −3.47 ± 5.90 %, p>0.05 versus both baseline and repeat room air).

Figure 7:

The percentage change in measured T1 from baseline room air, in healthy volunteers, in the lungs (gray bars), left ventricular (LV) blood pool (green bars), and the right ventricular (RV) blood pool (blue bars).

Missing and excluded data

Excluded and missing data are described in a flowchart in Supplemental Digital Content 1 and 2.

Discussion

The major finding of this study was that the contribution of lung proton density increase induced by pulmonary hyperoxic vasodilation to OE-MRI signal was 2.0 ± 1.8 percentage points in healthy volunteers, which is measurable but modest in magnitude relative to the OE-MRI signal attributable to T1-shortening during hyperoxia. The relative contribution of pulmonary hyperoxic vasodilation to OE-MRI had not previously been characterized.

The notion that PDw enhancement is sensitive to changes in pulmonary blood volume was supported by a small but consistent negative PDw PSE during IVC occlusion and a dose-dependent positive PDw PSE after infusion of hetastarch. We chose to use a colloid fluid for the volume challenge because of its osmotic propensity to remain in the blood vessels compared to saline, thereby decreasing rapid diffusion into the interstitium. Because intravenous fluids have a longer T1 than blood, there was a discrepancy in T1w and PDw enhancement after hetastarch while breathing room air. With the addition of 100% oxygen post 20 ml/kg of hetastarch infusion, lung T1 returned close to baseline. Under these conditions, there was a similar positive PSE in both PDw and T1w enhancement. As swine have been previously estimated to have a blood volume of approximately 65 ml/kg¹⁸, the hetastarch doses infused in this study would correspond to ~7.5% and ~30% of total blood volume, respectively. Importantly, this increase would be more than the anticipated pulmonary blood volume increase following vasodilation from 100% FiO₂ used in the context of OE-MRI. Still, our results demonstrate that our images are sensitive to the pulmonary blood volume increase associated with hyperoxia which generated ~2% PDw PSE in our study.

We found no evidence of incremental hyperoxic vasodilation above 50% oxygen, as measured by either PVR in anesthetized swine or by PSE in PDw images compared to baseline room air in anesthetized swine. This finding is in line with previous studies which suggest that the vasodilation response curve to oxygen is relatively flat in the hyperoxic interval.^19,20 In a clinical setting of uneven pulmonary oxygenation, e.g. asthma^3,7,21, the addition of 100% oxygen may potentially cause sizable local pulmonary blood volume change as individual lung regions move along the steep hypoxic range of the oxygen-vasodilation stimulus-response curve.

When measuring PSE in our control repeat room air state in anesthetized swine, we observed a substantial spread in the data. This suggests that the reproducibility over several hours was somewhat limited, which may be caused by prolonged anesthesia. In healthy volunteers, the increase in PDw PSE associated with hyperoxia was reversed upon return to room air, although the spread in inter-subject PSE was again larger in repeat room air compared to PSE between consecutive baseline room air acquisitions.

We utilized a high-performance prototype 0.55T MRI system in order to take advantage of the increased oxygen T1-relaxivity at lower field strengths compared to 1.5T and higher.¹² At higher clinical field strengths where T1-relaxivity is lower, the contribution of signal increase from pulmonary vasodilation may be larger relative to oxygen-induced T1-shortening.

In the present study, we used a free-breathing UTE sequence, as opposed to prior studies using a breath-held approach.^10,11 A free-breathing approach should be less susceptible to inconsistent respiratory position from repeated breath holds and the associated misregistration and improves the clinical applicability in patients with reduced capabilities of holding their breath. We rejected data with large changes in lung volume between experimental states. In a previous study, an accepted interval of 0.9 ≤ mean ($\mathrm{d}\mathrm{e}\mathrm{t}\left(J_T\right)$) ≥ 1.1 has been used, indicating lung volume change of ≤10 % compared to baseline.¹⁰ In order to minimize data rejection while, we used a 15% threshold. OE-MRI has also been previously attempted using T1-mapping, which is insensitive to proton density, but is associated with large measurement variability in the lungs², which may be related to the effect of respiratory phase on T1²².

Pulmonary-specific vasodilatory agents, such as nitric oxide, could be considered to replicate hyperoxic pulmonary vasodilation without the T1-shortening effect of oxygen. Systemic vasodilators are discouraged as these may generate an unpredictable balance between pulmonary and systemic vasodilation. The latter may shift fluid away from the pulmonary circulation, analogous to the acute effect of furosemide in routine treatment of pulmonary edema.

Clinical implications

Clinical application of T1w OE-MRI could benefit from concurrent acquisition of PDw images to isolate signal enhancement related to change in pulmonary blood volume. This information could be clinically meaningful in patients with regional hypoxic vasoconstriction sensitive to induced hyperoxia, such as asthma^3,7,21. For example, inhalation of oxygen may cause previously hypoxic (and therefore vasoconstricted) lung regions to undergo rapid dilation, creating a local increase in blood volume relative to baseline room air. This principle could also potentially be used to reduce variability in OE-MRI images and thereby mitigate an obstacle to clinical adoption.

In the present study, we only used PDw images to assess the global contribution of pulmonary blood volume changes to the OE-MRI signal. The added-value of PDw images requires further experimental validation in animal models with induced regional hypoxia, and in patient cohorts.

Limitations

The study had some important limitations. Our small sample size prohibited adequate statistical hypothesis-testing in many experimental states. In particular, the hypoxia state was associated with large lung volume changes which may have caused poor image registration quality. Physiological changes may have occurred in between PDw and T1w acquisitions, particularly in less stable experimental states such as hypoxia and IVC occlusion, and conclusions from experimental provocations in swine may not be generalizable to humans.

Conclusions

Our results suggest that T1-weighted oxygen-induced signal enhancement is affected by blood volume changes in the lungs that can be measured using proton density weighted signal enhancement. This effect is small but measurable in waking healthy volunteers. These findings suggest the future potential for using proton density weighted images to control for the effect of vasodilation-related confounding factors that may affect oxygen enhancement signal.

Abbreviations:

Det(J_T)

Jacobian determinant of the deformation field

FiO₂

Fraction of Inspired Oxygen

IVC

Inferior Vena Cava

MOLLI

Modified Look Locker Inversion Recovery

mPAP

Mean Pulmonary Arterial Pressure

OE-MRI

Oxygen Enhanced Magnetic Resonance Imaging

PDw

Proton Density Weighted

PSE

Percent Signal Enhancement

PVR

Pulmonary Vascular Resistance

SaO₂

Arterial Oxygen Saturation

T1w

T1 Weighted

UTE

Ultrashort Echo Time