Longitudinal Lung Function Assessment of Patients Hospitalized With COVID-19 Using 1H and 129Xe Lung MRI

Background Microvascular abnormalities and impaired gas transfer have been observed in patients with COVID-19. The progression of pulmonary changes in these patients remains unclear. Research Question Do patients hospitalized with COVID-19 without evidence of architectural distortion on structural imaging exhibit longitudinal improvements in lung function measured by using 1H and 129Xe MRI between 6 and 52 weeks following hospitalization? Study Design and Methods Patients who were hospitalized with COVID-19 pneumonia underwent a pulmonary 1H and 129Xe MRI protocol at 6, 12, 25, and 51 weeks following hospital admission in a prospective cohort study between November 2020 and February 2022. The imaging protocol was as follows: 1H ultra-short echo time, contrast-enhanced lung perfusion, 129Xe ventilation, 129Xe diffusion-weighted, and 129Xe spectroscopic imaging of gas exchange. Results Nine patients were recruited (age 57 ± 14 [median ± interquartile range] years; six of nine patients were male). Patients underwent MRI at 6 (n = 9), 12 (n = 9), 25 (n = 6), and 51 (n = 8) weeks following hospital admission. Patients with signs of interstitial lung damage were excluded. At 6 weeks, patients exhibited impaired 129Xe gas transfer (RBC to membrane fraction), but lung microstructure was not increased (apparent diffusion coefficient and mean acinar airway dimensions). Minor ventilation abnormalities present in four patients were largely resolved in the 6- to 25-week period. At 12 weeks, all patients with lung perfusion data (n = 6) showed an increase in both pulmonary blood volume and flow compared with 6 weeks, although this was not statistically significant. At 12 weeks, significant improvements in 129Xe gas transfer were observed compared with 6-week examinations; however, 129Xe gas transfer remained abnormally low at weeks 12, 25, and 51. Interpretation 129Xe gas transfer was impaired up to 1 year following hospitalization in patients who were hospitalized with COVID-19 pneumonia, without evidence of architectural distortion on structural imaging, whereas lung ventilation was normal at 52 weeks.

In patients hospitalized with pneumonia caused by infection with SARS-CoV-2, the existing literature and clinical experience suggest that there is considerable overlap in clinical presentation with typical pneumonia and ARDS in patients exhibiting hyperinflammation and progressive hypoxemia.2][3] Abnormal pulmonary vasoregulation has been observed in patients in the acute phase of COVID-19 1 and may be a pathophysiologic mechanism contributing to the progressive hypoxemia seen in these patients.
Abnormalities on chest radiograph or CT scan imaging at 12 weeks following hospitalization due to COVID-19 are present in some patients, particularly those with more severe disease who require ICU treatment. 4However, for patients without radiographic abnormalities, sensitive techniques for monitoring longitudinal change in lung function are needed.
[7][8][9][10][11] In addition, 129 Xe can image gas diffusion within the lung airspace (diffusion-weighted MRI [DW-MRI]), and the derived apparent diffusion coefficient (ADC) and mean diffusive length scale (Lm D ) provide threedimensional in vivo information of the underlying lung microstructure that is highly sensitive to changes in patients with emphysema 12 and fibrotic lung disease. 13n addition, 129 Xe is soluble in the interstitium/ membrane (M) and in the RBCs, and the signal from 129 Xe in these dissolved compartments can be distinguished spectroscopically.The ratio of the 129 Xe MRI signal observed in the lung airspaces (gas), the lung M, and bound to the RBCs can thus be determined with magnetic resonance spectroscopic imaging.7][18] RBC:M has been shown to correlate highly with the transfer factor for carbon monoxide (TL CO ). 6,19ke-home Points Study Question: Do patients hospitalized due to COVID-19 with no evidence of architectural distortion exhibit longitudinal improvements in 129 Xe gas transfer to within a normal range between 6 and 52 weeks following hospitalization?Results: At 12 weeks, significant improvements in 129 Xe gas transfer were observed compared with 6-week examinations.However, 129 Xe gas transfer remained abnormally low at weeks 12, 25, and 51.Interpretation: In a cohort of patients with moderate severity disease, 129 Xe gas transfer improved but did not return to within a normal range within 1 year following hospitalization.T., and J. M. W.), University of Sheffield, Sheffield, England; Sheffield Teaching Hospitals (J.G. R. W., J. E. M., Z. G., T. N., M. P., P. W., S. S., L. G., S. R., L. W., and R. L.), NHS Foundation Trust, Sheffield, England; National Heart and Lung Institute (A.L. and G. J.), Imperial College London, London, England; Department of Oncology (J.G. and F. G.), Oxford NHS Foundation Trust, Oxford, England; Oxford Centre for Clinical Magnetic Resonance Research (J.T. G.), University of Oxford, Oxford, England; Department of Physiology, Anatomy, and Genetics (J.T. G), University of Oxford, Oxford, England; Department Previous studies have reported reduced RBC:M in patients with COVID-19, [20][21][22][23] including in patients with normal chest CT scan imaging but ongoing dyspnea. 21In patients with residual lung abnormalities on CT scans, decreased RBC:M may be due to an increase of xenon uptake in the interstitial lung tissue.However, in the absence of CT scan abnormalities, we propose that a decreased RBC:M instead indicates microvascular (capillary) abnormalities.Therefore, RBC:M in particular may be a sensitive metric suitable for longitudinal assessment of regional gas exchange abnormalities in patients who have had COVID-19 and have normal structural imaging.It is currently unknown whether RBC:M improves longitudinally following COVID-19 pneumonia.Age-related reductions in the RBC:M ratio may be relevant in other cohorts, 24 and control cohorts well-matched for age are therefore needed for accurate interpretation of RBC:M in post-COVID-19 studies.It is also unclear whether patients with abnormal RBC:M have concurrent abnormalities in lung perfusion or ventilation.1 H lung MRI is able to assess changes in lung structure and perfusion.Ultra-short echo time (UTE) imaging enables good visualization of the lung parenchyma and has shown excellent agreement with CT scan imaging in the visualization of lesions in patients with COVID-19. 25Dynamic contrastenhanced (DCE) 1 H lung MRI allows the assessment of lung perfusion, with high sensitivity and specificity in detecting perfusion defects without exposing the participant to ionizing radiation, 26 and it is therefore well suited for patient follow-up studies.Increased lung perfusion transit times (time to peak) have been reported in both an acute hospitalized patient with COVID-19 and in nonhospitalized male patients with breathlessness who have had COVID-19. 27,28e current study used a 1 H and 129 Xe MRI protocol that combines hyperpolarized 129 Xe imaging methods sensitive to ventilation, lung microstructure (DW-MRI), and gas exchange (dissolved xenon spectroscopic imaging) alongside 1 H DCE perfusion and UTE lung structural imaging to assess pathophysiologic changes in patients who had been hospitalized with COVID-19 pneumonia during the postacute period.The primary hypothesis of this work was that abnormal imaging and pulmonary function test (PFT) markers of lung function would increase to within a normal range over the course of 1 year in patients without structural abnormalities seen on CT scan or proton structural imaging.Patients underwent up to four follow-up MRI examinations at approximately 6, 12, 24, and 52 weeks following hospitalization.

Participants
Patients with acute COVID-19 pneumonia and no previously diagnosed respiratory disease (excluding mild asthma) were recruited from Sheffield Teaching Hospital's pulmonology and infectious diseases wards from November 2020 to February 2022 for this prospective cohort study, prior to or shortly following discharge.Follow-up 129  Patients with evidence of interstitial lung damage at 12 weeks following hospital admission were recruited into the parallel UK Interstitial Lung Disease Consortium (UKILD) study. 29Standard MRI exclusion criteria were applied to all subjects.In addition, patients were excluded if they were unable to tolerate a test inhalation of 129 Xe gas according to the supervising clinicians' judgment or if they had a chest size exceeding the 129 Xe chest coil circumference (76 cm).
Where possible, PFTs were acquired on the same day as the MRI examination at each visit.Spirometry and transfer factor were performed, and from these tests, the metrics FEV 1 , FVC, FEV 1 /FVC, TL CO , and carbon monoxide transfer coefficient were calculated and presented as z scores and % predicted using Global Lung Function Initiative reference ranges. 30,31This study was approved by the London-Hampstead Research Ethics Committee (REC reference: 9/LO/1115).

MRI Acquisition
Patients underwent scanning on either an HDx 1.5T (N ¼ 7) or a 450w 1.5T (N ¼ 2) (GE Healthcare) MRI scanner. 32The 129 Xe images were acquired with the patient in a flexible quadrature transmit/receive vest coil (Clinical MR Solutions).Patients' vital signs were monitored throughout the MRI examination.Each patient underwent MRI examinations on the same scanner for baseline visits and follow-up.Figure 1 presents an illustrative diagram of the lung MRI methods used in this study.129 Xe doses were polarized to approximately 30% using a home-built high-performance spin-exchange optical pumping polarizer. 32This had regulatory approval for manufacture of hyperpolarized 129 Xe for clinical lung MRI by the UK Medicines and Healthcare Products Regulatory Agency.
MR imaging was conducted as summarized in the following text (parameters are detailed in e-Table 1).
A structural 1 H scan was acquired following inhalation of a bag of air to match the lung inflation state of the subsequent xenon sequences. 129Xe ventilation images were acquired using a three-dimensional imaging sequence with whole-lung coverage following inhalation of a 1 L maximum mixture of 129 Xe and nitrogen (titrated if subject height < 160 cm 33 ) and inhaled from functional residual capacity; patients were coached in the required breathing maneuver prior to their MRI examination. 34 129 X DW-MRI to assess alveolar microstructural change was acquired following inhalation of a maximum 1 L mixture of 129 Xe and nitrogen (three-dimensional spoiled gradient echo [SPGR] multiple b-value sequence with compressed sensing with wholelung coverage).35 Three-dimensional spectroscopic imaging of the gas and dissolved phase xenon resonances (dissolved xenon in lung M and in blood RBCs) was acquired by using a maximum 1 L dose of hyperpolarized 129 Xe (repetition time ¼ 15 milliseconds, flip angle ¼ 22 [threedimensional acquisition with whole-lung coverage]).6 1 H MRI was acquired by using an eight-element cardiac array (GE Healthcare).UTE images were acquired with a three-dimensional radial sequence during 8 min of free-breathing with prospective respiratory bellows gating on expiration. 36 hree-dimensional variable flip angle SPGR images 37,38 were acquired (flip angle ¼ 2 , 4 , 10 , and 30 ) to allow for the correction of lung T 1 and proton density.DCE lung perfusion MRI was acquired (threedimensional volumetric time-resolved SPGR acquisition).A half dose (0.05 mL/kg) of Gadovist (Bayer) was administered at an injection rate of 4 mL/s followed by a 20 mL saline flush at 4 mL/s.Patients were advised to hold their breath for as long as possible and breathe shallowly thereafter.

A B
Dissolved phase 129 Xe MRI Dissolved phase MRI measures changes in the proportion of 129 Xe in the airspace (gas), membrane (M), and red blood cells (RBC).

Lung perfusion
Measures concentration of gadolinium-based contrast agent in the bloodstream. 129Xe Ventilation Measures density and distribution of 129 Xe in the airspace.

Contrast-enhanced lung perfusion
Decrease in microvascular perfusion.

Image Analysis
Qualitative assessments of the UTE 1 H structural, 129 Xe ventilation, and DCE lung perfusion images were made by two radiologists with 10 and 14 years of experience, respectively.UTE images were assessed for parenchymal changes, and ventilation and perfusion images were assessed for defects.
Metrics of ventilation defect percentage (VDP), low ventilation percentage (VP), normal VP, and hyper VP for each patient were calculated by using linear binning (see the Methods section of e-Appendix 1).The coefficient of variation of the segmented lung ventilation images was also calculated from the 129 Xe ventilation images as a marker of ventilation heterogeneity.
Maps of 129 Xe ADC and Lm D from a stretched exponential model of 129 Xe gas diffusion in the lungs were calculated on a voxel-by-voxel basis. 39ps of gas transfer ratios (RBC:M, RBC:gas, and M:gas) were calculated from three-dimensional spectroscopic imaging.The transverse relaxation time (T 2 *) of the RBC and M spectroscopic peaks was also calculated.
Mean values of all global metrics were calculated for each patient.A sample size calculation was not performed because this was an exploratory study.
Global MRI metrics from visits 1, 2, 3, and 4 were compared by using a Skillings-Mack test due to the presence of missing data 40 with pairwise Wilcoxon tests and a correction for multiple testing, 41 implemented by using R software. 42Data are presented as median (range), unless otherwise stated.
Mixed linear effect models were set up using a random intercept model to test the relationship between RBC:M and the following: (1) pulmonary blood volume; (2) pulmonary blood flow; (3) mean transit time; (4) VDP; and (5) TL CO z score.IBM SPSS Statistics 27 (IBM SPSS Statistics, IBM Corporation) was used for analysis.A P value < .05 was considered statistically significant.

Age-and Sex-Matched Healthy Volunteer Metrics
Median ADC and Lm D values for an age-and sex-matched control cohort were determined by retrospective analysis of previously published data. 43Eleven subjects from this previously published work were selected based on matching median and interquartile range (IQR) of age and sex ratio from a cohort of 23 subjects while blinded to MRI metrics; the control cohort had a median age of 63 (40-70) years, and 73% were male.
Median RBC:M, RBC:gas, and M:gas for an age-and sex-matched control cohort were determined by retrospective analysis of a healthy cohort data set, with control subjects chosen based on matching median and IQR of age and sex ratio while blinded to MRI metrics.Twelve subjects were selected (median age, 57 [41-68] years), and 67% were male.

Results
Of the 16 recruited patients, 14 showed no signs of interstitial lung damage at 12 weeks and were therefore included as part of this study.Nine of 14 patients had follow-up examinations and were included for analysis (     1. 44,45 No patients received any trial of pharmacologic treatment for post-COVID-19 symptoms following discharge.Two of the patients commenced treatment for diabetes during the follow-up period. UTE and 129 Xe MRI were successfully acquired in all patients at all visits.DCE lung perfusion imaging was successfully acquired in six of nine patients at visit 1, eight of nine patients at visit 2, six of seven patients at visit 3, and five of eight patients at visit 4. The reasons for unsuccessful lung perfusion imaging were patients failing screening (6 visits), patient motion (2 visits), and technical issues (1 visit).
Quantitative metrics of ventilation improved at visits 2, 3, and 4 compared with visit 1; however, this was not statistically significant following adjustment for multiple corrections (Table 2).The T 2 * of the M and RBCs was calculated.M T 2 * showed a significant longitudinal decrease across visits, with lower M T 2 * at visit 4 compared with visits 1, 2, and 3 (P adj ¼ .023,P adj ¼ .023,and P adj ¼ .047,respectively) and between visits 1 and 3 (P adj ¼ .031)(Table 2).No other significant changes in the T 2 * of the RBC or M were seen (e-Fig 3).

H MRI
Structural Changes: The UTE image of Patient 3 showed abnormal linear parenchymal changes at visit 1, which improved but remained abnormal at visits 2 and 3 and were resolved at visit 4. Patients 2, 6, 7, and 8 displayed air trapping on their UTE image at visit 1, which resolved at visit 2 for patients 6, 7, and 8. Patient 2 continued to have air trapping present at visits 3 and 4.
The UTE images of Patients 1, 4, 5, and 9 were normal at all visits (e-Table 2).

DCE (Perfusion):
Patient 1 showed a segmental perfusion defect at visit 1 that was resolved at visit 2. No other patients showed any substantial regional perfusion defects.Median pulmonary blood volume and flow increased in all patients (n ¼ 6) at visit 2 compared with visit 1; however, the increase was not statistically significant.For the six patients with DCE MRI at visits 1 and 2, median pulmonary blood volume was 37.8 (11.7-53.5)mL/100 mL at visit 1 and 47.6 (15.0-60.2) mL/100 mL at visit 2, and pulmonary blood flow was

Pulmonary Function Tests
Data were available on PFTs for six of nine patients at visit 1, six of nine patients at visit 2, seven of seven patients at visit 3, and seven of eight patients at visit 4; all z scores and % predicted data are shown in Figure 7.
2][23] Because xenon gas transfer depends on both the xenon uptake in the lung tissue and the xenon uptake in the RBCs, a combination of lung perfusion abnormalities and/or alveolar/interstitial endothelial changes may be mechanistically driving the reduced xenon gas transfer seen in patients following COVID-19.1][22] In the current study, the inclusion of data from age-and sex-matched healthy control subjects shows that these changes are not due to age or sex differences between control subjects and patients in this study.RBC:gas and M:gas did not show significant longitudinal change once adjusted for multiple comparisons, implying that the change in RBC:M was a combined effect of changes in both M and RBC.A significant reduction in M T 2 * at visit 2 was also found.The physiological mechanisms behind changes in M T 2 * are not well established and are discussed further in the Discussion section of e-Appendix 1 We also found that changes in xenon gas transfer increased significantly with increased TL CO z score, VDP, and lung perfusion metrics (pulmonary blood volume and pulmonary blood flow).All patients with DCE data available displayed an increase in regional pulmonary blood flow and volume between visits 1 and 2, despite only one having a substantial perfusion defect.This may indicate microvascular improvements at 12 weeks, and that microvascular recovery may be partially driving changes in RBC:M in these patients.In parallel, a concomitant reduction in M signal due to resolution of postinfection endothelial inflammation could contribute to the increase in RBC:M with time.
Although we see global correlations between RBC:M, ventilation, and perfusion, regional heterogeneity in RBC:M did not visually agree with ventilation or perfusion heterogeneity; for example, as shown in Figure 3, Patient 8 has a visually heterogeneous RBC:M map but no visual concordance with pulmonary blood flow heterogeneity and homogeneous ventilation on the similar slices presented.Further work assessing regional distributions seen in the different functional magnetic resonance images available here is warranted to evaluate regional correlations quantitatively.
In this study, most patients (seven of nine) did not report significant breathlessness at visit 2 (12 weeks), despite lower RBC:M than the control reference data.  alterations in lung microstructure measured using 129 Xe MRI, 6 reductions in lung ventilation, 46 and reductions in lung perfusion. 47lthough this means that there is considerable promise for lung MRI to provide longitudinal biomarkers in patients with signs of interstitial lung damage, it also suggests that persistent perfusion, ventilation, gas transfer, and lung microstructure abnormalities may be mechanistically related to the visible tissue changes within a cohort with structural lung abnormalities.Further work using a 1 H and 129 Xe protocol in patients with established pulmonary fibrosis due to COVID-19 on CT scan imaging is the subject of an ongoing study (UKILD). 29nor ventilation heterogeneity and defects were present in this cohort shortly following acute illness; these defects improved over time, which is consistent with the findings of Grist et al 21 and of Li et al. 20 Overall, the current study and the findings from previous literature suggest it is unlikely that impaired lung ventilation is the primary cause of ongoing symptoms following the acute stage of COVID-19 and that the pathophysiology is not primarily of the airways.
The main limitation of the current study is the limited number of participants, which was largely caused by the challenging nature of recruiting patients for scanning directly following a recent hospitalization due to COVID-19 in the first wave of the pandemic.In addition, not all patients had DCE lung perfusion imaging or PFTs at all examinations (due to aerosolization constraints).The numbers recruited limit correlations with symptoms, activity, and lung function, as well as the statistical tests used to test for change.All 129 Xe acquisitions were acquired at functional residual capacity plus 1 L, resulting in some variability between patients in the lung inflation state.A final potential source of bias in this study is that five patients who were potentially eligible for the study were excluded due to chest size exceeding the size of the xenon MRI coil.

Interpretation
This study found that in a cohort of patients who were hospitalized with COVID-19 pneumonia of moderate severity who had normal CT scan/lung structural imaging, 129 Xe gas transfer improved at 12 weeks but did not return to within a normal range within 1 year following hospitalization.Improvements in 129 Xe gas transfer were associated with an increased lung perfusion on DCE-MRI and increased TL CO z score; therefore, abnormalities in 129 Xe gas transfer may be a marker of ongoing microvascular abnormalities post-COVID-19.
TL CO z score was within a normal range for seven of eight patients with available data at 51 weeks posthospitalization.This indicates that 129 Xe gas transfer may be a more sensitive measure of gas exchange in this population and that it may be able to identify abnormalities that routine clinical tests overlook.
We believe this to be the first follow-up study of similar patients with such an extensive range of functional lung imaging techniques.Our findings show the sensitivity and complementary nature of functional MRI to followup post-COVID-19 lung pathophysiology in a clinical setting.

Figure 1 -
Figure 1 -A-B, Illustrative diagram showing how the lung MRI techniques used in this article measure lung perfusion, ventilation, lung microstructure (diffusive length scale), and xenon gas transfer (the transfer of xenon between the airspace, membrane, and RBCs).A, Techniques in a healthy alveolus.B, Possible interpretation of the findings of this article in patients who have had COVID-19, with reduced RBC:M due to damage to pulmonary microcirculation but preserved acinar airway dimensions.RBC:M ¼ RBC to membrane fraction.

Figure 3 -
Figure 3 -Example of UTE images, RBC:M maps, 129 Xe ventilation images, and maps of pulmonary blood flow at visit 1 and visit 2, for each patient.The white arrow indicates a segmental perfusion defect visible at visit 1, which improves at visit 2. M ¼ membrane; PBF ¼ pulmonary blood flow; RBC:M ¼ RBC to membrane fraction; UTE ¼ ultra-short echo time.

4 Figure 4 -
Figure4-Spaghetti plots of ventilation, dissolved phase xenon, and dynamic contrast-enhanced lung perfusion metrics at visits 1 to 4. CV ¼ coefficient of variation of lung ventilation; M ¼ membrane; M:gas ¼ membrane to gas fraction; RBC:gas ¼ RBC to gas fraction; RBC:M ¼ RBC to membrane fraction; VDP ¼ ventilation defect percentage.

Figure 5 -Figure 6 -
Figure 5 -Lung RBC:M maps in three patients with four MRI visits at 6, 12, 25, and 51 weeks following hospital admission.Mean RBC:M at each visit is shown.M ¼ membrane; RBC:M ¼ RBC to membrane fraction.

TABLE 1 ]
Patient Demographic Data

TABLE 2 ]
Median Metrics for All MRI Parameters at Visits 1, 2, 3, and 4 The number of patients who had RBC:M below the median AE IQR of the age-and sex-matched healthy volunteers was eight of nine at visit 1, seven of nine at visit 2, five of seven at visit 3, and six of eight at visit 4.

TABLE 2 ]
Data are presented as median (range) of all patients with available data for each visit.If a Skillings-Mack test determined that there was a significant difference between at least two variables, chestjournal.org 5 interquartile range; star denotes data > 3 interquartile range.M ¼ membrane.;M:gas ¼ membrane to gas fraction; RBC:gas ¼ RBC to gas fraction; RBC:M ¼ RBC to membrane fraction.

TABLE 3 ]
35fect of Pulmonary Blood Volume, Pulmonary Blood Flow, Mean Transit Time, VDP, and TL CO z Score on RBC:M Tested Using Linear Mixed-Effect Model Analysis With a Random Intercept RBC to membrane fraction; TL CO ¼ transfer factor for carbon monoxide; VDP ¼ ventilation defect percentage.Median patient ADC and Lm D were within the ageand sex-matched reference range at visits 1 and 2,35with no significant change at visit 2, indicating that airway dimensions were not increased in these nine patients who had COVID-19 but no signs of interstitial lung damage on structural imaging.This study excluded patients with signs of interstitial lung damage, as previous work has shown that patients with interstitial lung diseases can have reduced xenon gas transfer,