Authors: James T. Grist BSc PhD1,2,3,4, Guilhem J. Collier PhD5, Huw Walters MBBS 1,
Mitchell Chen BMBCh MEng DPhil FRCR1, Gabriele Abu Eid BSc1, Aviana Laws1, Violet
Matthews BSc1, Kenneth Jacob BSc1, Susan Cross BSc1, Alexandra Eves BSc1, Marianne
Durant BSc1, Anthony Mcintyre BAppSci1, Roger Thompson PhD6, Rolf F. Schulte PhD7,
Betty Raman MBBS DPhil3, Peter A. Robbins PhD2, Jim M. Wild PhD MSc MA5, Emily Fraser
PhD MBChB BSc8, Fergus Gleeson MBBS1,9.
1 Department of Radiology, Oxford University Hospitals NHS Trust, Oxford, UK
2 Department of Physiology, Anatomy, and Genetics, University of Oxford, UK
3 Radcliffe Department of Medicine, Oxford Centre for Clinical Magnetic Resonance
Research, University of Oxford, UK
4 Institute of Cancer and Genomic Sciences, University of Birmingham, Birmingham, UK
5 POLARIS, Department of Infection Immunity and Cardiovascular Disease, University of
6 Department of Infection, Immunity, and Cardiovascular Disease, University of Sheffield
7 GE Healthcare, Munich, Germany
8 Oxford Interstitial Lung Disease Service, Oxford University Hospitals NHS Trust, Oxford, UK
9 Department of Oncology, University of Oxford, Oxford
Professor Fergus Gleeson, Department of Oncology, University of Oxford.
This work was funded by the Oxford NIHR Biomedical Research Centre, the National
Consortium of Intelligent Medical Imaging and the National Institute of Health Research,
British Heart Foundation Oxford Centre of Research Excellence. We would like to thank CMORE
research team for their support of this study. The Sheffield collaborators are
supported by MRC/ MR/M008894/1.
Long-COVID is an umbrella term used to describe ongoing symptoms following COVID-19
infection after four weeks. Symptoms are wide-ranging but breathlessness is one of the
most common and can persist for months after the initial infection. Investigations including
Computed Tomography (CT), and physiological measurements (lung function tests) are
usually unremarkable. The mechanisms driving breathlessness remain unclear, and this may
be hindering the development of effective treatments.
Eleven non-hospitalised Long-COVID (NHLC, 4 male), 12 post-hospitalised COVID-19 (PHC,
10 male) patients were recruited from a Post-COVID Assessment clinic, and thirteen healthy
controls (6 female) were recruited to undergo Hyperpolarized Xenon Magnetic Resonance
Imaging (Hp-XeMRI). NHLC and PHC participants underwent contemporaneous CT, Hp-
XeMRI, lung function tests, 1-minute sit-to-stand test and breathlessness questionnaires.
Statistical analysis included group and pair-wise comparisons between patients and
controls, and correlations between patient clinical and imaging data.
NHLC and PHC patients were 287 ± 79 [range 190-437] and 149 ± 68 [range 68-269] days
from infection, respectively. All NHLC patients had normal CT scans, and the PHC had
normal or near normal CT scans (0.3/25 ± 0.6 [range 0-2] and 7/25 ± 5 [range 4-8],
respectively). There was a significant difference in TLco (%) between NHLC and PHC patients
(76 ± 8 % vs 86 ± 8%, respectively, p = 0.04) but no differences in other measurements of
lung function. There were significant differences in RBC:TP mean between volunteers (0.45
± 0.07, range [0.33-0.55]) and PHC (0.31 ± 0.11, [range 0.16-0.37]) and NHLC (0.35 ± 0.09,
[range 0.26-0.58]) patients, but not between NHLC and PHC (p = 0.26).
There are RBC:TP abnormalities in NHLC and PHC patients, with NHLC patients also
demonstrating lower TLco than PHC patients despite their having normal CT scans. These
abnormalities are present many months after the initial infection.
Hyperpolarized Xenon MRI and TLco demonstrate significantly impaired gas transfer in nonhospitalised
long-COVID patients when all other investigations are normal.
1. There are significant differences in RBC:TP mean between healthy controls and
PHC/NHLC patients (0.45 ± 0.07, range [0.33-0.55], 0.31 ± 0.11, [range 0.16-0.37],
0.35 ± 0.09, [range 0.26-0.58], respectively, p < 0.05 after correction for multiple
comparisons) indicating a change in lung compartment volumes between groups.
2. There was a significant difference in TLco (%) between NHLC and PHC patients (76 ±
8 % vs 86 ± 8%, respectively, p = 0.04), despite normal or near normal FEV (%) (100 ±
13% [range 72-123%] and 88 ± 21% [range 62-113%], p>0.05.
3. There were significant differences in CT abnormalities between NHLC and PHC
patients (0.3/25 ± 0.6 [range 0-2] and 7/25 ± 5 [range 4-8], respectively) despite
similarly impaired RBC:TP.
On 11th March 2020, COVID-19 was declared a global pandemic by the World Health
Organisation (WHO). Beyond the acute respiratory manifestations of COVID-19 infection,
which can result in severe illness, hospitalisation and death, the medium and long-term
problems experienced by people following COVID-19 can be considerable(1). Large cohort
studies have revealed that symptoms can persist months after initial infection in both
patients hospitalised with COVID-19 pneumonia and those managed in the community. The
presence of ongoing symptoms related to prior COVID-19 infection is colloquially known as
Long-COVID. Although over 200 symptoms have been reported, the most common
problems are that of breathlessness, fatigue and brain fog. Long-COVID presents a global
health burden, with many people unable to return to normal activities or employment
months after becoming unwell.
The Chest X-ray (CXR) is the commonest imaging modality used in the diagnostic work-up of
acute COVID-19 pneumonia and this is often repeated three months after the acute
infection in patients requiring hospital admission. Computed Tomography (CT) may be
performed to investigate persistent breathlessness if the CXR is normal or there are other
concerns regarding COVID-19-related lung damage. In a small proportion of patients,
interstitial lung abnormalities persist and evidence of post-COVID fibrosis is
demonstrated(2). These abnormalities may account for dyspnoea, but in the majority of
individuals with Long-COVID, CT scans are normal or near normal. Similarly, lung function
tests are usually within the normal range. A recent study looking at a small cohort of posthospitalised
COVID-19 patients at 3 months, reported that hyperpolarised Xenon MRI (Hp-
XeMRI), was able to detect abnormalities of alveolar gas transfer even when the CT scans
and lung function tests were normal or near normal(3). HP-XeMRI enables the assessment
of ventilation and gas transfer across the alveolar epithelium into red blood cells. It provides
regional information of pulmonary vasculature integrity and may be able to identify lung
abnormalities not apparent on CT(4).
In Long-COVID, breathing pattern disorder is commonly identified and contributes to
breathlessness in a significant proportion of patients(5), but whether there are additional
reasons for their breathlessness, such as longer term pulmonary damage is currently
unclear. Specifically, whether the lung abnormalities on Hp-XeMRI in post-hospitalised
COVID-19 patients are present in non-hospitalised patients with Long-COVID has not been
previously evaluated and is the aim of this study.
Patient recruitment and screening
This study was approved by the HRA (REC reference 20/NW/0235), and all participants gave
informed consent. Participants were recruited from the Oxford Post-COVID Assessment
clinic, with the following inclusion criteria:
A) Post-hospitalised COVID (PHC) patients: PCR proof of SARS-CoV-2 infection; no history of
intubation; more than 3 months post discharge; no prior history of interstitial lung or
airways disease* or a smoking history >10 pack years; and a normal or near normal CT scan.
B) Non-hospitalised Long-COVID (NHLC) patients: PCR or positive antibody proof of SARSCoV-
2 infection; not hospitalised during acute infection; no evidence of interstitial lung or
airways disease*, or a smoking history >10 pack years; and a normal or near normal CT scan.
Diagnosis of Long-COVID made after referral to a specialist clinic with medically unexplained
dyspnoea as a symptom.
C) Healthy volunteers were recruited from the local staff pool at University of Sheffield and
the University of Oxford. Volunteers had to have no previous evidence of COVID-19
infection with PCR testing and no significant history of lung or cardiovascular disease or
*with the exception of mild well-controlled asthma with no evidence of airways obstruction
See Figure 1 for the flow diagram of this study.
Imaging protocol and physiological measurements
Imaging was performed at 1.5T (HDx, GE Healthcare, Chicago, IL) using a dedicated xenon
Transmit/Receive coil (CMRS, Brookfield, WI) and 1H images acquired on the body coil.
A 3D 4-echo flyback radial acquisition was used to acquire dissolved phase hyperpolarized
129Xe Images as previously described(6). Sequence parameters were:
acquired/reconstructed image resolution of 2/0.875 cm in all dimensions, repetition time
per spoke = 23ms, nominal flip angle per excitation on dissolved and gas phases = 40 and 0.7
degrees, respectively, total scan time = 16s. Gas, Tissue/Plasma (TP), and Red Blood Cell
(RBC) images were reconstructed.
The noise level for each image was calculated on a slice-by-slice basis. Any voxels in
the TP mask in each slice that were less than 5 times the median noise level in the slice were
discarded. Ratiometric maps (RBC:TP) were then calculated on a voxel-by-voxel basis. The
mean, standard deviation, and coefficient of variation of each ratiometric map was then
calculated on a patient-by-patient basis.
All trial participants underwent a contemporaneous low dose Computed Tomography (CT)
scan (GE Healthcare, Chicago, IL) following inspiration of 1L of room air, with slice thickness
of 0.625mm. Images were reviewed by a radiologist, blinded to clinical data and Xenon
results, as previously described(3).
They also underwent spirometry, gas transfer and Dyspnoea-12 score and 1 minute sit-tostand
test (STST). The number of repetitions were recorded, alongside the modified Borg
(mBorg) score and oxygen saturations pre and post a one-minute sit-to-stand test.
Initial analysis was performed on each participant cohort independently with correlation
between variables assessed using Spearman’s correlation, with a subsequent linear fit
performed for significantly correlated variables.
Participant data was separated into non-hospitalised Long-COVID (NHLC) and posthospitalized
COVID (PHC) groups and the above analysis re-performed for group-dependent
associations with clinical symptoms.
Comparisons between RBC:TP in patients and volunteer groups was assessed using a nonparametric
ANOVA and Tukey post hoc tests with Bonferroni correction for multiple
comparisons. A p value of < 0.05 was assumed for statistical significance. All analysis was
performed in R (The R Project).
A total of 11 NHLC (7 female) and 12 PHC (2 female) participants were recruited, with a
mean age of 43 ± 11 and 57 ± 12 years (p = 0.05), respectively. CT, proton, and fused RBC:TP
and proton images from participants with NHLC and PHC are shown in Figures 2 and 3,
respectively. Thirteen healthy volunteers (mean age: 41 ± 11 years, 6 female) were recruited
and underwent Hp-XeMRI. Example proton and fused RBC:TP and proton images for a
volunteer in this study are shown in Figure 4. The mean time from infection for the NHLC
and PHC participants was 287 ± 79 [range 190-437] and 149 ± 68 [range 68-269] days (p <
Average Hb for NHLC and HLC was 144 ± 15, range [122-166] and 145 ± 14, range [130-167],
respectively. NHLC and PHC participants exhibited breathlessness with a mean Dyspnoea-12
score of 9 ± 5 [range 0-21] and 10 ± 5 [range 1-15] (p = 0.67) and mBORG pre- and post-sit
stand test of 2 ± 2 [range 0-5] and 7 ± 1 [range 4-8] and 2 ± 2 [range 0-5] and 5 ± 1 [range 2-
8], respectively (p > 0.05 in all cases). The majority (9/11 and 4/5) of NHLC and PHC
participants were in the bottom 2.5th percentile for the number of repetitions they could do
for the mBORG sit-stand test, range 2.5-75 and range 2.5-25, respectively(7). There were
significant differences between NHLC and the PHC participants’ CT score (0.3 ± 0.6 [range 0-
2] and 7 ± 5 [range 0-15] respectively, p < 0.01). All data are presented in tables 1 and 2, for
NHLC and PHC participants, respectively.
Lung function and Imaging results
The mean, standard deviation, and coefficient of variation of NHLC, PHC participants, and
healthy volunteers RBC:TP are shown in Figure 5. There were significant differences in
RBC:TP mean between volunteers (0.46 ± 0.07, range [0.33-0.55]) and PHC (0.31 ± 0.11,
[range 0.16-0.37]) and NHLC (0.35 ± 0.09, [range 0.26-0.58]) patients (adjusted p<0.05 in all
cases), however not between PHC and NHLC (p = 0.29). Of note, 7/11 of NHLC and 11/12 of
PHC patients were beyond 2 standard deviations of the mean from normal volunteer mean
There was no significant difference in mean percent Forced Expiratory Volume (FEV), 100 ±
13% [range 72-123%] and 88 ± 21% [range 62-113%] (p>0.05), or FVC (NEED STATS) but
there was a significant difference mean gas transfer (TLco), 78 ± 8% [range 66-89%] vs 86 ±
9% [range 71-100%] (p = 0.04) between NHLC and PHC participants respectively.
In NHLC participants, there was a significant correlation between TLco (%) and RBC:TP
standard deviation (cc = 0.78, p = 0.02), RBC:TP mean and RBC:TP standard deviation (cc =
0.63, p = 0.05). Further correlations between RBC:TP and Dyspnoea-12 score and RBC:TP
and mBORG post-sit stand were close to significance (p = 0.06 and 0.08, respectively). All
other correlation results are shown in Tables 1 and 2. A linear model between RBC:TP mean
and RBC:TP standard deviation in Supplementary Figure 1A, and TLco (%) and RBC:TP
standard deviation in Figure 6A.
The significant correlations in the PHC participants were between patient age and dissolved
phase mean (cc = -0.82, p < 0.01, Supplementary Figure 1B), CT score and RBC:TP standard
deviation (cc = 0.54, p = 0.04, Figure 6B), RBC:TP mean and RBC:TP standard deviation (cc =
0.76, p = 0.03, Supplementary Figure 1C), and RBC:TP mean and RBC:TP coefficient of
variation (cc = -0.73, p = 0.04, Supplementary Figure 1D).
This pilot study utilised Hp-XeMRI to evaluate the lungs of NHLC patients with unexplained
breathlessness following clinical evaluation in a dedicated post-COVID clinic. The findings
were compared to PHC with normal or near normal CT scans, and to age and gender
matched healthy controls. We found that the Hp-XeMRI results were abnormal in the
majority (7/11) of NHLC patients, indicating that pulmonary gas transfer is impaired many
months and in some, over a year after their initial infection. Importantly, despite having
completely normal CT scans with no evidence of prior pneumonia, there were no discernible
differences in the type of Hp-XeMRI abnormality detected between the NHLC and the PHC
All the NHLC participants in this study with abnormal Hp-XeMRI were imaged more than 6
months after their initial infection, indicating that these abnormalities are not a transient
phenomenon following acute infection. The NHLC patients were also on average further
from their initial infection than the PHC group (287 v 149 days). Interestingly, the measured
abnormality on Hp-XeMRI appears to be only marginally greater in the PHC than the NHLC
patients despite those admitted to hospital having had a presumed clinically more severe
The participants in this study were well phenotyped with symptoms typical of nonhospitalised
Long-COVID patients who did not require hospital admission(8). The
relationship of the Hp-XeMRI abnormalities detected and the breathlessness experienced by
the wider population of Long-COVID patients, managed both in hospital and in the
community during their acute infection is unclear. Additionally, the pathophysiological
mechanisms that underlie the changes in Hp-XeMRI post COVID-19 infection remain to be
fully elucidated. Although, it is possible to make some inferences regarding the nature of
the underlying defect based on our results. It is known that inert gases (those that do not
chemically react with blood) equilibrate rapidly in the lung(9), with Xenon reaching the red
blood cells within 30-50 milliseconds(10). RBC:TP is a ratio of two tissue volumes (the
pulmonary capillary (plus potentially some pulmonary venous) blood volume to the alveolar
membrane volume) measured using Hp-XeMRI. A lower figure suggesting that infection with
Sars-CoV-2 may have induced some microstructural abnormality to either one or both
volumes, causing a reduction in blood volume for example due to widespread
microclots(11) and/or a thickening of the alveolar membrane, both of which would be
expected to cause a reduction in diffusing capacity(12).
It is potentially possible that in patients hospitalised with COVID-19 pneumonia, our PHC
group, that the direct damage to the lungs caused by the virus and resultant inflammatory
sequelae may cause longer-lasting microstructural abnormalities. Indeed, although the CT
scans were normal or near normal in the PHC patients, a faint ‘footprint’ of prior COVID-19
pneumonia, when present, may possibly at least partially explain the abnormal RBC:TP and
pulmonary gas transfer values. In contrast, in the NHLC patients, all the CT scans were
normal and none of the participants had evidence of having had pneumonia (accepting that
this may have been because they were not imaged during their acute infection). This could
potentially indicate that the abnormalities detected in the NHLC cohort have a different
pathophysiological basis. Furthermore, the gas transfer (TLCO), which also provides a
measure of pulmonary vascular integrity, was lower in the NHLC group than the PHC group
and correlates with the RBC:TP ratio, reinforcing the significance of the findings and the
need for further investigation to delineate the nature of the abnormality.
Outside the setting of SARS-CoV2 infection it is known from prior studies with Hp-XeMRI in
patients with CT diagnosed interstitial lung disease that more severe disease determined by
TLCO correlates with worsening RBC:TP and that Hp-XeMRI may identify lung abnormality in
areas that are normal on CT(13). Hp-XeMRI appears to also be a more sensitive way of
detecting disease than CT in patients with Long COVID and may be a useful tool in its
diagnosis, quantification and follow-up. But caution is necessary as it is unknown whether
patients with other respiratory tract infections such as flu have abnormal Hp-XeMRI gas
transfer months after infection even when non-hospitalised and with a normal CT. It is also
not known whether the abnormalities we have detected are of clinical significance, nor
whether Hp-XeMRI is an over-sensitive test, although the correlation with TLCO argues
To better understand the significance of our findings, our study has now expanded to recruit
a larger cohort of patients that includes NHLC patients without significant breathlessness
alongside participants with prior proven COVID-19 infection who have fully recovered. We
will also be performing repeat imaging at different time intervals up to 12 months to
determine whether the abnormalities detected persist or resolve over time.
In conclusion, Hp-XeMRI has identified objective impairment in gas transfer in the lungs of
non-hospitalised breathless Long-COVID patients with normal CT scans, providing
preliminary evidence that lung abnormalities exist that cannot be detected with
conventional imaging. The significance and underlying pathophysiology of this abnormality
is currently unknown and highlights the need for further research in this field.
1. Huang C, Huang L, Wang Y, et al. 6-month consequences of COVID-19 in patients
discharged from hospital: a cohort study. Lancet. Elsevier Ltd; 2021;397(10270):220–
232. doi: 10.1016/S0140-6736(20)32656-8.
2. Han X, Fan Y, Alwalid O, et al. Six-Month Follow-up Chest CT findings after Severe
COVID-19 Pneumonia Manuscript type: Original Research. (1).
3. Grist JT, Chen M, Collier GJ, et al. Hyperpolarized 129Xe MRI Abnormalities in
Dyspneic Participants 3 Months after COVID-19 Pneumonia: Preliminary Results.
4. Qing K, Ruppert K, Jiang Y, et al. Regional mapping of gas uptake by blood and tissue
in the human lung using hyperpolarized xenon-129. J Magn Reson Imaging.
2014;39(2):346–359. doi: 10.1002/jmri.24181.Regional.
5. Motiejunaite J, Balagny P, Arnoult F, et al. Hyperventilation as one of the mechanisms
of persistent dyspnoea in SARS-CoV-2 survivors. Eur Respir J. 2021;58(2). doi:
6. Collier GJ, Eaden JA, Hughes PJC, et al. Dissolved 129Xe lung MRI with four-echo 3D
radial spectroscopic imaging: Quantification of regional gas transfer in idiopathic
pulmonary fibrosis. Magn Reson Med. 2020;(October):1–12. doi:
7. Strassmann A, Steurer-Stey C, Lana KD, et al. Population-based reference values for
the 1-min sit-to-stand test. Int J Public Health. 2013;58(6):949–953. doi:
8. NICE, RCGP and SIGN publish guideline on managing the long-term effects of COVID19.
9. Piiper J, Scheid P. Blood-gas equilibration in lungs. In: Pulmonary Gas Exchange.
Academic Press Inc; 1980.
10. Hyppönen V, Stenroos P, Nivajärvi R, et al. Metabolism of hyperpolarised [1– 13
C]pyruvate in awake and anaesthetised rat brains . NMR Biomed.
2021;(September):1–10. doi: 10.1002/nbm.4635.
11. Pretorius E, Vlok M, Venter C, et al. Persistent clotting protein pathology in Long
COVID/Post-Acute Sequelae of COVID-19 (PASC) is accompanied by increased levels
of antiplasmin. Cardiovasc Diabetol. BioMed Central; 2021;20(1):1–18. doi:
12. Roughton F., Forster RE. Relative importance of diffusion and chemical reaction rates
in determining rate of exchange of gases in the human lung, with special reference to
true diffusing capacity of pulmonary membrane and volume of blood in the lung
capillaries. J Appl Physiol. 1957;11:290–302.
13. Mammarappallil J, Rankine L, Wild JM, Driehuys B. New Developments in Imaging
Idiopathic Pulmonary Fibrosis With Hyperpolarized Xenon Magnetic Resonance
Imaging. J Thorac Imaging. 2019;34(2):136–150.
Figure 1. The study flowchart
Figure 2. Example CT, proton, Gas, TP, and RBC imaging from Long-COVID patients. The top
row is a patient with RBC:TP = 0.49, the middle row is a patient with RBC:TP of 0.31, and the
bottom row is a patient with RBC:TP = 0.24.
Figure 3. Example CT, proton, Gas, TP, and RBC imaging from post-Hospitalized patients. The
top row is a patient with RBC:TP = 0.59, the middle row is a patient with RBC:TP of 0.31, and
the bottom row is a patient with RBC:TP = 0.16.
Figure 4. Example Proton (1H) and fused RBC:TP map of a healthy control in this study.
Figure 5 – Comparison of RBC:TP mean (A), standard deviation (SD) (B), and Coefficient of
Variation (C) between healthy, post-hospitalized COVID, and non-hospitalized Long-COVID
patients. * = significant after correction for multiple-comparisons.
Figure 6. Correlation results. A – A significant positive correlation between Tlco (%) and
RBC:TP Standard Deviation (STD) in the NHLC group. B – a significant positive correlation
between RBC:TP Standard Deviation (SD) and CT score in the PHC group.
Supplementary Figure 1 – a significant positive correlation between SBC:TP mean and
Standard Deviation (SD) in the NHLC group (A). Significant correlations between Age and
RBC:TP mean (B), RBC:TP Sd and TBC:TP mean (C), and RBC:TP Coefficient of Variation (CoV)
and RBC:TP mean (D) in the PHC group.
Table 1 – Data for each non-hospitalized Long-COVID patient, N/A = not acquired.
Table 2 – Data for each post-hospitalized COVID patient, N/A = not acquired.