Introduction
In December 2019, Wuhan City (Hubei Province, China) reported a febrile respiratory
tract illness of unknown origin in a cluster of patients. Bronchoalveolar lavage of
the patients isolated a novel strain of coronavirus (SARS-coronavirus-2 [SARS-CoV-2])
as the pathogen.
1
The pulmonary infection caused by SARS-CoV-2 was named coronavirus disease 2019 (COVID-19)
by the World Health Organization (WHO). Despite the imposition of strict quarantine
rules and restricted travel within and from China, the infection has spread rapidly
across the globe, with continuing escalation in the UK, necessitating a plan by NHS
England to build several temporary “Nightingale” hospitals (of up to 4,000 beds each),
to provide mass-scale extra capacity.
Following on from three initial statements2, 3, 4 this Editorial from the British
Society of Thoracic Imaging (BSTI) will collate ongoing knowledge, guidance, and developing
conundrums related to COVID-19 imaging.
COVID-19 chest radiograph reporting template
The unprecedented surge in UK COVID-19 cases has seen the chest radiograph (CXR) emerge
as the frontline diagnostic imaging test,
5
in conjunction with clinical history and key blood biomarkers: C-reactive protein
(CRP) and lymphopenia. In early March 2020, BSTI developed a simple, internationally
recognised CXR reporting template to help facilitate standardised reporting with embedded
CXR reporting codes allowing retrospective radiology information system keyword searches
for audit purposes
6
(Electronic Supplementary Material Appendix S1).
This template stratifies patients into one of four groups based on CXR findings: classic/probable
COVID-19 findings (Fig 1
); indeterminate for COVID-19 (Fig 2
); normal (COVID-19 not excluded in the correct clinical context); and features most
consistent with a non-COVID-19 diagnosis (Fig 3
). Frontline doctors have found this a useful adjunct to clinical assessment, particularly
when CXRs are “hot-reported”. More nuanced severity grades can also be applied, in
particular for the classic/probable COVID-19 CXRs: these include a simple qualitative
scoring method (mild, moderate, severe) or the quantitative modified Radiographic
Assessment of Lung Edema (RALE) scoring system.
7
Figure 1
A 68-year-old intubated man with fever and cough. Classic COVID-19 appearances on
CXR. This patient had a positive RT-PCR for COVID-19 on the second throat swab after
an initial negative throat swab.
Figure 1
Figure 2
A 39-year-old woman with fever and ill-defined left perihilar opacity. The RT-PCR
throat swab was COVID-19 positive.
Figure 2
Figure 3
A 77-year-old woman with large left pleural effusion and no significant alveolar opacity.
The CXR revealed a non-COVID diagnosis.
Figure 3
COVID-19 CT reporting template
Several authors
8
and professional societies have agreed computed tomography (CT) should not be used
as a first-line or pure diagnostic test. Although sensitive, CT findings of COVID-19
are not specific. In addition, false-negative CT rates vary in the literature, ranging
from 3–56% in reverse transcriptase (RT)-polymerase chain reaction (PCR)-positive
patients.
9
In contrast, mainland China frequently employed CT as a first-line investigation for
COVID-19; however, such practice was burdensome on radiology departments and hugely
challenging for infection control. It is also noteworthy that the use of CT as the
main diagnostic tool for COVID-19 over and above RT-PCR in China has been challenged
subsequently.
10
Moreover, use of chest CT findings as a diagnostic criterion were removed in the more
recent sixth and seventh editions of the Diagnosis and Treatment Program of 2019 New
Coronavirus Pneumonia proposed by The National Health Commission of China.
4
CT in COVID-19 shows typical findings of ground-glass opacity, peripheral consolidation,
or a combination of both. Ai et al. reported CT sensitivity from their Wuhan cohort
of 97% when compared to RT-PCR.
11
In light of this conflicting evidence, BSTI issued a statement on the potential role
of CT in COVID-19, highlighting that any design of local diagnostic imaging algorithms
would be contingent on RT-PCR sensitivity (dependent upon both the assay and the technique)
and turnaround times,
4
which have geographical variation. At the time of writing, the UK government is planning
on increasing RT-PCR testing 10-fold, to approximately 25,000 tests a day.
12
Where CT does emerge as important is in the BSTI/NHS England COVID-19 diagnostic algorithm
(Electronic Supplementary Material Appendix S2). Patients who are unwell with an indeterminate
or normal CXR may then be considered for CT imaging. This will help decide subsequent
management and need for isolation. CT can also be justifiably used in a cohort of
patients with high clinical probability but suspected false-negative RT-PCR results.
In early March, BSTI suggested a CT reporting template
6
(Electronic Supplementary Material Appendix S3) that could help as a guide to facilitate
consistency of reporting. This has recently been simplified. It should be noted that
CT severity scoring is not validated as linked to potential outcomes in COVID-19 and
should be reserved, if used, for qualitative assessment of parenchymal involvement.
Further potential templates for CXR and CT in the intensive care setting are also
included in this paper for reference (Electronic Supplementary Material Appendices
S4 and S5).
When performed, a low-dose unenhanced CT (LDCT) technique should be utilised and consideration
given to an additional CT pulmonary angiogram (PA; if renal function permits).
6
The need for CTPA is to ensure occult pulmonary embolism (PE) is excluded, particularly
in light of increasing reports of pro-thrombotic risk in COVID-19. The unenhanced
LDCT prior is performed for two main reasons
1
: dependent ground glass is often exaggerated on contrast-enhanced CT and can be mistaken
for true ground glass
2
; having a baseline unenhanced examination can be useful if subsequently performing
unenhanced imaging to look for fibrosis. It is also recognised that on-table “hot”
review of lung windows, including lung bases on CT abdominal examinations, can be
beneficial. Such rapid review permits quick detection of incidental COVID-19 cases
on designated “clean” scanners, allowing appropriate decontamination.
Three-step imaging requesting tool
With RT-PCR turnaround times at >24–48 h in large portions of the country, CRP, lymphocyte
count, and CXR have emerged as the bedrock of COVID-19 patient triage.
13
Evidence from the current UK epicentre, London, indicates that several patients undergoing
imaging investigations for other indications, including for abdominal pain in the
absence of respiratory symptoms, frequently demonstrate COVID-19 pulmonary features
on CT. This “unexpected” COVID-19 diagnosis on CT has necessitated stricter vetting
strategies for all imaging requests with the aim of avoiding further exacerbation
of COVID-19 spread amongst staff and COVID-19 naive patients. Such is the fear around
contaminating non-infected patients with COVID-19 on supposedly “clean” CT machines,
stringent vetting of imaging requests is essential. We suggest a three-step triage
tool (Electronic Supplementary Material Appendix S6), developed in collaboration with
the Royal Free London NHS Trust (RFL NHS Trust), provides a simple, effective method
of screening patients who are referred for imaging:
(1)
Does the patient have a new cough or fever? Yes or No
(2)
Is there a lymphopenia or raised CRP? Yes or No
(3)
Is the CXR consistent with COVID-19? Yes or No
This tool has allowed radiology departments to triage imaging requests to determine
whether they should take place on the designated “COVID” or the “clean” CT machine.
The tool can also be incorporated into digital imaging requesting systems as a mandated
field.
The role of CT in COVID-19 infection control management
An emerging bottleneck to effective care is dealing with hospital isolation capacity.
Inpatients who are truly RT-PCR negative can be moved to a non-isolation ward, freeing
up isolation beds for confirmed COVID-19 positive patients, also reducing the risk
of virus transmission within the hospital setting. Patients with a high index of clinical
suspicion of COVID-19, but who test negative on initial RT-PCR test, continue to be
managed with respiratory isolation precautions, often undergoing repeat RT-PCR testing;
however, this means yet further delay whilst awaiting test results.
A further concern being recognised regarding de-isolation is patient discharge to
either home or community care: a confirmed COVID-19 diagnosis or a confirmed non-COVID-19
diagnosis is key to impart infection control advice to both families and carers of
discharged patients.
With stepdown from isolation to de-isolation beds becoming a rate-limiting step, an
effective pathway to deal with negative RT-PCR results in the setting of high clinical
probability of COVID-19 infection has been developed at the RFL NHS Trust (Electronic
Supplementary Material Appendix S7).
14
This pathway, which is predicated on two negative RT-PCR tests and two normal/indeterminate
CXRs in the setting of high clinical probability of COVID-19 infection, facilitates
a systematic approach to patient de-isolation and stepdown from the COVID ward to
non-COVID wards, whilst also providing a means to exclude an alternative diagnosis.
CT chest imaging in the acute emergency surgical abdomen
In the patient with a surgical abdomen requiring potential emergency surgery, intubation
and ventilation could generate COVID-19 aerosols. Some reports have also emerged of
increased mortality in COVID-19-positive patients undergoing surgery.
15
As such, BSTI and the British Society of Gastrointestinal and Abdominal Radiology
(BSGAR) have suggested a template for this scenario (Electronic Supplementary Material
Appendix S8).
A complete CT thorax is advocated if the clinical decision has already been made to
send the patient for a CT abdomen and pelvis. It should be stressed, however, that
this recommendation does not apply to patients in whom abdominal CT (or magnetic resonance
imaging [MRI]) is being performed for other reasons, or electively, and that this
pathway is purely to facilitate decisions in severely unwell patients where the clinical
diagnosis of an acute abdomen is requires emergency surgery.
BSTI disagrees with current Intercollegiate General Surgery Guidance on COVID-19,
16
which recommends that “any patient undergoing abdominal CT scan should also have a
CT chest”. As previously mentioned, we do not advocate CT chest screening for COVID-19
(indeed neither is there a role for multiple RT-PCR testing in asymptomatic individuals).
CT chest is only 54% sensitive in asymptomatic patients who are RT-PCR positive for
COVID-19.
17
Therefore, a negative CT thorax cannot be considered to have sufficient negative predictive
value to exclude COVID-19; however, in the context of an emergent surgical abdomen,
a negative CT thorax, as well as low clinical probability of COVID-19, could aid the
confidence in the surgical decision to take the patient to theatre or manage the patient
conservatively.
Other situations where CT screening has been notionally considered include prior to
non-emergency surgery; interventional radiology mechanical thrombectomy for proximal
vessel occlusion in ischaemic stroke; and primary percutaneous coronary intervention
for ST-elevation myocardial infarction. The general principles regarding “CT screening”
hold true for these cohorts: there is a significant risk of a false-negative study,
resulting in inappropriate reassurance and the risk of a patient potentially being
denied treatment for findings that are not specific for COVID-19. It is also important
to emphasise that it remains uncertain how CT manifestations correlate with COVID-19
“infectivity”. Even if there were CT changes compatible with COVID-19, it would not
necessarily be possible to determine where that patient lies in the temporal evolution
of the changes with a single time-point analysis and changes may persist on CT for
>28 days.
18
BSTI/NHSX collaboration
When confronted by a new disease, sharing data and expertise is essential to maintain
homogeneous diagnostic and management standards across a country. This is particularly
true in a condition such as COVID-19 where imaging has a central diagnostic role.
Given the scale of the pandemic, radiology workforces have been configured so that
non-chest specialist radiologists now have to report chest imaging. It has become
apparent that rapid deployment of technology to upskill the UK radiology workforce
in COVID-19 imaging interpretation is both necessary and urgent.
Although, in theory, existing image transfer capabilities across the NHS should have
facilitated widespread data sharing across sites, in practice, this has never been
coordinated on a regional or national scale in the UK. The transfer of imaging studies
directly from a picture archiving and communications system (PACS) viewer to a central
imaging repository would make teaching and research in the NHS more efficient and
accurate.
3
By bringing together thoracic CT expertise across the UK, a central COVID-19 imaging
repository and database could help reduce unwarranted diagnostic variation, thereby
providing better quality of care for patients, and provide a resource to accumulate
up-to-date information in a rapidly changing clinical setting.
Through a partnership with NHSX, BSTI have created a national database of CXR and
CT images alongside a linked clinical dataset. The purpose of the National COVID-19
Chest Imaging Database (NCCID) initiative is to enable the development and validation
of artificial intelligence technologies in the domains of medical image analysis and
data science and the enabling of research projects in response to the COVID-19 pandemic.
Further to the creation of a database, NHSX will be able to accelerate the adoption
of technologies that may prove effective in supporting COVID-19 care pathways. Our
anticipation is that these data will help influence management decisions during the
COVID-19 pandemic in “real time”.
With the recognition that front-line clinical staff will have to prioritise patient
management, the NCCID initiative has been built to run without significant input from
clinical teams. Images can be uploaded by hospital PACS teams through the Image Exchange
Portal network. Clinical data will be collected by research staff, who have been redeployed
for COVID-19-related research, and uploaded to a parallel central repository linked
to imaging data. In time, more formal linkage with other clinical COVID-19 databases
is anticipated. In addition to data, the NCCID resource also aims to link researchers
asking similar clinical questions around the country. We hope to bring the brightest
minds to work on the challenges posed by COVID-19. By doing this on a national level,
we hope to speed up scientific discovery by utilising economies of scale and resource
across UK institutions and researchers.
Emerging imaging conundrums in COVID-19: pulmonary embolism in COVID-19
Most of the published CT literature in COVID-19 has been in unenhanced CT; however,
there is increasing concern about venous thromboembolism (VTE) risk in COVID-19. Acute
pulmonary embolism (PE) could be incidental or related to COVID-19. Within the latter,
this might be due to generic increased risk of VTE with infection
19
or due to a specific pro-thrombotic state caused by or resulting from COVID-19 infection.
The reports of acute coronary syndromes with unobstructed coronary arteries and thrombotic
gastrointestinal complications in COVID-19 patients lend weight to the latter. The
question whether microvascular pulmonary arterial occlusion is responsible for some
of the clinical deterioration in patients with COVID-19 is a hot topic. Whether this
epiphenomenon occurs in the absence of macrovascular disease with PE and/or deep venous
thrombosis is uncertain, but likely. The role of empirical anticoagulation in some
COVID-19 subgroups is being considered, but anecdotal reports of spontaneous haemorrhage,
including sub-arachnoid haemorrhage, make decision-making challenging.
Point-of-care ultrasound could help in potentially two situations
1
: Doppler ultrasound of the deep veins of the legs to assess for deep-vein thrombosis
or
2
focussed right-heart echocardiography to assess right ventricular findings in acute
PE; however, the former will not answer the question of microvascular disease and
the latter may be confounded by concomitant acute or chronic lung disease. In the
presence of ground glass and/or consolidation, ventilation–perfusion (V/Q) assessment
for macro/microvascular pulmonary arterial disease will be limited. CTPA will depict
PE out to the subsegmental level, providing imaging quality is sufficient. The recent
study of Chen et al.
20
comprising 1,008 patients admitted to hospital with COVID-19 pneumonia, reported the
findings in 25 patients who underwent CTPA. Acute PE was found in 10/25 with a high
D-dimer level in those with PE compared to those without. Traditional ancillary imaging
findings of microvascular disease, for example, in the context of chronic thromboembolic
disease, may be challenging to assess reliably in the context of concomitant viral
pneumonia and/or acute respiratory distress. Dual-energy CTPA has been demonstrated
to predict regional perfusion heterogeneity in a porcine model of acute lung injury,
21
but it is uncertain whether it would accurately detect microvascular pulmonary arterial
occlusion. Practically speaking, consideration will need to be given to the use of
intravenous iodinated contrast medium in patients at risk of acute kidney injury (AKI).
Although there will be understandable concerns of contrast-induced nephropathy, a
recent meta-analysis of 28 studies and 107,335 participants found no significant differences
in AKI, need for renal replacement therapy, or all-cause mortality between patients
receiving contrast-enhanced CT versus those receiving unenhanced CT, suggesting other
patient and illness factors, rather than contrast material are likely to contribute
to AKI.
22
It should be noted that patients with COVID-19, or suspected COVID-19, requiring high-flow
nasal cannula oxygen delivery or non-invasive ventilation pose a risk of aerosolisation.
23
Transferring such patients to radiology departments for CTPA would have implications
for infection control and cleaning as per local institutional policy. At the time
of writing, we therefore can only endorse CTPA imaging where the result will alter
clinical management and it is considered safe to administer intravenous contrast media.
Nevertheless, it is acknowledged that the threshold for requesting and performing
CTPA may be lower than originally thought, especially in cases with “disproportionate”
hypoxia, and potentially even in unexpectedly very high D-dimers (>5,000 ng/ml) or
a new “significant rise” in D-dimer.
Percutaneous lung biopsy in COVID-19
The Society of Interventional Radiology recommends the use of appropriate personal
protective equipment for interventional radiologists “performing any aerosol-generating
procedure (i.e., any procedure that is likely to induce coughing)”.
24
BSTI endorses the classification of percutaneous lung biopsy (PLB) as a potential
aerosol-generating procedure due to the increased risk of patient coughing during
and immediately after the procedure. We would also suggest a period of 7 days of self-isolation
after an uncomplicated PLB in order to mitigate the risk of a COVID-19 naive patient
acquiring COVID-19 infection in the post-procedure period; it is felt that this pragmatic
advice potentially reduces the risk of COVID-19 pulmonary infection as an insult on
top of any small post-biopsy pneumothorax or perilesional haemorrhage. It should also
be recognised that ambulatory lung biopsy
25
,
26
permits PLB to be performed in an entirely outpatient setting. This may allow continuation
of lung cancer diagnostic work-up at a time when bed availability is likely to be
scarce and liable to risk of nosocomial CVOID-19 transmission. The potential introduction
of serology-based COVID-19 testing may be useful in the triage of PLB requests by
indicating those patients that may have immunity from earlier, mild COVID-19 infection.
Conclusion
The management strategies in COVID-19 patients continue to evolve. Despite the vast
numbers of worldwide cases, it remains difficult to standardise care and guidelines.
New challenges will emerge and, with them, further dilemmas are faced. Radiologists
should be aware of the current issues of potential increased thrombotic risk in COVID-19,
not only affecting the pulmonary arteries but also other organs. Although the role
for CTPA remains unclear, it is quite possible that the threshold for considering
it will be lower than originally anticipated. Through the collaboration of BSTI with
NHSX, there will be a refined dataset of imaging, clinical parameters, and outcome
measures that can hopefully answer some of these questions.
Conflict of interests
The authors declare no conflict of interest.