Imaging COVID-19 — Radiology’s Evolving Role in Diagnosing and Managing the Coronavirus
By Beth W. Orenstein
Radiology Today
Vol. 21 No. 5 P. 10

Because the newly discovered coronavirus, SARS-CoV-2, can be spread from human to human, it is critical to know whether someone has been infected and should be quarantined. According to the latest guidelines published by the government of China, where the outbreak began, the diagnosis of COVID-19—the disease caused by the virus—must be confirmed by reverse-transcription polymerase chain reaction (RT-PCR) or gene sequencing for respiratory or blood specimens. However, the lab test has limitations and takes time.

Given the limitations of sample collection and transportation, as well as kit performance, the total positive rate of RT-PCR for throat swab samples has been reported to be between 30% and 60% at initial presentation. Could an X-ray or CT of the chest help determine whether a patient has been infected with this new coronavirus, more rapidly and more accurately than the lab tests? Paras Lakhani, MD, an associate professor of radiology at Thomas Jefferson University Hospitals in Philadelphia, sees a potential role of importance for CT.

“Sometimes the RT-PCR test is negative early in the disease course, but the chest CT is positive for findings that could represent coronavirus infection,” Lakhani says. “If you were to see an abnormality on the CT and the patient has been in contact with someone who had been in areas of outbreaks of the coronavirus, you could put on precautions and isolate the patient until repeat confirmatory RT-PCR testing is done.”

However, Edith Marom, MD, a diagnostic radiology specialist associated with the University of Texas MD Anderson Cancer Center in Houston, says the exact role of CT in the diagnosis of patients infected with the coronavirus has not yet been determined and is still evolving; none of the imaging findings of patients with suspected COVID-19 has been shown to be characteristic of a particular infection.

“These imaging findings are similar to those reported with other coronaviruses [SARS-CoV and MERS-CoV] and can be seen in other more commonly encountered respiratory viruses,” Marom says. She adds that some patients even present with normal chest CT scans. “Thus, imaging cannot diagnose this particular infection,” she says. Marom notes, however, that a chest radiograph or CT may suggest COVID-19 as a possibility in the correct clinical context.

Early Research
Around the world, radiologists are researching the use of chest CT in COVID-19 cases to help determine the indications of the disease and the role imaging information might play.

In late February, researchers at Mount Sinai in New York reported on a retrospective study of chest CTs of 121 symptomatic patients infected with COVID-19. The patients were treated at four centers in China from January 18 to February 2, 2020. The scans were reviewed for common CT findings in relationship to the time between symptom onset and the initial CT scan (early, 0 to 2 days, 36 patients; intermediate, 3 to 5 days, 33 patients; and late, 6 to 12 days, 25 patients).

The hallmarks of COVID-19 that were seen on the images were bilateral and peripheral ground-glass and consolidative pulmonary opacities. More than one-half of the early patients (20 out of 36) had a normal CT. With a longer time after the onset of symptoms, CT findings were more frequent, including consolidation, bilateral and peripheral disease, greater total lung involvement, linear opacities, a “crazy-paving” pattern, and the “reverse halo” sign, the researchers wrote. Bilateral lung involvement was observed in 10 out of 36 early patients (28%), 25 out of 33 intermediate patients (76%), and 22 out of 25 late patients (88%).

In another study, published online in Radiology the day before the Mount Sinai study, researchers in China cited the need for CT in diagnosing patients with suspected COVID-19, particularly when a lab test is negative. The researchers compared the performance of CT vs viral nucleic acid detection using RT-PCR for COVID-19 infection. They compared the results of 51 patients (29 men and 22 women) with likely symptoms who had a chest CT and RT-PCR assay within three days or less. The sensitivity of CT for COVID-19 infection was 98%, while the RT-PCR sensitivity was 71%.

“In our series, the sensitivity of chest CT was greater than that of RT-PCR,” the authors wrote. “Our results support the use of chest CT for screening for COVID-19 for patients with clinical and epidemiologic features compatible with COVID-19 infection, particularly when RT-PCR testing is negative.” The findings, the researchers contend, suggest that if patients are only tested with RT-PCR and receive negative results, they may be released and unwittingly spread the infection.

The authors offered several reasons why the lab test may be less sensitive than chest CT: The nucleic acid–detection technology is immature, the detection rate may vary from different manufacturers, patients may have a low viral load in early stages of the disease, and/or the clinical sampling may have been improper. Some of the findings on axial chest CTs that suggested COVID-19 included bilateral subpleural ground-glass opacities and consolidation, small bilateral areas of peripheral ground-glass opacities with minimal consolidation, and a right lung region of peripheral consolidation.

Radiology Resources
In a statement released in March, the ACR recommended that CT not be used to screen for or as a first-line test to diagnose COVID-19. According to the statement, “CT should be used sparingly and reserved for hospitalized, symptomatic patients with specific clinical indications for CT. Appropriate infection control procedures should be followed before scanning subsequent patients.”

Facilities may consider deploying portable radiography units in ambulatory care facilities for use when chest X-rays are considered medically necessary. The surfaces of these machines can be easily cleaned, avoiding the need to bring patients into radiography rooms, the ACR says. The organization also recommends that radiologists familiarize themselves with the CT appearance of COVID-19 infection to be able to identify findings consistent with infection in patients imaged for other reasons.

As new research is coming fast and furious, RSNA has developed a resource, Radiology of Coronavirus: Spectrum of Imaging Findings, which physicians can use for assistance with cases. “This resource was developed for radiologists who are dealing with emergency situations and who may have the need to view a number of case studies quickly and in one place,” says Linda Brooks, senior manager of public information and communications for RSNA.

The site has a feature that allows radiologists to “flip” through numerous case images that have been published in Radiology research articles, special reports, or as “Images in Radiology.” The page will be continually updated as new images are published, Brooks says. Radiologists can also find all the latest Radiology coronavirus research at Special Focus: 2019-nCoV.

In addition, RSNA published a consensus statement on March 25—developed by experts at nine US academic centers and endorsed by the ACR and the Society of Thoracic Radiology—that “aims to help radiologists recognize findings of COVID-19 pneumonia and provide guidance on reporting CT findings potentially associated with COVID-19, including standardized language to reduce reporting variability.” On March 30, RSNA also announced that it is developing an open data repository for COVID-19 imaging that “will compile images and correlative data from institutions, practices, and societies around the world to create a comprehensive source for COVID-19 research and education efforts.”

Also, on April 7, the Fleischner Society published a consensus statement on the role of chest imaging in the management of COVID-19 patients. The statement, which was published jointly in Radiology and Chest, draws opinions from clinicians in 10 countries with expertise in thoracic radiology, pulmonology, intensive care, emergency medicine, laboratory medicine, and infection control. The experts’ recommendations are that imaging not be used for asymptomatic or mildly symptomatic patients but for those with worsening respiratory status or who exhibit moderate to severe features of COVID-19, regardless of lab test results.

AI the Answer?
Researchers also are looking at whether AI could be used to help make chest CTs more reliable in diagnosing the new coronavirus. Infervision, an international company with US headquarters in Philadelphia, has a coronavirus AI solution that has been in use at the center of the pandemic outbreak at Tongji Hospital in Wuhan and other sites in China, such as the Third People’s Hospital of Shenzhen. Imaging departments are under tremendous pressure, says Matt (Yufeng) Deng, PhD, director of Infervision North America. “They are reading more than a thousand cases a day,” he says. Infervision’s AI is shortening diagnosis time for each case, and “each minute saved is critical to decrease the chance of cross-contamination at the hospital.”

Deng says it took Infervision’s researchers “three weeks to gather data, annotate it, and develop an algorithm to help those reading the images to process them more efficiently.” With every new case they screen, “it’s going to get better,” he says. “This is really urgent so we put out this algorithm as soon as we could.”

Currently, Deng says, the algorithm is in use in China only, but “we are having conversations with other countries that have not formally deployed it yet.” Infervision developed its algorithm at the request of providers in China. “When the outbreak started, they came to us and told us they needed a tool to tell quickly whether a person is at high risk and asked how quickly we could roll out an algorithm,” Deng says.

RADLogics of Silicon Valley and Tel Aviv, Israel has also launched an AI platform for automatic medical imaging analysis, including a specially designed algorithm to automatically detect early coronavirus findings on chest CT scans. The platform was launched in cooperation with Suzhou, China–based Chainz Medical Technology Co. “Our COVID-19 app was developed over a couple of weeks based on our extensive library of deep learning algorithms that have been developed over the last 10 years,” says Moshe Becker, CEO and cofounder of RADLogics. The platform is operating in a few hospitals at the center of the pandemic in the Hubei province, as well as in Jiangsu.

Lakhani says that, in theory, the algorithm could work to help identify likely cases of coronavirus more quickly. However, he says, algorithms are only as good as the data on which they are based.

“Because it was trained on a lot of scans performed in China, there may be a bias to that,” he says. “Were you to apply it to patients in America, you might find it flags influenza as coronavirus or as something else. We’re finding that AI solutions can have different outputs depending on where and how they are deployed, so they should be trained at multiple sites. With the coronavirus, ideally, it should be developed with patients in different countries that have different patient populations. As of now, we just don’t know how applicable that solution would be in other countries outside of China. The potential is there for solutions like this; we just don’t know yet if it will generalize to other countries without testing.”

Marom agrees. “Since we as humans perceive the imaging findings of [SARS-CoV-2] as not specific for this virus, there is always hope that, perhaps, AI will better distinguish between [SARS-CoV-2] patterns of disease than our human eye,” she says. However, she adds, to train the system, “one would need a library of many CT scans with [SARS-CoV-2] infection as well as many CT scans with other infections.”

Moving Forward
Could there be a role for chest CT for follow-up care of patients with the COVID-19 infection? As with diagnosis, the role of imaging in the management of these patients has not been determined, Marom says.

“Sometimes, imaging may aid in decision making as to severity of disease, which may determine which patient is admitted to an intensive care unit, though clinical signs and laboratory blood tests are more important in this decision making,” she says. As researchers and clinicians learn more about this disease, “we may determine the exact role of CT imaging in the future.

“Having said that,” Marom continues, “radiologists should be familiar with the findings described so far, as well as familiar with the specific imaging features that were absent in these patients. We do know that in the majority of these patients, the disease is bilateral, involves the posterior portion of the lungs, is peripheral, and usually of ground-glass opacities. We also know that discrete nodules, pleural effusions, and lymphadenopathy are usually absent in patients presenting with this disease. This is particularly important, when the clinical team does not suspect [SARS-CoV-2] infection, and these CT findings are seen, that the radiologist consider [SARS-CoV-2] infection in the differential and alert the treating clinicians.”

Marom is cautiously optimistic that, with continued research, imaging may eventually “play a role in the management of the hospitalized patients with COVID-19, help determine severity of their disease, or help determine complications, such as secondary superimposed bacterial infections or complications related to mechanical ventilation, just as can be seen with any acute respiratory distress syndrome patients in the intensive care unit.”

However, she says, “as there is an easily obtained laboratory test for [SARS-CoV-2] infection, and more than half of patients with COVID-19 infection have normal CT scans when screened initially, there is little role for imaging in this disease as a screening tool, from the epidemiological point of view. As for the future, it is just too early to say.”

— Beth W. Orenstein of Northampton, Pennsylvania, is a freelance medical writer and regular contributor to Radiology Today.