Radiological Society of North America Expert Consensus Document on Reporting Chest CT Findings Related to COVID-19: Endorsed by the Society of Thoracic Radiology, the American College of Radiology, and RSNA

CT Imaging Features

Several papers have found that COVID-19 typically presents with GGO with or without consolidation in a peripheral, posterior, and diffuse or lower lung zone distribution (6–11). GGO has also been frequently reported to have round morphology or a “crazy paving” pattern (6,8). However, a significant portion of cases have opacities without a clear or specific distribution (8). A predominant perihilar pattern was not reported (8). Bronchial wall thickening, mucoid impactions, and nodules (“tree-in-bud” and centrilobular) seen commonly in infections, are not typically observed (8). Lymphadenopathy and pleural effusion have been rarely reported (6,21).

The frequency of imaging findings also depends on when infected patients are imaged. A slight majority of patients had a negative CT during the first 2 days after symptom onset with GGO usually developing between day 0 and 4 after symptom onset and peaking at 6–13 days (8,9,22–24). Therefore, a negative CT should not be used to exclude the possibility of COVID-19, particularly early in the disease. Later in the course of the disease, the frequency of consolidation increases as does the likelihood of seeing a reverse halo or atoll sign, typically absent near the time of symptom onset (8). Available evidence regarding these CT findings is limited, and new patterns of pulmonary involvement may eventually be reported (25).

Diagnostic CT Performance and Screening

Chest CT findings can precede positivity on reverse-transcription polymerase chain reaction testing (RT-PCR). Early reports of RT-PCR sensitivity vary considerably, ranging from 42% to 71% (26,27), and an initially negative RT-PCR may take up to 4 days to convert in a patient with COVID-19 (26). The reported sensitivities and specificities of CT for COVID-19 vary widely (60%–98% and 25%–53%, respectively) (26–29), likely due to the retrospective nature of the currently published studies, including lack of strict diagnostic criteria for imaging and procedural differences for confirming infection. The positive and negative predictive value of chest CT for COVID-19 are estimated at 92% and 42%, respectively, in a population with high pretest probability for the disease (eg, 85% prevalence by RT-PCR) (27). The relatively low negative predictive value suggests that CT may not be valuable as a screening test for COVID-19 at least in earlier stages of the disease.

Literature comparing individual CT features of COVID-19 or radiologists’ performance in correctly choosing COVID-19 as a first-choice diagnosis on imaging is limited. In one study, six of seven radiologists demonstrated 93%–100% specificity in correctly distinguishing CT features of COVID-19 from other viral infections (10). A peripheral distribution of GGO was found to correctly distinguish COVID-19 from other viral causes 63%–80% of the time. However, the authors did not include high numbers of influenza-A or any noninfectious causes such as drug reaction, which could degrade radiologists’ performance.

Viral Testing—Implications for CT

In reviewing CT publications on COVID-19, it is important to consider the accuracy of the laboratory viral testing used. This applies both to the collection method and the laboratory testing method (30), as many articles published on chest CT do not specify the sample collection or RT-PCR method used (31). With respect to collection method, bronchoalveolar lavage fluid (BALF) testing is the most sensitive, but not for general use given the invasive nature of fluid collection, and because it is an aerosol-generating procedure that could place health care workers at greater risk. Sputum and nasopharyngeal swab collection are considered equivalent in sensitivity, while throat swab testing is less sensitive. As viral pneumonias typically do not result in production of purulent sputum, nasopharyngeal swab is the preferred method for sample collection (31). As an example, in a recently published series of 1070 patients, the majority of samples collected were throat swabs, and throat swabs detected only approximately half of the positive cases that were detected by nasal swabs (32).

Rapid antigen tests are fast but have poor sensitivity. While RT-PCR is the most accurate, not all tests are equivalent. Eleven different RT-PCR tests were approved for use in China between January 26 and March 12, 2020, with varying levels of sensitivity. In a report of CT findings in 1014 patients (26), with 59% having a positive RT-PCR and 88% having a positive chest CT, the method of swab collection was not described. Two different RT-PCR tests were used, one of which does not appear on the list of approved tests, and the other approved for use in nasal, throat and sputum collection. The sensitivity of tests approved for use in the United States is high, with emergency use authorizations available on the website of the U.S. Food and Drug Administration (33).

Structured Reporting

Rationale and use.—The goal of structured reporting in the setting of COVID-19 pneumonia is to help radiologists recognize the findings seen, decrease reporting variability, reduce uncertainty in reporting findings potentially attributable to this infection, and enhance the referring provider’s understanding of those radiologic findings, thereby allowing better integration into clinical decision making. While we do not currently recommend the use of CT screening for COVID-19 pneumonia, we suggest using a standardized language when specifically asked to address whether or not findings of COVID-19 pneumonia may be present on CT and propose language that could be placed in the impression of the report.

How to report incidentally discovered features potentially attributable to COVID-19 pneumonia is more complex. When typical features of COVID-19 pneumonia are present in an endemic area as an incidental finding, we recommend direct communication with the referring provider to discuss the likelihood of viral infection and to try to reach consensus. As always, radiologists should follow the ACR Practice Parameter for Communication of Diagnostic Imaging Findings (34). These incidental findings do not necessarily need to be reported as COVID-19 pneumonia, with “viral pneumonia” as a reasonable and inclusive alternative. However, if consensus is reached, and COVID-19 is mentioned as a potential diagnosis in the radiology report, we suggest using the provided standardized reporting language. Additionally, staff at the site performing the examination should be notified to initiate standard operating procedures (SOP) for potential exposure.

It should be noted that viral pneumonias have a wide variety of imaging manifestations, some of which are atypical or less common in COVID-19 such as tree-in-bud opacities and other small nodules, bronchial wall thickening, and bronchial mucus plugs (12). Thus, the term viral pneumonia encompasses a range of imaging findings some of which are not typical for COVID-19. It is also important to describe other lung abnormalities that may be associated with increased morbidity in the setting of COVID-19, such as emphysema and diffuse parenchymal lung disease.

Categories.—We propose four categories for reporting CT imaging findings potentially attributable to COVID-19, each with suggested standardized language (Table 1). The reporting language does not offer an exact likelihood for COVID-19 pneumonia, which depends on several factors including prevalence in a community, exposure, risk factors, and clinical presentation. Rather, the reporting language focuses on CT findings reported in the literature and the typicality of these features in COVID-19 pneumonia rather than other diseases. Included in the reporting language are unique coding identifiers in brackets that can then be used for future data mining.

Table 1: Proposed Reporting Language for CT Findings Related to COVID-19, including Rationale, CT Findings, and Suggested Reporting Language for each Category

Typical features are those that are reported in the literature to be frequently and more specifically seen in COVID-19 pneumonia in the current pandemic (10,11). (Figs 1–4). The principal differential diagnosis includes some viral pneumonias, especially influenza, and acute lung injury patterns, particularly organizing pneumonia, either secondary, such as from drug toxicity and connective tissue disease, or idiopathic.

Figure 1:

Download as PowerPoint

Figure 2:

Download as PowerPoint

Figure 3:

Download as PowerPoint

Figure 4:

Download as PowerPoint

Indeterminate features are those that have been reported in COVID-19 pneumonia but are not specific enough to arrive at a relatively confident radiologic diagnosis. An example would be diffuse GGO without a clear distribution (Figs 5, 6). This finding is common in COVID-19 pneumonia but occurs in a wide variety of diseases such as acute hypersensitivity pneumonitis, Pneumocystis infection, and diffuse alveolar hemorrhage, which are difficult to distinguish by imaging alone.

Figure 5:

Download as PowerPoint

Figure 6:

Download as PowerPoint

Atypical features are those that are reported to be uncommon or not occurring in COVID-19 pneumonia and are more typical of other diseases such as lobar or segmental consolidation in the setting of a bacterial pneumonia, cavitation from necrotizing pneumonia, and tree-in-bud opacities with centrilobular nodules, as can occur with a variety of community acquired infections and aspiration (Figs 7–9).

Figure 7:

Download as PowerPoint

Figure 8:

Download as PowerPoint

Figure 9:

Download as PowerPoint

Negative for pneumonia implies that there are no parenchymal abnormalities that could be attributable to infection. Specifically, GGO and consolidation are absent. Importantly, there may be no findings on CT early in COVID-19. Conversely, CT has been reported to be more sensitive than RT-PCR earlier in the course of the disease (29), although this result may change with local RT-PCR test characteristics.

Pros, Cons, and Limitations of Standardized Reporting

There are compelling arguments both for and against the use of standardized reporting language in describing CT findings potentially attributable to COVID-19 (Table 2). This project largely stemmed from the expectation that despite most current professional guidelines recommending against the routine use of screening CT for COVID-19, CT may be requested for potential assistance in diagnosis and management, particularly if RT-PCR is not readily available.

Table 2: Pros and Cons of Standardized Reporting for Chest CT Findings Related to COVID-19

Pros.—Without expert consensus, radiologists may be left with uncertainty as to how to convey the presence, absence, or likelihood of COVID-19 when confronted with this as a specific indication or as an incidental finding. Standardized reporting can provide guidance and confidence to radiologists as well as increased clarity to providers through reduced reporting variability. Providing unique identifiers for each category facilitates mining data for future educational, research, and quality improvement. In addition, standardized radiology reports combined with clinical assessment may enable future care pathways to determine which patients may preferentially undergo RT-PCR should testing capacity be exceeded. Initial RT-PCR testing may also be negative, and typical imaging findings may encourage repeat testing.

Cons.—The true sensitivity and specificity of CT for COVID-19 remains relatively unknown. One study showed that radiologists identified COVID-19 versus other viral pneumonias correctly 60%–83% of the time based on typical CT imaging features (10). However, the results of this study must be evaluated cautiously as all of the COVID-19 cases came from one country (China) and most of the control cases from a single institution in another country (United States). Additionally, this moderate level of distinction may be reduced in clinical practice as the control cases included a low proportion of influenza-A, which is the major viral pneumonia that must be differentiated from COVID-19 during the winter and spring months across the northern hemisphere. Reporting “atypical features” may result in false-negative cases, and the risk of missing COVID-19 can have broad implications. Ordering providers may also feel that having “COVID-19” or “coronavirus” documented in a radiology report constrains their clinical decision making and treatment options. This concern is less relevant in PUIs, as clinical suspicion already exists. However, difficulties may arise in patients with findings suggestive of COVID-19 that are incidentally detected. Direct communication with the referring provider about the likelihood of COVID-19 is recommended to avoid surprising providers and patients. We again emphasize that as an incidental finding, particularly with indeterminate or atypical features, “viral pneumonia” may be preferable to “COVID-19” or “coronavirus.”

Limitations.—We anticipate cases with mixed imaging findings, that is, those that have both typical and atypical imaging features for COVID-19. Recent analysis suggests that over 20% of patients with COVID-19 may have coexistent infections complicating the categorization of imaging observations (35). The radiologist will have to determine whether or not these findings are part of the same process or are unrelated. For example, a hospitalized patient undergoing chest CT for fever could have lower lobe tree-in-bud opacities as well as peripheral GGO, which could reflect aspiration superimposed on viral pneumonia. It is also possible that atypical features such as lobar consolidation may reflect a secondary bacterial pneumonia even in patients who test positive for COVID-19. Available evidence is still limited concerning the appearance of COVID-19 in the presence of secondary disease processes such as coexistent infections and aspiration. In scenarios such as these, discussion with the treating team would be prudent.

Imaging appearances in the standardized reporting language are based upon available literature at the time of writing in March 2020, noting the retrospective nature of many reports, including biases related to patient selection in cohort studies, examination timing, and other potential confounders. As radiologists’ experience with COVID-19 increases, our categorization of these findings as typical, indeterminate, or atypical may evolve.