Secondary Logo

Journal Logo

Original Articles

Performance of Chest Computed Tomography in Differentiating Coronavirus Disease 2019 From Other Viral Infections Using a Standardized Classification

Borges da Silva Teles, Gustavo MD*,†; Kaiser Ururahy Nunes Fonseca, Eduardo MD*,‡; Yokoo, Patricia MD*; Marques Almeida Silva, Murilo MD*; Yanata, Elaine MD*; Shoji, Hamilton MD*,§; Bastos Duarte Passos, Rodrigo MD*; Caruso Chate, Rodrigo MD*,‡; Szarf, Gilberto MD, PhD*,∥

Author Information
doi: 10.1097/RTI.0000000000000563
  • Free

Abstract

At the beginning of this year, humanity was confronted with a rapidly progressive outbreak of a new coronavirus infection, the so-called coronavirus disease 2019 (COVID-19). Since then, it got a pandemic status and prompted societies of different countries to adapt to a different reality, as a contagious disease that poses an enormous demand on health systems and limited supplies. Its diagnosis is confirmed with SARS-CoV-2 RNA detection by reverse transcription-polymerase chain reaction (RT-PCR) after collection of a nasopharyngeal swab, oropharyngeal swab, or lower respiratory tract specimens.1 Reports of RT-PCR sensitivity vary substantially (42% to 71%),2,3 and low sensitivities can possibly be attributed to low viral load in test specimens or laboratory error.4,5

In many places, the results of RT-PCR can take hours or days, so radiologists are expected to extract as much information as possible after images are obtained on COVID-19 suspected patients. Recently, an expert consensus proposed a standardized COVID-19 computed tomography (CT) reporting language in order to help radiologists recognize and report CT findings of COVID-19 pneumonia and to aid their communication with referring providers.6

Although most medical specialty societies currently do not recommend the use of CT as a screening method for COVID-19, a recent multidisciplinary panel recommended that imaging is indicated for medical triage of patients with suspected COVID-19, who present with moderate to severe clinical features and high-pretest probability of disease in a resource-constrained environment.7 The utility of CT in this scenario increases depending on its ability to distinguish COVID-19 pulmonary features from noninfectious and infectious causes, such as other viral pneumonia.

The aim of this study was to evaluate the performance of chest CT in differentiating COVID-19 from other viral infections in our institution, using a RSNA-endorsed reporting classification system,6 in consecutive patients evaluated during the same time period.

METHODS

Patients

This retrospective study was performed in a single tertiary-care medical center in Sao Paulo, Brazil, and it was approved by the local Institutional Review Board. Written informed consent was waived. Data were collected from 350 consecutive patients with suspected acute respiratory infection and available chest CT from March 15 to March 24, 2020. Patients with no COVID-19 RT-PCR test within 7 days (n=119) were excluded. Patients with negative COVID-19 RT-PCR and negative results from respiratory pathogen panel (RPP) (n=46) or patients with a positive COVID-19 RT-PCR and positive results from RPP (n=10), indicating coinfection, were also excluded. Thus, the final cohort consisted of 175 patients (Fig. 1).

FIGURE 1
FIGURE 1:
Illustration of the study flow.

As there was a shortage of COVID-19 RT-PCR kits during this time period, chest CT was used as part of the initial assessment of patients with respiratory symptoms, even in mild cases. Also, due to its low availability, COVID-19 RT-PCR tests were not repeated after an initial negative test in patients with positive RPP.

The following clinical information were extracted from patients’ electronic medical records: age, sex, presence of comorbidities (systemic arterial hypertension, diabetes, dyslipidemia, obesity, defined as body mass index >30, cardiopathy, including coronary artery disease, arrhythmias and cardiac failure, asthma, thyroidopathy, and cancer), COVID-19 RT-PCR and RPP results and date of acquisition, and duration between the symptom onset and chest CT.

RT-PCR AND RPP

RT-PCR assays were performed using MagNA Pure 96 DNA and Viral NA Small Volume Kit (Roche Molecular Systems Inc.) and real-time PCR amplification and detection according to Charité protocol. The tests of ePlex Respiratory Panel (GenMark Diagnostics, Carlsbad, CA) were performed in the Microbiology Lab at our institution.

Chest CT Scans

CT examinations were acquired using multidetector CT scanners with 40, 80, and 320 detector rows (Biograph mCT, Siemens Healthcare, Erlangen, Germany, Aquilion Prime and Aquilion ONE, Canon Medical Systems, Tochigi, Japan). All scans were obtained in supine position during end-inspiration, without intravenous contrast material. Acquisition parameters for all CT scans were as follows: reconstructed slice thickness of 1 mm; voltage of 80 to 120 kVp; automatic milliampere setting with a range of 10 to 440 mA.

CT Analysis

Two thoracic radiologists (with 11 and 2 y of experience interpreting chest imaging), blinded to RT-PCR and RPP results, reviewed all chest CT images independently, in a standard clinical Picture Archiving and Diagnostic System workstation. They assessed and classified the image features according to the Radiological Society of North America Statement on Reporting Chest CT Findings Related to COVID-19.6

According to this statement, it was proposed 4 categories for reporting CT imaging findings potentially attributable to COVID-19: “typical” appearance (Fig. 2)—peripheral or bilateral ground glass opacities (GGO), with or without consolidation or “crazy-paving” pattern, or multifocal GGO of rounded morphology with or without consolidation or “crazy paving” pattern, or reverse halo sign or other signs of organizing pneumonia; “indeterminate” (Fig. 3)—nonrounded or nonperipheral multifocal, diffuse, perihilar or unilateral GGO, with or without consolidation lacking a specific distribution, or few very small GGO with a nonrounded and nonperipheral distribution; “atypical” appearance (Fig. 3)—isolated lobar or segmental consolidation without GGO, or discrete small nodules (centrilobular, “tree-in-bud”), or lung cavitation, or smooth interlobular septal thickening with pleural effusion; and “negative” for pneumonia—no CT features to suggest pneumonia. The presence of pleural effusion, mediastinal, or hilar lymphadenopathy was also assessed. Final classification was based on consensus between the readers.

FIGURE 2
FIGURE 2:
Examples of CT “typical” appearance. A and B, Multifocal peripheral bilateral pulmonary GGO and consolidation. Final diagnosis: COVID-19. C and D, Multifocal bilateral pulmonary GGO with reverse halo sign. Final diagnosis: metapneumovirus pneumonia.
FIGURE 3
FIGURE 3:
A and B, Example of CT “indeterminate” appearance. Few small pulmonary GGO with a nonrounded and nonperipheral distribution in the left lower lobe (black circles). Final diagnosis: COVID-19. C and D, Example of CT “atypical” appearance. Bronchial thickening and discrete centrilobular nodules and “tree-in-bud” opacities in the lower lobes (black circles). Final diagnosis: rhinovirus/enterovirus infection.

Statistics

Continuous variables were displayed as mean±SD and categorical variables were reported as counts and percentages. Sex, comorbidities, and chest CT categories were compared using χ2 test and Fisher exact test between COVID-19 and non-COVID-19 infections. Age was compared using Student t test and duration of symptoms using Mann-Whitney test. P-values <0.05 were considered statistically significant. Inter-reader agreements of the CT classification were assessed with linearly weighted κ statistics. Values >0.80 were considered excellent: 0.61 to 0.80 good, 0.41 to 0.60 moderate, 0.21 to 0.40 fair, ≤0.20 poor.

Final CT classification based on the consensus was separated into 2 scenarios: the first scenario considered the “typical” appearance as positive test (for COVID-19), and the other categories (“indeterminate,” “atypical,” and “negative” for pneumonia) as a negative test. The second scenario considered the “typical” and the “indeterminate” appearances as a positive test, and the others as negative. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated for each scenario. Data were analyzed with SPSS version 21.0 (SPSS Inc., Chicago, IL).

RESULTS

Our final cohort consisted of 175 patients, including 87 with positive RT-PCR test for COVID-19 and 88 with negative COVID-19 RT-PCR test but positive RPP (the pathogens detected are detailed in Fig. 1). The average number of days between CT and RT-PCR test results was 1.0±2.0 days. Patients with COVID-19 were older than patients with other viral infections (mean age of 46.7 vs. 40.0 y, P=0.038) (Table 1). The prevalence of cancer was higher in non-COVID-19 patients compared with COVID-19 patients (P=0.031). There were no statistical differences between sex, other comorbidities, and duration of symptoms between the groups. There were also no statistical differences for the presence of pleural effusion, mediastinal or hilar lymphadenopathy between the groups.

TABLE 1 - Clinical Information and CT Classification of Individuals, Stratified by Final Diagnosis
Variables Non-COVID-19 COVID-19 Total P
Sex, n (%) 0.69
 Male 51 (49.0) 53 (51.0) 104
 Female 37 (52.1) 34 (47.9) 71
Mean age (SD) (y) 40.0 (25.4) 46.7 (15.9) 43.3 (21.4) 0.038
Comorbidities, n (%) 0.200
 No 37 (45.1) 45 (54.9) 82
 Yes 51 (54.8) 42 (45.2) 93
Systemic arterial hypertension, n (%) 0.396
 No 74 (49.0) 77 (51.0) 151
 Yes 14 (58.3) 10 (41.7) 24
Diabetes, n (%) 0.340
 No 84 (51.2) 80 (48.8) 164
 Yes 4 (58.3) 7 (63.6) 11
Dyslipidemia, n (%) 0.790
 No 81 (50.0) 81 (50.0) 162
 Yes 7 (53.8) 6 (46.2) 13
Obesity (BMI >30), n (%) 0.555
 No 68 (49.3) 70 (50.7) 138
 Yes 13 (43.3) 17 (56.7) 30
Cardiopathy, n (%) 0.790
 No 81 (50.0) 81 (50.0) 162
 Yes 7 (53.8) 6 (46.2) 13
Asthma, n (%) 0.551
 No 73 (49.3) 75 (50.7) 148
 Yes 15 (55.6) 12 (44.4) 27
Thyroidopathy, n (%) >0.999*
 No 83 (50.3) 82 (49.7) 165
 Yes 5 (50.0) 5 (50.0) 10
Cancer, n (%) 0.031
 No 79 (48.2) 85 (51.8) 164
 Yes 9 (81.8) 2 (18.2) 11
Duration of symptoms, mean (dSD) (d) 4.6 (4.1) 5.0 (5.7) 4.8 (4.9) 0.333
Pleural effusion, n (%) 0.329*
 No 85 (51.2) 81 (48.8) 166
 Yes 3 (33.3) 6 (66.7) 9
Mediastinal or hilar lymphoadenopathy, n (%) 0.770
 No 82 (50.0) 82 (50.0) 164
 Yes 6 (54.5) 5 (45.5) 11
CT classification (consensus), n (%) <0.001
 Negative for pneumonia 53 (80.3) 13 (19.7) 66
 Atypical 24 (92.3) 2 (7.7) 26
 Indeterminate 9 (52.9) 8 (47.1) 17
 Typical 2 (3.0) 64 (97.0) 66
Total 88 (50.3) 87 (49.7) 175
Bold values indicate significance (P< 0.05).
χ2 test.
*Fisher’s exact test.
Student t test.
Mann-Whitney test.
BMI indicates body mass index.

The inter-rater agreements of CT categories between the 2 readers ranged from good to excellent, κ=0.80 (0.73 to 0.87) (Table 2). Final CT categories based on consensus were statistically different among COVID-19 and non-COVID-19 groups (P<0.001). CT “typical” appearance was more prevalent in the COVID-19 group (64/87, 73.6%) than in the non-COVID-19 group (2/88, 2.3%). CT classifications of “negative” for pneumonia and “atypical” appearance were less prevalent in COVID-19 patients (13/87, 14.9%, and 2/87, 2.3%, respectively) than in non-COVID-19 patients (53/88, 60.2%, and 24/88, 27.3%, respectively). The prevalence of the CT “indeterminate” criteria was similar between the groups (8/87, 9.2%, in COVID-19 patients and 9/88, 10. 2%, in non-COVID patients) (Table 1).

TABLE 2 - CT Categories Assigned Per Reader and Inter-rater Agreements
CT Classification—Reader 2, n (%)
CT Classification—Reader 1 Negative Atypical Indeterminate Typical Total
Negative 58 (33.1) 2 (1.1) 0 (0.0) 0 (0.0) 60 (34.3)
Atypical 3 (1.7) 22 (12.6) 1 (0.6) 0 (0.0) 26 (14.9)
Indeterminate 7 (4.0) 5 (2.9) 8 (4.6) 1 (0.6) 21 (12.0)
Typical 0 (0.0) 0 (0.0) 5 (2.9) 63 (36.0) 68 (38.9)
Total 68 (38.9) 29 (16.6) 14 (8.0) 64 (36.6) 175 (100.0)
κ, CI (95%) 0.801 (0.730-0.872)
CI indicates confidence interval.

When considering CT “typical” appearance as a positive test and the other categories as a negative test (scenario 1), it was observed a sensitivity of 73.6% (95% confidence interval [CI]: 63%-82.4%), specificity of 97.7% (95% CI: 92%-99.7%), PPV of 97% (95% CI: 89.5%-99.6%), NPV of 78.9% (95% CI: 70%-86.1%), and accuracy of 85.7% (95% CI: 80.5%-85.7%). In the second scenario, including the “typical” and “indeterminate” appearance as a positive test, it was obtained a sensitivity of 82.8% (95% CI: 73.2%-90%), specificity of 87.5% (95% CI: 78.7%-93.6%), PPV of 86.7% (95% CI: 77.5%-93.2%), NPV of 83.7% (95% CI: 74.5%-90.6%), and accuracy of 85.1% (CI 95%: 79.8%-85.1%) (Table 3).

TABLE 3 - Performance Results of CT Classification for COVID-19 Diagnosis
Test Performance
Sensitivity Specificity PPV NPV Accuracy
Variable CI (95%) CI (95%) CI (95%) CI (95%) CI (95%)
Scenario 1 73.6 97.7 97 78.9 85.7
(63-82.4) (92-99.7) (89.5-99.6) (70-86.1) (80.5-85.7)
Scenario 2 82.8 87.5 86.7 83.7 85.1
(73.2-90) (78.7-93.6) (77.5-93.2) (74.5-90.6) (79.8-85.1)
Scenario 1: CT “typical appearance” as positive test, other categories as negative test.
Scenario 2: CT “typical” and “indeterminate appearance” as positive test, other categories as negative test.
CI indicates confidence interval.

The CT categories in the COVID-19 group were statistically different (P=0.007) regarding the duration of symptoms (Table 4). In early disease (duration of symptoms 0 to 5 d), 12 patients (19.7%) had a negative CT, 8 patients (13.1%) had an “indeterminate” appearance, and 41 patients (67.2%) had a “typical” appearance. In COVID-19 patients with symptom duration of more than 5 days, a “negative” CT was found in 1 patient (4.3%), an “atypical” appearance was found in 1 patient (4.3%), and a “typical” appearance was found in 21 patients (91.3%). No statistically significant differences were found between the duration of symptoms and the radiologic classification in non-COVID-19 patients (Table 5).

TABLE 4 - CT categories and Duration of Symptoms in COVID-19 Patients
Duration of Symptoms (d)
CT Classification (Consensus) 0-5 d >5 d Total P
Negative for pneumonia, n (%) 12 (19.7) 1 (4.3) 13 (15.5) 0.007
Atypical, n (%) 0 (0.0) 1 (4.3) 1 (1.2)
Indeterminate, n (%) 8 (13.1) 0 (0.0) 8 (9.5)
Typical, n (%) 41 (67.2) 21 (91.3) 62 (73.8)
Total 61 (100.0) 23 (100.0) 84 (100.0)
Bold value indicates significance (P < 0.05).
Likelihood ratio test.

TABLE 5 - CT Categories and Duration of Symptoms in Non-COVID-19 Patients
Duration of Symptoms (d)
CT Classification (Consensus) 0-5 d >5 d Total P
Negative for pneumonia, n (%) 37 (57.8) 15 (68.2) 52 (60.5) 0.375
Atypical, n (%) 20 (31.3) 4 (18.2) 24 (27.9)
Indeterminate, n (%) 5 (7.8) 3 (13.6) 8 (9.3)
Typical, n (%) 2 (3.1) 0 (0.0) 2 (2.3)
Total 64 (100.0) 22 (100.0) 86 (100.0)
Likelihood ratio test.

DISCUSSION

Our results showed relevant differences between the tomographic presentation of pulmonary involvement in COVID-19 compared with other viral infections in our institution, during the same time interval, particularly when a “typical” appearance was detected. We also found good to excellent inter-rater agreement of the CT standardized classification between the 2 readers.

The majority of RT-PCR positive patients for COVID-19 in our study had a “typical” appearance (73.6%) on CT, according to the standardized classification. This is consistent with previous articles, which have found that COVID-19 typically presents with pulmonary GGO, frequently with round morphology, with or without consolidation or “crazy paving” pattern, and with a predominant peripheral zone distribution.8–11 We found the same CT “typical” appearance in only 2 patients from the non-COVID-19 group (2.3%), confirming that these findings are possible but rare in other viral infections. Bai et al12 also showed that a peripheral distribution of GGO correctly distinguished COVID-19 from other viral pneumonias in 63% to 80% of the cases.

“Indeterminate” CT appearance had a similar prevalence in COVID-19 patients and in non-COVID-19 patients in our study. It reinforces the fact that some cases of both COVID-19 and other viral pneumonias may have few small GGO or GGO without a clear or specific distribution.10,12 An “atypical” appearance of CT was more prevalent in non-COVID-19 viral infections in comparison to COVID-19. This is also in agreement with other articles, which demonstrated that some imaging features seen commonly in other infections, such as nodules (“tree-in-bud” and centrilobular), are not typically observed in COVID-19.10 Interestingly to note is that, during this time interval, the absence of abnormalities consistent with pneumonia on CT, classified as “negative” for pneumonia, was also more prevalent in other viral infections than in COVID-19.

Our study evaluated CT classification in relation to the duration of symptoms. CT scans were obtained relatively early during the clinical course (mean: 4.8 d). This is due to the fact that chest CT was used as part of the initial assessment of patients with respiratory symptoms since there was low availability of COVID-19 RT-PCR kits. A relevant proportion of COVID-19 patients in early disease had a negative CT or an “indeterminate” appearance, significantly higher than in patients with later symptoms. This is consistent with previous articles that showed that the appearance of COVID-19 pneumonia may be different in the different stages of the disease process.10,13 In contrast, we did not find significant differences of the CT appearance of patients with other viral infections according to the duration of symptoms.

Chest CT showed high specificity and high PPV when considering a “typical” appearance as a positive test and the other categories as a negative test. Previous articles have described lower specificities of chest CT (25% to 53%).2,3,14 One of the possible reasons for that is the lack of a strict imaging diagnostic criteria in those reports. In the literature it was also described a range of specificities between 93% and 100% from 6 of 7 radiologists in differentiating COVID-19 from non-COVID-19 viral pneumonias, similar to our results.12 When we included the “indeterminate” appearance with a “typical” appearance as a positive test for COVID-19, we found a relevant decrease in CT specificity and PPV, without significant increase in sensitivity. This strengthens the role of a typical imaging appearance in the right clinical context to suggest the diagnosis of COVID-19.

We also observed that patients with COVID-19 were older than non-COVID-19 patients. A recent article also found many differences in clinical presentations between patients with COVID-19 and H1N1, including that the median age of patients with COVID-19 was higher, but it evaluated a different scenario of hospitalized patients with acute respiratory distress syndrome.15 No significant differences were found regarding sex and symptoms duration in our analysis.

Our study had limitations. First, it was a retrospective analysis in a single center, with thoracic radiologists who already developed experience with COVID-19. Also, the study was carried out during an epidemic, including a population with a high pretest probability of COVID-19. Besides that, part of the COVID-19 RT-PCR positive patients did not undergo RPP, so it was not possible to exclude coinfection in those patients. Lastly, in our sample there was just a small number of influenza infections among non-COVID-19 patients, whose differentiation from COVID-19 may be more challenging, as CT manifestations of those 2 infections can have a large amount of overlap.16

In conclusion, the novel standardized chest CT classification, besides enabling better communication between radiologists and other health care providers, demonstrated high specificity and PPV in differentiating COVID-19 from other viral infections when presenting a “typical” appearance. This is especially important in a resource-constrained setting and in a high pretest probability environment.

REFERENCES

1. Center for Disease Control and Prevention. Interim Guidelines of Collecting, Handling and Testing Clinical Specimens from Persons Under Investigation (PUIs) for Coronavirus Disease 2019 (COVID-19). 2020. Available at: https://www.cdc.gov/coronavirus/2019-ncov/lab/guidelines-clinical-specimens.html. Accessed April 14, 2020.
2. Ai T, Yang Z, Hou H, et al. Correlation of chest CT and RT-PCR testing for coronavirus disease 2019 (COVID-19) in China: a report of 1014 cases. Radiology. 2020;296:E32–E40.
3. Wen Z, Chi Y, Zhang L, et al. Coronavirus disease 2019: initial detection on chest CT in a retrospective multicenter study of 103 Chinese subjects. Radiol Cardiothorac Imaging. 2020;2:e200092.
4. Fang Y, Zhang H, Xie J, et al. Sensitivity of chest CT for COVID-19: comparison to RT-PCR. Radiology. 2020;296:E115–E117.
5. Xie X, Zhong Z, Zhao W, et al. Chest CT for typical coronavirus disease 2019 (COVID-19) Pneumonia: relationship to negative RT-PCR testing. Radiology. 2020;296:E41–E45.
6. Simpson S, Kay FU, Abbara S, et al. Radiological Society of North America Expert Consensus Statement on Reporting Chest CT Findings Related to COVID-19. Endorsed by the Society of Thoracic Radiology, the American College of Radiology, and RSNA—Secondary Publication. J Thorac Imaging. 2020;35:219–227.
7. Rubin GD, Ryerson CJ, Haramati LB, et al. The role of chest imaging in patient management during the COVID-19 pandemic: a multinational consensus statement from the Fleischner Society. Radiology. 2020;296:172–180.
8. Chung M, Bernheim A, Mei X, et al. CT imaging features of 2019 novel coronavirus (2019-nCoV). Radiology. 2020;295:202–207.
9. Kong W, Agarwal PP. Chest imaging appearance of COVID-19 infection. Radiol Cardiothorac Imaging. 2020;2:e200028.
10. Bernheim A, Mei X, Huang M, et al. Chest CT findings in coronavirus disease-19 (COVID-19): relationship to duration of infection. Radiology. 2020;295:200463.
11. Salehi S, Abedi A, Balakrishnan S, et al. Coronavirus disease 2019 (COVID-19): a systematic review of imaging findings in 919 patients. AJR Am J Roentgenol. 2020;215:87–93.
12. Bai HX, Hsieh B, Xiong Z, et al. Performance of radiologists in differentiating COVID-19 from non-COVID-19 viral pneumonia at chest CT. Radiology. 2020;296:E46–E54.
13. Wang Y, Dong C, Hu Y, et al. Temporal changes of CT findings in 90 patients with COVID-19 pneumonia: a longitudinal study. Radiology. 2020;296:E55–E64.
14. Inui S, Fujikawa A, Jitsu M. Chest CT findings in cases from the cruise ship “Diamond Princess” with coronavirus disease 2019 (COVID-19). Radiol Cardiothorac Imaging. 2020;2:e200110.
15. Xiao T, Ronghui D, Rui W, et al. Comparison of hospitalized patients with ARDS caused by COVID-19 and H1N1. Chest. 2020;S0012-3692:30558–4.
16. Lin L, Fu G, Chen S, et al. CT manifestations of coronavirus disease (COVID-19) pneumonia and influenza virus pneumonia: a comparative study. AJR Am J Roentgenol. 2020. [Epub ahead of print].
Keywords:

tomography; x-ray computed; coronavirus disease 2019; pneumonia; viral; virus diseases; diagnostic imaging

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.