CT Findings Don't Differentiate COVID-19 Lung Injury from Other Causes of Pneumonia
Key findings
- This study, the first of its kind, assessed six radiologists' ability to differentiate CT findings in patients previously diagnosed with COVID-19 pneumonia, influenza pneumonia or non-infectious organizing pneumonia (OP)
- Radiologists showed poor diagnostic accuracy when evaluating a sample containing equal percentages of these three types of pneumonia, with average accuracy of 70% for COVID-19 pneumonia and 68% for both influenza pneumonia and OP
- Across all six readers, the average percentage of correct diagnoses was 44% correct for COVID-19, 29% for influenza and 39% for OP
- Between COVID-19 pneumonia and influenza pneumonia, there were significant differences in the frequency of seven individual CT features, but only four individual findings significantly differed between COVID-19 pneumonia and OP
- Radiologists should remain aware of these differences in CT features and the difficulties in distinguishing types of pneumonia, especially COVID-19 pneumonia versus non-infectious OP
The specificity of chest CT for COVID-19 pneumonia is lower than its sensitivity because many other conditions are associated with similar imaging findings. Brent P. Little, MD, assistant program director of the Radiology Residency Program in the Department of Radiology at Massachusetts General Hospital, and colleagues recently became the first to evaluate radiologists' ability to differentiate the organizing pneumonia (OP)-like imaging pattern that occurs in COVID-19 from other causes of OP.
Subscribe to the latest updates from Radiology Advances in Motion
In the American Journal of Roentgenology, they present findings that may help distinguish COVID-19 from influenza pneumonia, but they say the extent of similarities between COVID-19 pneumonia and OP underscore the difficulties in distinguishing those two diagnoses by imaging alone.
Study Methods
The researchers queried a Mass General imaging database for patients who underwent chest CT and were diagnosed with COVID-19 pneumonia, influenza pneumonia or OP. COVID-19 pneumonia diagnoses were made between March 1 and May 1, 2020; influenza pneumonia and OP were diagnosed between January 1, 2011, and December 31, 2019.
Fifty examinations were selected for each of the three diagnoses. Six thoracic fellowship-trained radiologists reviewed each of the examinations and used a standardized spreadsheet to record the presence or absence of 14 CT features.
COVID-19 vs. Influenza
COVID-19 and influenza differed significantly on seven of the 14 CT features:
- Axial distribution—More commonly peripheral in COVID-19 than influenza (27% vs. 15%); more commonly central and peripheral (55% vs. 34%) in COVID-19 than influenza; more commonly diffuse (31%) in influenza than COVID-19 (7%) (P<0.001)
- Craniocaudal distribution—Lower lobe predominant distribution more common in COVID-19 (35%) than influenza (25%); upper lobe predominant distribution more common in influenza (14%) than COVID-19 (5%) (P=0.03)
- Tree-in-bud nodules—More common in influenza (36%) than COVID-19 (6%) (P<0.001)
- Diffuse ground-glass opacities—More common in influenza (20%) than COVID-19 (7%) (P=0.02)
- Perilobular pattern or peripheral band-like opacities—More common in COVID-19 (42%) than influenza (28%) (P<0.001)
- Reverse halo sign—More common in COVID-19 (34%) than influenza (12%) (P<0.001)
- Bilateral pleural effusions—More common in influenza (38%) than COVID-19 (14%) (P=0.004)
COVID-19 vs. OP
COVID-19 and OP differed significantly on four features:
- Axial distribution—Solely central distribution was more common in OP (20%) than COVID-19 (10%), as was diffuse distribution (21% vs. 7%) (P<0.001)
- Laterality—Unilateral findings were more common in OP (7%) than COVID-19 (1%) (P=0.01)
- Nodules (not tree-in-bud)—More common in OP (53%) than COVID-19 (32%) (P=0.003)
- Tree-in-bud nodules—More common in OP (14%) than COVID-19 (6%) (P=0.03)
RSNA Categories
Each reader assigned each case to one category from the Radiological Society of North America (RSNA) consensus guidelines for the diagnosis of COVID-19 ("typical," "indeterminate," "atypical" or "negative"):
- 70% of COVID-19 cases were considered "typical" vs. 33% for influenza and 47% for OP
- Among cases assigned a "typical" CT pattern for COVID-19, 47% had previously been diagnosed as COVID-19, 22% as influenza and 31% as OP
Favored Diagnoses
Each reader also assigned each case a favored diagnosis from a list of options:
- Across all six readers, the average percentage of correct diagnoses was 44% for COVID-19, 29% for influenza and 39% for OP
- The average diagnostic accuracy of favored diagnoses was 70% for COVID-19 and 68% for both influenza and OP
Guidance for Clinical Practice
The RSNA "typical" category and the presence of a reverse halo sign or perilobular distribution may help distinguish COVID-19 pneumonia from influenza pneumonia. However, the extent of similarities between the CT findings of COVID-19 pneumonia and OP underscore the difficulties in distinguishing these two diagnoses by imaging alone. Physicians should also consider clinical presentation, chronicity of imaging findings and conditions or treatments predisposing to OP, such as connective tissue disease or certain drug therapies.
This may be a particular challenge when patients are receiving chemotherapy, immunotherapy or targeted therapy for cancer, as many of those agents are associated with an OP pattern of lung injury.
view original journal article Subscription may be required
Learn more about the Department of Radiology at Mass General
Refer a patient to the Department of Radiology