RSNA COVID-19 Chest CT Classification System Leads to Similar Interpretations

June 15, 2020

Radiologists using RSNA classification system largely agree on image assessments.

Radiologists who use an RSNA-developed chest CT classification system to interpret possible cases of COVID-19 pneumonia can feel an increased level of confidence in their evaluations, according to newly published research.

In a study from the June 11 Radiology: Cardiothoracic Imaging, investigators from the Zuyderland Medical Center in The Netherlands reported that the agreement between the RSNA system and reading radiologists underscores how the system can help providers with assessing and diagnosing patients who have possible COVID-19 infection.

Even though the RT-PCR test is still considered the gold standard for diagnosing COVID-19 infection, it is not always sensitive to early-stage disease, and getting the results can be a slow process. In these cases, said study lead Tom de Jaegere, M.D., chest CT – as the quicker, more COVID-sensitive test – can be an effective tool. These results should bolster provider confidence in the results of these scans.

“Because disagreement between CT interpreters can result in dissimilar diagnoses and subsequent patient management recommendations, high interobserver agreement is crucial before chest CT can be routinely used in practice,” wrote de Jaegere’s team.

According to the researchers, reading and interpreting chest CT scans for COVID-19 infection can be tricky, particularly because there is little diagnostic clarity for using chest CT with COVID-19 and the virus can mimic other respiratory conditions. Also, in many instances, readers do not have the experience needed for confident interpretation.

Previously, RSNA did publish a four-category system, offering guidance on how to report COVID-19 chest CT results, the team noted. But, to date, that system has not yet been appropriately evaluated. Consequently, de Jaegere’s team designed a study to do just that.

Related Content: 4 Categories for Reporting COVID-19 CT Findings

Using two chest radiologists and a 5th-year radiology resident to evaluate chest CT scans, investigators used the Kappa co-efficient measure to determine reader agreement – a score of 1.0 indicates perfect agreement. The providers assessed the interobserver agreement on interpretations of chest CT scans from 96 patients who were suspected of COVID-19 infection, 45 of whom received COVID-19-positive results via the RT-PCR test.

Related Content: Dutch Group Creates CO-RADS Classification System

The team compared agreement between the two radiologists, as well as the radiologists and the resident. They looked at results using both the RSNA classification system and the CO-RADS system developed by Dutch investigators. According to the results, the two radiologists had a high level of agreement using the RSNA (0.663 kappa score) and the CO-RADS system (0.773 kappa score).

They identified moderate agreement between the radiologists and the resident. With the first radiologist, they calculated a 0.570 kappa score with the RSNA system and a 0.658 kappa score with CO-RADS. In comparison with the second radiologist, the investigators found a 0.564 kappa score with RSNA and a 0.648 kappa score with CO-RADS.

It is still important to note, de Jaegere’s team said, that some cases of COVID-19 confirmed with RT-PCR will not show up as positive on CT scans. This can happen in a significant number of cases, they wrote.

“Radiologists and clinicians should take into account that the proportion of RT-PCR confirmed COVID-19 cases in categories "atypical appearance" and "negative for pneumonia" is non-negligible,” they said.