29.04.2020

CO-RADS - A categorical CT assessment scheme for patients with suspected COVID-19: definition and evaluation

Diagnostic Radiology
Prokop M et al
Radiology

Main result

1. Scoring presentation:
<P>
  • CO-RADS 0: incomplete or poor-quality scanner.
  • CO-RADS 1 (very low level of suspicion): non-infectious, unambiguous or normal scanner.
  • CO-RADS 2 (low level of suspicion): typical infectious lesions not compatible with COVID-19.
  • CO-RADS 3 (equivocal/uncertain): equivocal results for pulmonary involvement of COVID-19 based on CT characteristics, which can also be found in other etiologies (non-infectious and viral infections).
  • CO-RADS 4 (high level of suspicion): presence of typical COVID-19 lesions but some overlap with other (viral) pneumonia.
  • CO-RADS 5 (very strong suspicion level): typical lesions. Mandatory features are frosted glass opacities, with or without consolidation, in the lung regions near the visceral pleural surfaces, including fissures.
  • CO-RADS 6 multifocal bilateral distribution: RT-PCR positive

<P>

  1. Agreement among observers: value of kappa coefficient 0.47 (95% CI 0.45-0.49). Moderate agreement. There is substantial agreement for the categories CO-RADS 1 (0.58 (0.54-0.62)) and CO-RADS 5 (CO-RADS 5: 0.68 (0.65-0.72)).
  2. Diagnostic performance: CO-RADS was able to distinguish patients with PCR+ from those with PCR- with a mean AUC of 0.91 (95% CI, 0.85-0.97).The mean AUC increased to 0.95 (95% CI, 0.91-0.99) if a clinical diagnosis of COVID-19 was also accepted.

Takeaways

Interesting classification to diagnose SARS-CoV-2 infection. However, radiological signs may take some time to appear after the onset of infection. Therefore, the CO-RADS scores 1 and 2 should be tempered if they are close to the onset of symptoms.

Agreement among observers appears moderate with the exception of CO-RADS scores 1 and 5 where agreement is substantial.

Strength of evidence Moderate

- Multicenter, retrospective study
- Blinded evaluation
- Randomized image selection
- Small effective: 105 scanners, 8 observers
- Acute phase of the pandemic (generalization not evident in other contexts)
- Clinical diagnosis of COVID-19 taking into account radiology

Objectives

Presentation of CO-RADS, a standardized assessment scheme for pulmonary involvement of COVID-19 on chest CT, and observational study assessing interobserver variability and diagnostic accuracy.

Method

  • Presentation of CO-RADS, a 5-level score (+ 2 technical levels)  assessing the probability of SARS-CoV-2 infection based solely on chest scan, and following previous standardization efforts such as Lung-RADS, PI-RADS or BI-RADS. The classification follows the April 7th consensus.
  • Retrospective, multicenter study (7 hospitals in the Netherlands) of 105 chest scanners of suspected COVID-19 patients admitted to the emergency room. Only patients with RT-PCR were included. The selection was random. Stratification into three groups: PCR+, clinical PCR- and clinical symptoms for COVID-19, PCR- and no clinical symptoms for COVID-19. Scoring of anonymized images by 8 different radiologists.
  • Establishing a ROC curve for each observer. Comparison of CO-RADS performance with two references: PCR and combination of PCR with diagnosis of clinical COVID-19.
  • Use of kappa coefficient to measure agreement between radiologists.
Read paper

bibliovid.org and its content are bibliovid property.

Legal Notice