Methods
The retrospective review was approved by the ethics committee through the quality assurance pathway. Chest imaging from January 2020 to October 2020 in patients who tested COVID-PCR positive between January 2020 to August 2020 who presented to an Australian tertiary hospital was reviewed. The imaging was evaluated by a senior radiology registrar and four subspecialty chest radiologists including one dual trained radiologist and nuclear medicine specialist. Imaging performed within 5 days prior to the COVID-PCR positive test was also included to account for the potential lag time between onset of infection and a COVID-PCR positive test.
A set of descriptors were developed for the purposes of labelling based on descriptors in the current literature.
Chest Radiograph Descriptors
Chest radiograph descriptors were divided into zone involvement, distribution and findings. The chest radiograph was divided into 6 zones, right upper, right middle, right lower, left upper, left middle, and left lower (Fig. 1). This method of dividing the lungs into thirds is commonly used in reporting and has been proposed for the reporting of COVID-19 chest radiographs[3]. The distribution of the lung changes were also classified into three categories: peripheral, central and diffuse. Other findings that were also recorded included the presence of pleural effusions or interstitial change.
CT Descriptors
CT characteristics were classified according to pattern, distribution, lobe involvement and other findings. The CT descriptors were used for all chest CT imaging including pre and post contrast CT chests such as pulmonary angiograms.
CT patterns reviewed were based on patterns described in COVID-19 CT imaging in the literature[2]. Patterns that were included were ground glass opacities, consolidation, ground glass with consolidation, crazy paving pattern, reverse halo, interlobular septal thickening and air bronchograms. Distribution on CT was divided into peripheral, central, peribronchovascular or diffuse. Lobar involvement was also recorded based on the normal anatomical division of the right and left lung lobes (right upper, middle and lower and left upper and lower).
Other findings that were recorded included pleural effusions, lymphadenopathy, pulmonary nodules, presence of pulmonary embolism in CT pulmonary angiograms and scarring.
V/Q Descriptors
V/Q scans were analysed as either with findings or without findings. The V/Q scans recorded in this study were either performed just prior to a positive COVID swab or after the acute episode of COVID as follow up imaging.
Evaluation of readers
Between the 5 assessors, the senior radiology registrar reviewed all 661 studies. The consultant radiologist each reviewed 20 studies, 20 of which was also reviewed by the radiology registrar and 10 reviewed by another consultant radiologist. Fleiss kappa values were calculated to assess inter-rater variability between all 5 assessors for chest radiograph distribution (peripheral, central and diffuse) as well as for each of the 6 lung zones. A qualitative assessment of the responses of the 5 assessors was performed with all discrepancies reviewed.