On February 6, 2020, the Korean Society of Thoracic Radiology proposed a multi-institutional collaboration for reporting the radiologic findings of COVID-19. Seven members with different affiliations decided to participate in this study, and the working group submitted a document for review by the Institutional Review Board of each affiliated hospital. Due to between-hospital differences in the required period for Institutional Review Board review, we expedited the process to report cases for which approval was received, while excluding cases under review. Consequently, we included nine consecutive patients (median age, 54 years; four men and five women) with COVID-19 infections confirmed through a pan-coronavirus conventional polymerase chain reaction assay from January 2020 through February 9, 2020 from Incheon Medical Center (7), Seoul National University Hospital, and Seoul National University Bundang Hospital.

All CT examinations were performed using a multi-detector CT scanner with 64 or more channels (Somatom Definition, Somatom Definition AS+, or Somatom Force, Siemens Healthineers, Erlangen, Germany). The detailed parameters for CT acquisition were as follows: tube voltage, 120 kVp; tube current, standard (reference mAs, 60–120) to low-dose (reference mAs, 30) with automatic exposure control; slice thickness, 1.0 mm; reconstruction interval, 1.0–3.0 mm; and a sharp reconstruction kernel. CT images were obtained with the patient in the supine position at full inspiration and without contrast medium.

All patients underwent a baseline digital anteroposterior chest radiography at full inspiration using a mobile chest radiograph machine (FLUOROSPOT Compact FD, Siemens Healthineers; DRX-Revolution Mobile X-ray System, Carestream Health, Rochester, NY, USA; or Optima XR220, GE Healthcare, Milwaukee, WI, USA).

Eight of the nine patients had parenchymal abnormalities on their baseline chest CT scans. In another patient, the baseline chest CT scan was normal, and parenchymal abnormalities were observed on a follow-up CT scan one week later. Accordingly, we analyzed eight baseline chest CT scans and one follow-up CT scan with abnormal findings of COVID-19 pneumonia.

Two attending radiologists (with 15 and 12 years of experience in chest imaging, respectively) reviewed the chest radiographs and CT images by consensus on a picture archiving and communication system (PACS, INFINITT Healthcare co., Ltd., Seoul, Korea). The readers assessed the presence, location, and density of parenchymal abnormalities on chest radiographs without reviewing the CT images, and graded the abnormalities using the following 5-point scale (8): 1, normal; 2, patchy atelectasis and/or hyperinflation and/or bronchial wall thickening; 3, focal alveolar consolidation involving no more than one segment or one lobe; 4, multifocal consolidation; and 5, diffuse alveolar consolidation. After the initial assessment of the chest radiographs, we checked whether the abnormalities on chest radiographs corresponded to abnormalities on chest CT images.

The CT images were evaluated with both lung (width, 1500 HU; level, −600 HU) and mediastinal (width, 400 HU; level, 40 HU) window settings. The readers identified all separate pulmonary lesions on CT and analyzed the longest diameter, location, shape, density, and margin of the lesions. The locations of the lesions were specified as lobar, axial, and anteroposterior. For the axial location, the peripheral lung was defined as the outer one-third of the lung, and the central lung was defined as the inner two-thirds of the lung. According to a horizontal line dividing the lung into anterior and posterior halves, locations were classified as anterior or posterior. The shapes of the lesions were first categorized as patchy, confluent, or nodular. Patchy lesions were defined as isolated focal lesions with no nodular shape in the segment, and confluent lesions were defined as large fused lesions involving multiple segments. The densities of the patchy to confluent lesions were classified as pure GGO, mixed GGO and consolidation, consolidation, or a crazy-paving appearance. The densities of the nodular lesions were classified as pure GGO, part-solid, or solid. The margins of the lesions were dichotomized as ill-defined or well-defined. We evaluated whether the pulmonary lesions had a predilection for the lower lobes, pleura, or bronchovascular bundles. Furthermore, the readers evaluated the presence of air-bronchograms, the reversed halo sign, cavities, micronodules, a tree-in-bud appearance, and pleural effusion based on the Fleischner Society glossary of terms for thoracic imaging (9).

Fisher's exact test was used to compare patchy to confluent lesions and nodular lesions in terms of the proportion of pure GGO lesions, axial location, anteroposterior location, margin, and predilection for the lower lobes, pleura, and bronchovascular bundles. Statistical analyses were performed using SPSS version 25.0 (IBM Corp., Armonk, NY, USA). P values < 0.05 were considered to indicate a statistically significant difference.

Five of the nine patients showed radiographic abnormalities (severity grade 2, one patient; grade 3, two patients; and grades 4 and 5, one patient each). Chest CT scans revealed that the grade 2 lesion and one of the grade 3 lesions as areas of prominent breast tissue mimicking faint GGO and post-inflammatory focal atelectasis on chest radiography, respectively. In summary, three of the nine patients (33.3%) were confirmed to have parenchymal abnormalities related to COVID-19 pneumonia on chest radiographs (Figs. 1A, 2A, 3A).

One of the three patients had a single nodular opacity in the left lower lung zone (Fig. 3A), and the other two patients had four and five patchy opacities in both lungs (Figs. 1A, 2A), respectively (Table 1). In a per-lesion analysis, 50% of the 10 opacities presented in the lower lung zones, 80% of the opacities were located in the peripheral lungs, and 70% of the opacities were areas of consolidation.

In total, 77 lung parenchymal lesions were identified in the nine patients, of whom eight had bilateral lung parenchymal abnormalities. The median numbers of parenchymal lesions and involved lobes were 5 (interquartile range, 2–13) and 2 (interquartile range, 2–5), respectively. The most frequently involved lobe was the right lower lobe (eight patients), followed by the left upper and lower lobes (six patients each). Among the 77 identified lesions, 39% were patchy, 13% were confluent, and 48% were nodular (Table 2). The average diameters of the patchy, confluent, and nodular lesions were 2.6 ± 1.5 cm, 9.8 ± 2.6 cm, and 1.3 ± 0.6 cm, respectively. The peripheral and posterior lung were involved in 78% and 67% of the lesions, respectively. Patchy to confluent lesions were more prevalent than nodular lesions per patient (median relative proportion of patchy to confluent lesions, 86% [interquartile range, 33–100%]).

Patchy to confluent lesions primarily manifested as mixed GGO and consolidative lesions (50%) (Fig. 1B, C), followed by pure GGO lesions (35%) (Fig. 2B, C), lesions with a crazy-paving appearance (10%), and areas of consolidation (5%). The most frequent shape of the lesions was wedge-shaped (42%), followed by elongated (33%) and confluent (25%). They frequently had an ill-defined margin (70%) and 28% of the lesions had an air-bronchogram. Nodular lesions primarily manifested as pure GGO lesions (57%), followed by predominantly GGO lesions (32%) and predominantly solid lesions (11%). Nodular lesions were round (95%) and had ill-defined margins (75%). There were no cavities, micronodules, lesions with a tree-in-bud appearance, or pleural effusions. Compared to nodular lesions, patchy to confluent lesions were primarily distributed along the pleura (80% vs. 35%; p < 0.001) and more commonly involved the lower lobes (60% vs. 35%; p = 0.040) (Fig. 4A). The nodular lesions were primarily distributed along the bronchovascular bundles (59% vs. 28%; p = 0.006) (Fig. 4B) and tended to manifest as pure GGO lesions (57% vs. 35%; p = 0.069).