Patients
The study was conducted retrospectively, during September and October 2020, with 73 critically ill patients, who were confirmed to have COVID-19 according to RT-PCR and were admitted to the ICU with progressive dyspnea and low O2 saturation on room air (PaO2 ≤ 93%). They included 53 males and 20 females (73%:27%) and their age ranged from 18 to 88 years (mean ± SD=53.3 ± 13.5). All patients were hospitalized and the patients’ demographic data, medical histories, and laboratory results were collected retrospectively from the hospital information system. Clinical evaluations were performed by a single consulting pulmonologist.
The study was approved by the Ethics Committee of our University Hospital. The need to obtain informed patient consent was waived by the Research Ethics Board with the assurance of the confidentiality of patient information and medical records.
The inclusion criteria were as follows: Critically ill patients who were confirmed to have COVID-19 based on RT-PCR and were admitted to the ICU with progressive dyspnea and low O2 saturation on room air (PaO2 < 93%), which necessitated external airway support with either high-flow nasal oxygen or mechanical ventilation (according to the universal criteria for the clinical classification of COVID-19 patients).
The exclusion criteria were as follows: (1) CT images quality was degraded by respiratory motion artifacts, (2) unavailable medical or laboratory records, and (3) cardiomegaly or a history of previous pulmonary vascular disease.
Imaging modalities
CT pulmonary angiography (CTPA) was performed using 64 detector CT scanners from GE Medical Systems (USA) and Siemens SOMATOM Sensation 64 (Germany). The scanning was performed in a supine position with full inspiration and a single breath-hold. The images were obtained in a caudo-cranial direction using the high-quality mode at 120 kVp and 50–100 mAs. The slice thickness was 1.25 mm, and the spacing between slices was 1 mm. The FOV was 350mm × 350 mm. Images were acquired after the IV injection of 60 ml of non-ionic contrast followed by a saline chaser at a flow rate of 5.0–6.0 ml/s using bolus tracking at the threshold of 100 HU on the pulmonary trunk.
Imaging data analysis
CT images were retrospectively analyzed by three consulting radiologists, who came to a consensus (14–15 years of experience in the field of chest imaging). Images were evaluated in three different greyscale windows: (1) the lung window using a width/level of (1500/−600), (2) the mediastinal window (350/40), and (3) the pulmonary embolism window (700/100). Images were assessed for both morphological and vasculopathy CT signs using minimal intensity projection (MIP) volume reconstruction.
The morphological signs included ground-glass opacities (GGOs), consolidations, the crazy-paving pattern (ground-glass opacities with inter-lobular septal thickening), and fibrosis. Extra-pulmonary signs included pleural effusion and mediastinal lymph node enlargement.
The vasculopathy signs included the following:
-
(1)
Pulmonary embolism which was defined as a contrast filling defect within the pulmonary arterial tree.
-
(2)
Pulmonary hypertension which was defined as dilated pulmonary trunk > 3 cm and that exceeded the internal diameter of the nearby portion of the aorta.
-
(3)
Pulmonary infarcts which was defined as wedge-shaped sub-pleural ground-glass opacities or consolidative patches.
-
(4)
Pulmonary vascular enlargement inside and/or outside the pathological parenchymal ground-glass opacities, which was manifested by asymmetrical internal dilatation of the pulmonary arterial branches. They branches abnormally extended peripherally until the pleural lining, with the loss of normal tapering. Comparison to nearby segments in the same lung or the same segment in the contralateral lung at the same horizontal or axial level was used for assessment [8].
-
(5)
The pulmonary vascular “tree-in-bud pattern,” which was identified by dilated, beaded, and branching peripheral pulmonary arterioles that communicated to the pulmonary arterial tree on MIP images [9].
Quantitative assessment using an artificial intelligence system
Raw CTPA data were automatically transferred from both MDCT machines to a PACS with an incorporated artificial intelligence system called the CT Pneumonia Analysis algorithm. It was designed by ‘Siemens Health engineers’ to automatically perform quantitative CT measurements in all patients. It did not reside on a specific workstation. This automated quantification of pathological lung volumes is more precise than any other routine human semi-quantitative or qualitative method. It is also essential for standardizing the pathological lung volume calculation between different radiologists.
The artificial intelligence algorithm automatically identifies the diseased parts of the lungs (mainly those with high opacity) using an input of 2D CT images and the output of computerized 3D CT images and quantifies the pathological lung volumes using 3D segmentation and two major calculations:
-
(1)
The opacity score/index was estimated for each lobe of both lungs and then finally summed for both lungs (total opacity score—TOS): score 0 (≤ 25%), score 1 (25–50%), score 2 (50–75%), score 3 (> 75%), and score 4 (100%).
-
(2)
The percentage of the predicted volume of abnormal opacity in comparison to the total lung volume was estimated.
Other measurements were also computed such as volume of the lung (ml), opacity volume (ml), high-opacity volume (ml), percentage with opacity within a given lung region (%), total and mean HU of the entire lung and any specific lung region, and the HU standard deviation of the entire lung and any specific lung region. The computerized software calculated all these measurements for both lungs together and per lung, lung lobe and segment.
Statistical analysis of the data
-
a.
The prevalences of abnormal clinical, laboratory, or radiological data (percentage of patients having each criterion or abnormality) were calculated.
-
b.
A statistical analysis was performed to determine if there were significant relationship between the radiological CT signs corresponding to pulmonary vascular angiopathy and the automatically calculated CT opacity score. This was performed with chi-squared tests, and the P values were obtained with an online calculator (https://www.socscistatistics.com). (A P value <0.05 was considered statistically significant.)