Pulmonary vascular disease in COVID-19: insights from artificial intelligence analysis in a large multicentre imaging database

ABSTRACT

Aims and ObjectivesAn increased incidence of pulmonary thrombosis (PT) and right ventricular (RV) dysfunction is reported in COVID-19. The clinical significance is not fully understood and there are few large, multicentre studies. The National Covid-19 Chest Imaging Database (NCCID) was analysed for prevalence of PT in COVID-19 patients;we hypothesised associations between macroscopic PT, severity of parenchymal disease, evidence of RV dysfunction on CT and mortality.MethodsNCCID is a multicentre UK-wide centralised database comprised of radiological images from hospitalised COVID-19 patients. 391 thoracic contrast CT scans from 14 centres across England and Wales performed between 2nd March 2020 – 10th September 2020 underwent automated post-processing software (IMBIO LLC.) to determine RVLV diameter ratio. Scans were manually reported for PT and quantitatively scored for arterial obstruction and severity of parenchymal involvement using CT- Severity Scoring (CT-SS)[1]. Imaging metrics were analysed for association with PT and 30 day mortality.ResultsAutomated RVLV analysis was successful in 90% (351/391) of scans. Mean age 64, 53% (186/351) male. Mortality data was available for 325 patients 22 died within 30 days of scan (6.7% (22/325)).Macroscopic PT was present in 16% (56/351). Median Qanadli score was 6% (IQR 3%-17.5%), indicating low burden arterial obstruction. PT was not associated with mortality (p=0.18).RVLV >1 on CT was observed in 59% (206/351) (mean RVLV 1.08). RVLV was significantly higher in the presence of PT (mean RVLV 1.17 vs 1.06 p=0.011, χ2(2) = 6.499). RVLV was not predictive of mortality (AUC 0.467, CI 0.358–0.576).CT-SS significantly predicted mortality (AUC 0.787, p=<0.0005, CI 0.693–0.881). However there was no correlation between severity of parenchymal involvement and RVLV (r 0.82, p=0.123), nor presence of PT (χ2(2) 2.305, p=0.129).ConclusionsRV dilatation and PT were prevalent in this multicentre cohort of COVID-19 patients, but were not associated with mortality or parenchymal disease severity. PT is frequently low burden and, in contrast to PT outside the context of COVID-19, RVLV >1 is not discriminatory for prognosis.ReferenceYang R., et al. Chest CT severity score an imaging tool for assessing severe COVID-19. Radiology Cardiothoracic Imaging 2020;2(2)e200047. doi 10.1148/ryct.2020200047