Prognostic Implications of CT Feature Analysis in Patients with COVID-19: a Nationwide Cohort Study

Abstract

Background: Few studies have classified chest computed tomography (CT) findings of coronavirus disease 2019 (COVID-19) and analyzed their correlations with prognosis. The present study aimed to evaluate retrospectively the clinical and chest CT findings of COVID-19 and to analyze CT findings and determine their relationships with clinical severity.

Methods: Chest CT and clinical features of 271 COVID-19 patients were assessed. The presence of CT findings and distribution of parenchymal abnormalities were evaluated, and CT patterns were classified as bronchopneumonia, organizing pneumonia (OP), or diffuse alveolar damage (DAD). Total extents were assessed using a visual scoring system and artificial intelligence software. Patients were allocated to two groups based on clinical outcomes, that is, to a severe group (requiring O₂ therapy or mechanical ventilation, n = 55) or a mild group (not requiring O₂ therapy or mechanical ventilation, n = 216). Clinical and CT features of these two groups were compared and univariate and multivariate logistic regression analyses were performed to identify independent prognostic factors.

Results: Age, lymphocyte count, levels of C-reactive protein, and procalcitonin were significantly different in the two groups. Forty-five of the 271 patients had normal chest CT findings. The most common CT findings among the remaining 226 patients were ground-glass opacity (98%), followed by consolidation (53%). CT findings were classified as OP (93%), DAD (4%), or bronchopneumonia (3%) and all nine patients with DAD pattern were included in the severe group. Uivariate and multivariate analyses showed an elevated procalcitonin (odds ratio [OR], 2.521; 95% confidence interval [CI], 1.001-6.303, P = 0.048), and higher visual CT scores (OR, 1.137; 95% CI, 1.042-1.236; P = 0.003) or higher total extent by AI measurement (OR, 1.048; 95% CI, 1.020-1.076; P < 0.001) were significantly associated with a severe clinical course.

Conclusion: CT findings of COVID-19 pneumonia can be classified into OP, DAD, or bronchopneumonia patterns and all patients with DAD pattern were included in severe group. Elevated inflammatory markers and higher CT scores were found to be significant predictors of poor prognosis in patients with COVID-19 pneumonia.

Keywords: COVID-19; Chest; Coronavirus; Pneumonia; Prognosis; Tomography, X-Ray Computed.

Fig. 1. Coronavirus disease 2019 pneumonia showing an organizing pneumonia pattern in a 73-year-old man. (A, B) Serial axial CT scans obtained 5 days after symptom onset show patchy areas of ground glass opacity with a crazy-paving appearance distributed mainly in subpleural areas of both upper lungs. Visual CT score and total extents of automatic measurement of lung parenchymal abnormalities on initial CT images were 15 and 23.7%, respectively. (C, D) Follow-up axial CT scan obtained 1 months later showing partially improved lesions with residual subpleural lines.

CT = computed tomography.

Fig. 2. Coronavirus disease 2019 pneumonia with a diffuse alveolar damage pattern in a 72-year-old man. (A) Initial chest radiograph obtained 5 days after symptom onset showing bilateral diffuse ground glass opacities in both lungs. (B, C) Serial axial CT scans obtained on the same day as (A) showing bilateral diffuse ground-glass opacity with consolidation without specific zonal predominance in both lungs. Visual CT score and total extents of automatic measurement of lung parenchymal abnormalities on initial CT images were 24 and 47.8%, respectively. (D) Follow-up chest radiograph showing lesion progression. The patient died 14 days after chest radiograph.

CT = computed tomography.

Fig. 3. Receiver operating characteristic analysis of the predictive modelsfor severe clinical course of coronavirus disease 2019 with visual CT scores and artificial intelligence measurement.

CT = computed tomography.