Automatic CT Quantification of Coronavirus Disease 2019 pneumonia: An international collaborative development, validation, and clinical implication (preprint)

Objectives    We aimed to develop and validate the automatic quantification of COVID-19 pneumonia on CT images. Methods This retrospective study included 176 chest CT scans of 131 COVID-19 patients from 13 Korean and Chinese institutions. Two experienced radiologists semi-automatically drew pneumonia, preparing 49,830 positive and negative CT slices to develop the 2D U-Net for segmenting pneumonia. The 2D U-Net was distributed as downloadable software. External validation for quantifications’ accuracy was performed using Japanese, Italian, Radiopaedia, Chinese datasets. Primary measures for the accuracy of the network were correlation coefficients for extent (%) and weight (g) of pneumonia. Logistic regression analyses were performed to evaluate the clinical implication of the extent and weight regarding the presence of symptoms in the Japanese dataset and the occurrence of composite outcome in the Spanish dataset. Results In the internal validation dataset, the intraclass correlation coefficients between the 2D U-Net and reference values for the extent and weight were 0.990 and 0.993, respectively. In the Japanese dataset, the Pearson correlation coefficients between the U-Net outcomes and visual CT severity scores were 0.908 and 0.899, respectively. In the other external validation datasets, the intraclass correlation coefficients between the U-Net and reference values were between 0.951-0.970 (extent) and between 0.970-0.995 (weight), respectively. In multivariate logistic regression analyses, the extent and weight of pneumonia were independently associated with symptoms (OR, 4.142 and 4.434; p=.013 and .009, respectively), and poor prognosis (OR, 7.446 and 4.677; p=.004 and .029, respectively).Conclusions CT extent and weight of COVID-19 pneumonia were automatically quantifiable and independently associated with symptoms and prognosis. 

Pre-existing liver disease is associated with poor outcome in patients with SARS CoV2 infection; The APCOLIS Study (APASL COVID-19 Liver Injury Spectrum Study) (preprint)

Background and Aims: COVID-19 is a dominant pulmonary disease, with multisystem involvement, depending upon co morbidities. Its profile in patients with pre-existing chronic liver disease (CLD) is largely unknown. We studied the liver injury patterns of SARS-Cov-2 in CLD patients, with or without cirrhosis. Methods: Data was collected from 13 Asian countries on patients with CLD, known or newly diagnosed, with confirmed COVID-19. Result: Altogether, 228 patients [185 CLD without cirrhosis and 43 with cirrhosis] were enrolled, with comorbidities in nearly 80%. Metabolism associated fatty liver disease (113, 61%) and viral etiology (26, 60%) were common. In CLD without cirrhosis, diabetes [57.7% vs 39.7%, OR=2.1(1.1-3.7), p=0.01] and in cirrhotics, obesity, [64.3% vs. 17.2%, OR=8.1(1.9-38.8), p=0.002) predisposed more to liver injury than those without these. Forty three percent of CLD without cirrhosis presented as acute liver injury and 20% cirrhotics presented with either acute-on-chronic liver failure [5(11.6%)] or acute decompensation [4(9%)]. Liver related complications increased (p<0.05) with stage of liver disease; a Child-Turcotte Pugh score of 9 or more at presentation predicted high mortality [AUROC-0.94, HR=19.2(95CI 2.3-163.3), p<0.001, sensitivity 85.7% and specificity 94.4%). In decompensated cirrhotics, the liver injury was progressive in 57% patients, with 43% mortality. Rising bilirubin and AST/ALT ratio predicted mortality among cirrhosis. Conclusions: SARS-Cov-2 infection causes significant liver injury in CLD patients, decompensating one fifth of cirrhosis, and worsening the clinical status of the already decompensated. The CLD patients with diabetes and obesity are more vulnerable and should be closely monitored. 

Clinical course and risk factors for mortality of COVID-19 patients with pre-existing cirrhosis: A multicenter cohort study (preprint)

Background: Patients with pre-existing cirrhosis are considered at increased risk of severe coronavirus disease 2019 (COVID-19) but the clinical course in these patients has not yet been reported. This study aimed to provide a detailed report of the clinical characteristics and outcomes among COVID-19 patients with pre-existing cirrhosis. Methods: In this retrospective, multicenter cohort study, we consecutively included all adult inpatients with laboratory-confirmed COVID-19 and pre-existing cirrhosis that had been discharged or had died by 24 March 2020 from 16 designated hospitals in China. Demographic, clinical, laboratory and radiographic findings on admission, treatment, complications during hospitalization and clinical outcomes were collected and compared between survivors and non-survivors. Findings: Twenty-one patients were included in this study, of whom 16 were cured and 5 died in hospital. Seventeen patients had compensated cirrhosis and hepatitis B virus infection was the most common etiology. Lymphocyte and platelet counts were lower, and direct bilirubin levels were higher in patients who died than those who survived (p= 0.040, 0.032, and 0.006, respectively). Acute respiratory distress syndrome and secondary infection were both the most frequently observed complications. Only one patient developed acute on chronic liver failure. Of the 5 non-survivors, all patients developed acute respiratory distress syndrome and 2 patients progressed to multiple organ dysfunction syndrome. Interpretation: Lower lymphocyte and platelet counts, and higher direct bilirubin level might represent poor prognostic indicators in SARS-CoV-2-infected patients with pre-existing cirrhosis.

Machine learning-based CT radiomics model for predicting hospital stay in patients with pneumonia associated with SARS-CoV-2 infection: A multicenter study (preprint)

Abstract Objectives To develop and test machine learning-based CT radiomics models for predicting hospital stay in patients with pneumonia associated with SARS-CoV-2 infection. Design Cross-sectional Setting Multicenter Participants A total of 52 patients with laboratory-confirmed SARS-CoV-2 infection and their initial CT images were enrolled from 5 designated hospitals in Ankang, Lishui, Zhenjiang, Lanzhou, and Linxia between January 23, 2020 and February 8, 2020. As of February 20, patients remained in hospital or with non-findings in CT were excluded. Therefore, 31 patients with 72 lesion segments were included in the final analysis. Intervention CT radiomics models based on logistic regression (LR) and random forest (RF) were developed on features extracted from pneumonia lesions in training and inter-validation datasets. The predictive performance was further evaluated in test dataset on lung lobe- and patients-level. Main outcomes Short-term hospital stay ([≤]10 days) and long-term hospital stay (>10 days). Results The CT radiomics models based on 6 second-order features were effective in discriminating short- and long-term hospital stay in patients with pneumonia associated with SARS-CoV-2 infection, with areas under the curves of 0.97 (95%CI 0.83-1.0) and 0.92 (95%CI 0.67-1.0) by LR and RF, respectively, in the test dataset. The LR model showed a sensitivity and specificity of 1.0 and 0.89, and the RF model showed similar performance with sensitivity and specificity of 0.75 and 1.0 in test dataset. Conclusions The machine learning-based CT radiomics models showed feasibility and accuracy for predicting hospital stay in patients with pneumonia associated with SARS-CoV-2 infection.