ABSTRACT
Deep learning methods have been extensively investigated for rapid and precise computer-aided diagnosis during the outbreak of the COVID-19 epidemic. However, there are still remaining issues to be addressed, such as distinguishing COVID-19 in the complex scenario of multi-type pneumonia classification. In this paper, we aim to boost the COVID-19 diagnostic performance with more discriminative deep representations of COVID and non-COVID categories. We propose a novel COVID-19 diagnosis approach with contrastive representation learning to effectively capture the intra-class similarity and inter-class difference. Besides, we design an adaptive joint training strategy to integrate the classification loss, mixup loss, and contrastive loss. Through the joint loss function, we obtain the high-level representations which are highly discriminative in COVID-19 screening. Extensive experiments on two chest CT image datasets, i.e., CC-CCII dataset and COV19-CT-DB database, demonstrate the effectiveness of our proposed approach in COVID-19 diagnosis. Our method won the first prize in the ICCV 2021 Covid-19 Diagnosis Competition of AI-enabled Medical Image Analysis Workshop. Our code is publicly available at https://github.com/houjunlin/Team-FDVTS-COVID-Solution.
ABSTRACT
BACKGROUND: Nearly 20% novel coronavirus disease 2019 (COVID-19) patients have abnormal coagulation function. Padua prediction score (PPS) is a validated tools for venous thromboembolism (VTE) risk assessment. However, its clinical value in COVID-19 patients' evaluation was unclear. METHODS: We prospectively evaluated the VTE risk of COVID-19 patients using PPS. Demographic and clinical data were collected. Association of PPS with 28-day mortality was analyzed by multivariate logistic regression and Kaplan-Meier analysis. RESULTS: Two hundred and seventy-four continuous patients were enrolled, with total mortality of 17.2%. Patients in high PPS group, with significantly abnormal coagulation, have a higher levels of interleukin 6 (25.27 vs. 2.55 pg/ml, P < 0.001), prophylactic anticoagulation rate (60.7% vs. 6.5%, P < 0.001) and mortality (40.5% vs. 5.9%, P < 0.001) when compared with that in low PPS group. Critical patients showed higher PPS (6 vs. 2 score, P < 0.001) than that in severe patients. Multivariate logistic regression revealed the independent risk factors of in-hospital mortality included high PPS [odds ratio (OR): 7.35, 95% confidence interval (CI): 3.08-16.01], increased interleukin-6 (OR: 11.79, 95% CI: 5.45-26.20) and elevated d-dimer (OR: 4.65, 95% CI: 1.15-12.15). Kaplan-Meier analysis indicated patients with higher PPS had a significant survival disadvantage. Prophylactic anticoagulation in higher PPS patients shows a mild advantage of mortality but without statistical significance (37.1% vs. 45.7%, P = 0.42). CONCLUSION: Higher PPS associated with in-hospital poor prognosis in COVID-19 patients. Prophylactic anticoagulation showed a mild advantage of mortality in COVID-19 patients with higher PPS, but it remain to need further investigation.