A Survey on DL Based Frameworks for COVID-19 Radiological Diagnosis

The ongoing Coronavirus disease (COVID-19) pandemic still necessitates emphasis on diagnosis and management of the outbreaks due to the emergence of new variants. This paper is an extensive survey on the implementation of Deep Learning (DL) models used for diagnosing COVID-19 from chest imaging, enriched with quantitative measures and regulatory aspects. The authors have searched, collated and categorised various models and techniques that reported different architectures with respect to COVID-19 diagnosis in the literature. This survey also briefs about quantifying metrics and the reported results are enumerated, also regulatory frameworks for public use of Artificial Intelligence (AI) in medical devices are comprehended. © 2022, IFIP International Federation for Information Processing.

Inflammation, Hyperglycemia, and Adverse Outcomes in Individuals With Diabetes Mellitus Hospitalized for COVID-19.

OBJECTIVE: Diabetes mellitus (DM) is a major risk factor for severe coronavirus disease 2019 (COVID-19) for reasons that are unclear. RESEARCH DESIGN AND METHODS: We leveraged the International Study of Inflammation in COVID-19 (ISIC), a multicenter observational study of 2,044 patients hospitalized with COVID-19, to characterize the impact of DM on in-hospital outcomes and assess the contribution of inflammation and hyperglycemia to the risk attributed to DM. We measured biomarkers of inflammation collected at hospital admission and collected glucose levels and insulin data throughout hospitalization. The primary outcome was the composite of in-hospital death, need for mechanical ventilation, and need for renal replacement therapy. RESULTS: Among participants (mean age 60 years, 58.2% males), those with DM (n = 686, 33.5%) had a significantly higher cumulative incidence of the primary outcome (37.8% vs. 28.6%) and higher levels of inflammatory biomarkers than those without DM. Among biomarkers, DM was only associated with higher soluble urokinase plasminogen activator receptor (suPAR) levels in multivariable analysis. Adjusting for suPAR levels abrogated the association between DM and the primary outcome (adjusted odds ratio 1.23 [95% CI 0.78, 1.37]). In mediation analysis, we estimated the proportion of the effect of DM on the primary outcome mediated by suPAR at 84.2%. Hyperglycemia and higher insulin doses were independent predictors of the primary outcome, with effect sizes unaffected by adjusting for suPAR levels. CONCLUSIONS: Our findings suggest that the association between DM and outcomes in COVID-19 is largely mediated by hyperinflammation as assessed by suPAR levels, while the impact of hyperglycemia is independent of inflammation.

Role of glycaemic control in predicting severity of illness in COVID 19 patients with diabetes mellitus

Introduction: Coronavirus Disease 2019 (COVID 19) pneumonia is a rapidly spreading disease and which causes morbidity and mortality of many patients. Diabetes mellitus is co-morbidity which is considered as the risk factor for COVID 19. Well-controlled diabetes is associated with better outcomes than poorly controlled diabetes. Measurement of glycated haemoglobin (HbA1c) is the standard method for assessing long term glycaemic control. Regardless of the level of hyperglycaemia, improvement in glycaemic control will lower the risk of diabetic complications. Aim: This study was conducted to identify the role of glycaemic control (HbA1c) in predicting the severity of illness in patients with COVID 19 pneumonia. Materials and Methods: This was a retrospective observational study of (51 diabetic and 51 were non diabetic) patients at Kamineni Academy of Medical Sciences, Hyderabad, India. The patients diagnosed with COVID 19 pneumonia, which includes both diabetics and non diabetics from June 2020 to September 2020. Patients age, sex, baseline HbA1c levels, and oxygen requirement during the hospital stay were analysed using Statistical Package for the Social Sciences (SPSS) software version 22.0. The Chi-Square test was used to analyse qualitative data and p-value significant at level <0.05. Results: In the study among diabetics (n=51), 20 (39.2%) were on room air, 24 (47.1%) required intermittent oxygen support, 3 (5.9%) high flow oxygen, and 4 (7.8%) non invasive ventilator support. Among non diabetics (n=51), 28 (54.9%) were on room air, 18 (35.3%) on intermittent oxygen, 2 (3.9%) high flow oxygen, and 3 (5.9%) Non Invasive Ventilator (NIV) support. It was observed that patients with HbA1c measurements with poor glycaemic control required more oxygen support during treatment in diabetics (p-value:0.469) Conclusion: In the present study, patients with poor glycaemic control required insignificantly, more oxygen support than patients with good glycaemic control.