Recognition System of Multi-model Chest X-Ray Diseases Based on YOLO v4

We use three models to build and design a multi-model identification system, and the identification results between the models are verified to increase accuracy. The identified lung diseases are classified into four categories, namely Normal, COVID-19, Tuberculosis, and Viral Pneumonia cases. After the user uploads the chest X-ray image, the system displays the results of the three identification types, and the calculation time is about 5 to 10 s. The accuracy of the multi-model system is better than that of the single-model system. If the Normal cases are included, the specificity is 77.41% for the traditional single-model system and 89.81% for the multi-model system. Additionally, if Normal cases are excluded, the F1 score is 70.00% for the single-model system and 80.7% for the multi-model system. Compared with the neural network with Faster R-CNN F1-Score of 90%, Mask R-CNN F1-Score of 85% and resNet-50 F1-Score of 80% are obtained. © 2022 IEEE.

Comparison of Scoring systems in Predicting Clinical Outcomes in Patients Hospitalized with COVID 19

Background. Previous scoring systems have been proposed to predict COVID19 outcomes, however none have been universally adopted. Two scoring systems of interest are Monoclonal Antibody Screening Score (MASS) and Oral Antiviral and Monoclonal Antibody Screening Score (OMASS).MASS prioritized patientsfor outpatient monoclonal antibody treatment based on risk of hospitalization, and OMASS was a modified version of MASS used to prioritize outpatient oral antivirals. We created a modified scoring system (UCH2021) incorporating vaccination status. These scores (table 1) have not been used to predict in-hospital clinical outcomes. We investigate these systems' abilities to predict mortality and oxygen requirements in hospitalized COVID19 patients. They do not require blood tests and allow for more rapid triage. Table 1: MASS, OMASS, UCH2021 Scoring Criteria Methods. A retrospective chart review was performed on 133 patients in two tertiary care centers between March and Sept. 2020 with RT-PCR confirmed SARS CoV2. Baseline risk factors were collected and MASS, OMASS, and UCH2021 were calculated. Primary outcomes included mortality, need for intubation, and need for supplemental oxygen >6L during hospitalization. Secondary analysis assessed if any individual risk factors were associated with those outcomes. These systems were evaluated via area under the curve calculations. Two groups based on an outcome were compared using two-sample t-tests for continuous variables and Fisher's exact tests for categorical variables. Results. All three systems demonstrated some discriminative power for mortality (table 2), but not for oxygen and intubation requirements. There was statistically significant difference in age between survivors and deceased (table 3), and BMI for oxygen requirements (table 4). Other risk factors were not predictive of mortality or oxygen requirement. Table 2: MASS, OMASS, UCH2021 Scores and Mortality in Hospitalized COVID19 Patients Table 3: Age and Mortality in Hospitalized COVID19 Patients Table 4: BMI and Oxygen Requirements in Hospitalized COVID19 Patients Conclusion. The MASS, OMASS, and UCH2021 score all had predictive power in determining in-hospital mortality, with moderate accuracy, however none were predictive of oxygen requirements. Age and BMI were also good predictors of mortality and oxygen requirements respectively. This study was completed prior to vaccine distribution in the US. Further studies would be helpful to assess if UCH2021 score has greater discriminative power in samples with vaccinated patients.

Control of COVID-19 Pandemic: Vaccination Strategies Simulation under Probabilistic Node-Level Model

This paper aims at constructing a probabilistic node-level time-dependent contagious disease spreading model for coronavirus disease (COVID-19) pandemic which is called SEINRVseinr by introducing exposed and asymptomatic infectious state, imperfect vaccination, reinfected possibility and weighted undirected graph for social network into the traditional probabilistic node-level Susceptible-Infectious-Recovered (SIR) network model. This paper simulates the effectiveness of five vaccination strategies (including random base, degree-target base, random acquaintance, first-neighbor and second neighbor strategies) in random network, small world network and scale-free network. Compared with the benchmark model, the results show that random acquaintance strategy is efficient strategy and neighbors' strategies perform better in certain interval. © 2021 IEEE.

38 Dementia, Apoe and COVID-19 Severity

Introduction During the COVID-19 pandemic, pre-existing dementia was associated with a 3x increase in risk of hospitalisation and (25.6%) of COVID-19 related deaths had dementia. However, it is unclear whether people living with dementia are at higher risk of COVID-19 due to dementia or whether there may be a biologically plausible link between dementia and COVID-19. The ApoE e4 allele is highly associated with dementia. We aimed to test the COVID-19 risk associated with dementia and the association between ApoE e4e4 allele and COVID-19 with the aim of clarifying biological vulnerability. Methods UK Biobank (England) participants baseline (2006 to 2010), plus secondary care data to 2017. Separate analysis tested dementia and ApoE genotype association with COVID-19 status (16th March-31st May 2020) or mortality (to March 31, 2020, plus incomplete deaths from April, 2020) in logistic models, adjusted for demographics and technical covariates. Results In 269,070 participants aged 65+, including 507(0.2%) hospitalized COVID-19 patients, those with pre-existing dementia were at increased risk of being hospitalized for COVID-19 (OR = 3.50 95% CI 1.93 to 6.34) and also for COVID-19 and death (OR = 7.30 95% CI 3.28–16.21). In 375,689 European-ancestry UKB participants, ApoE e4e4 homozygotes were more likely to be COVID-19 test positives (reaching genome-wide significance: OR = 2.24, 95% CI:1.72–2.93, p = 3.24 × 10–9) and of mortality with test-confirmed COVID-19 (OR = 4.29, 95% CI: 2.38–7.72, p = 1.22 × 10–6), compared to e3e3s homozygotes. The associations were little changed in subsets of participants who were free of diseases associated with ApoE e4 and COVID-19 severity. Conclusion Dementia was found to be disproportionally common in older adults who develop severe COVID-19. We have shown a plausible genetic pathway of increased COVID-19 risk with dementia, therefore suggesting that the positive association between dementia and COVID-19 is not just the result of high cases of COVID-19 in care homes.

DEMENTIA, ApoE AND COVID-19 SEVERITY...British Geriatrics Society Autumn Meeting, November 25-27 2020 (Virtual)

Introduction: During the COVID-19 pandemic, pre-existing dementia was associated with a 3x increase in risk of hospitalisation and (25.6%) of COVID-19 related deaths had dementia. However, it is unclear whether people living with dementia are at higher risk of COVID-19 due to dementia or whether there may be a biologically plausible link between dementia and COVID-19. The ApoE e4 allele is highly associated with dementia. We aimed to test the COVID-19 risk associated with dementia and the association between ApoE e4e4 allele and COVID-19 with the aim of clarifying biological vulnerability. Methods: UK Biobank (England) participants baseline (2006 to 2010), plus secondary care data to 2017. Separate analysis tested dementia and ApoE genotype association with COVID-19 status (16th March-31st May 2020) or mortality (to March 31, 2020, plus incomplete deaths from April, 2020) in logistic models, adjusted for demographics and technical covariates. Results: In 269,070 participants aged 65+, including 507(0.2%) hospitalized COVID- 19 patients, those with pre-existing dementia were at increased risk of being hospitalized for COVID-19 (OR=3.50 95% CI 1.93 to 6.34) and also for COVID-19 and death (OR=7.30 95% CI 3.28–16.21). In 375,689 European-ancestry UKB participants, ApoE e4e4 homozygotes were more likely to be COVID-19 test positives (reaching genome-wide significance: OR=2.24, 95% CI:1.72–2.93, p=3.24×10–9) and of mortality with testconfirmed COVID-19 (OR=4.29, 95% CI: 2.38–7.72, p=1.22×10–6), compared to e3e3s homozygotes.The associations were little changed in subsets of participants who were free of diseases associated with ApoE e4 and COVID-19 severity. Conclusion: Dementia was found to be disproportionally common in older adults who develop severe COVID-19. We have shown a plausible genetic pathway of increased COVID-19 risk with dementia, therefore suggesting that the positive association between dementia and COVID-19 is not just the result of high cases of COVID-19 in care homes.