Research and Optimization of a Lightweight Refined Mask-Wearing Detection Algorithm Based on an Attention Mechanism

To address the current problems of the incomplete classification of mask-wearing detection data, small-target miss detection, and the insufficient feature extraction capabilities of lightweight networks dealing with complex faces, a lightweight method with an attention mechanism for detecting mask wearing is presented in this paper. This study incorporated an "incorrect_mask” category into the dataset to address incomplete classification. Additionally, the YOLOv4-tiny model was enhanced with a prediction feature layer and feature fusion execution, expanding the detection scale range and improving the performance on small targets. A CBAM attention module was then introduced into the feature enhancement network, which re-screened the feature information of the region of interest to retain important feature information and improve the feature extraction capabilities. Finally, a focal loss function and an improved mosaic data enhancement strategy were used to enhance the target classification performance. The experimental results of classifying three objects demonstrate that the lightweight model's detection speed was not compromised while achieving a 2.08% increase in the average classification precision, which was only 0.69% lower than that of the YOLOv4 network. Therefore, this approach effectively improves the detection effect of the lightweight network for mask-wearing.

Diagnostic Utility of a Ferritin-to-Procalcitonin Ratio to Differentiate Patients With COVID-19 From Those With Bacterial Pneumonia: A Multicenter Study.

BACKGROUND: There is an urgent need for accurate, rapid, inexpensive biomarkers that can differentiate coronavirus disease 2019 (COVID-19) from bacterial pneumonia. We assess the role of the ferritin-to-procalcitonin (F/P) ratio to classify pneumonia cases into those due to COVID-19 vs those due to bacterial pathogens. METHODS: This multicenter case-control study compared patients with COVID-19 with those with bacterial pneumonia, admitted between March 1 and May 31, 2020. Patients with COVID-19 and bacterial pneumonia co-infection were excluded. The F/P in patients with COVID-19 vs with bacterial pneumonia were compared. Receiver operating characteristic curve analysis determined the sensitivity and specificity of various cutoff F/P values for COVID-19 vs bacterial pneumonia. RESULTS: A total of 242 COVID-19 pneumonia cases and 34 bacterial pneumonia controls were included. Patients with COVID-19 pneumonia had a lower mean age (57.1 vs 64.4 years; P = .02) and a higher body mass index (30.74 vs 27.15 kg/m2; P = .02) compared with patients with bacterial pneumonia. Cases and controls had a similar proportion of women (47% vs 53%; P = .5), and COVID-19 patients had a higher prevalence of diabetes mellitus (32.6% vs 12%; P = .01). The median F/P was significantly higher in patients with COVID-19 (4037.5) compared with the F/P in bacterial pneumonia (802; P < .001). An F/P ≥877, used to diagnose COVID-19, resulted in a sensitivity of 85% and a specificity of 56%, with a positive predictive value of 93.2% and a likelihood ratio of 1.92. In multivariable analyses, an F/P ≥877 was associated with greater odds of identifying a COVID-19 case (odds ratio, 11.27; 95% CI, 4-31.2; P < .001). CONCLUSIONS: An F/P ≥877 increases the likelihood of COVID-19 pneumonia compared with bacterial pneumonia.

Diagnostic utility of a Ferritin-to-Procalcitonin Ratio to differentiate patients with COVID-19 from those with Bacterial Pneumonia: A multicenter study.

IMPORTANCE: There is a need to develop tools to differentiate COVID-19 from bacterial pneumonia at the time of clinical presentation before diagnostic testing is available. OBJECTIVE: To determine if the Ferritin-to-Procalcitonin ratio (F/P) can be used to differentiate COVID-19 from bacterial pneumonia. DESIGN: This case-control study compared patients with either COVID-19 or bacterial pneumonia, admitted between March 1 and May 31, 2020. Patients with COVID-19 and bacterial pneumonia co-infection were excluded. SETTING: A multicenter study conducted at three hospitals that included UCHealth and Phoebe Putney Memorial Hospital in the United States, and Yichang Central People's Hospital in China. PARTICIPANTS: A total of 242 cases with COVID-19 infection and 34 controls with bacterial pneumonia. MAIN OUTCOMES AND MEASURES: The F/P in patients with COVID-19 or with bacterial pneumonia were compared. Receiver operating characteristic analysis determined the sensitivity and specificity of various cut-off F/P values for the diagnosis of COVID-19 versus bacterial pneumonia. RESULTS: Patients with COVID-19 pneumonia had a lower mean age (57.11 vs 64.4 years, p=0.02) and a higher BMI (30.74 vs 27.15 kg/m 2 , p=0.02) compared to patients with bacterial pneumonia. Cases and controls had a similar proportion of women (47% vs 53%, p=0.5) and COVID-19 patients had a higher prevalence of diabetes mellitus (32.6% vs 12%, p=0.01). The median F/P was significantly higher in patients with COVID-19 (4037.5) compared to the F/P in bacterial pneumonia (802, p<0.001). An F/P ≥ 877 used to diagnose COVID-19 resulted in a sensitivity of 85% and a specificity of 56%, with a positive predictive value of 93.2%, and a likelihood ratio of 1.92. In multivariable analyses, an F/P ≥ 877 was associated with greater odds of identifying a COVID-19 case (OR: 11.27, CI: 4-31.2, p<0.001). CONCLUSIONS AND RELEVANCE: An F/P ≥ 877 increases the likelihood of COVID-19 pneumonia compared to bacterial pneumonia. Further research is needed to determine if obtaining ferritin and procalcitonin simultaneously at the time of clinical presentation has improved diagnostic value. Additional questions include whether an increased F/P and/or serial F/P associates with COVID-19 disease severity or outcomes.

Comparison of Venous Thromboembolism Risks Between COVID-19 Pneumonia and Community-Acquired Pneumonia Patients.

OBJECTIVE: The objectives were to investigate and compare the risks and incidences of venous thromboembolism (VTE) between the 2 groups of patients with coronavirus disease 2019 (COVID-19) pneumonia and community-acquired pneumonia (CAP). Approach and Results: Medical records of 616 pneumonia patients who were admitted to the Yichang Central People's Hospital in Hubei, China, from January 1 to March 23, 2020, were retrospectively reviewed. The patients with COVID-19 pneumonia were treated in the dedicated COVID-19 units, and the patients with CAP were admitted to regular hospital campus. Risks of VTE were assessed using the Padua prediction score. All the patients received pharmaceutical or mechanical VTE prophylaxis. VTE was diagnosed using Duplex ultrasound or computed tomography pulmonary angiogram. Differences between COVID-19 and CAP groups were compared statistically. All statistical tests were 2 sided, and P<0.05 was considered as statistically significant. All data managements and analyses were performed by IBM SPSS, version 24, software (SPSS, Inc, Chicago, IL). Of the 616 patients, 256 had COVID-19 pneumonia and 360 patients had CAP. The overall rate of VTE was 2% in COVID-19 pneumonia group and 3.6% in CAP group, respectively (P=0.229). In these two groups, 15.6% of the COVID-19 pneumonia patients and 10% of the CAP patients were categorized as high risk for VTE (Padua score, >4), which were significantly different (P=0.036). In those high-risk patients, the incidence of VTE was 12.5% in COVID-19 pneumonia group and 16.7% in CAP group (P=0.606). Subgroup analysis of the critically ill patients showed that VTE rate was 6.7% in COVID-19 group versus 13% in CAP group (P=0.484). In-hospital mortality of COVID-19 and CAP was 6.3% and 3.9%, respectively (P=0.180). CONCLUSIONS: Our study suggested that COVID-19 pneumonia was associated with hypercoagulable state. However, the rate of VTE in COVID-19 pneumonia patients was not significantly higher than that in CAP patients.