A novel machine learning-based video classification approach to detect pneumonia in COVID-19 patients using lung ultrasound

Context: Efficiently diagnosing COVID-19-related pneumonia is of high clinical relevance. Point-of-care ultrasound allows detecting lung conditions via patterns of artifacts, such as clustered B-lines. Aims: The aim is to classify lung ultrasound videos into three categories: Normal (containing A-lines), interstitial abnormalities (B-lines), and confluent abnormalities (pleural effusion/consolidations) using a semi-automated approach. Settings and Design: This was a prospective observational study using 1530 videos in 300 patients presenting with clinical suspicion of COVID-19 pneumonia, where the data were collected and labeled by human experts versus machine learning. Subjects and Methods: Experts labeled each of the videos into one of the three categories. The labels were used to train a neural network to automatically perform the same classification. The proposed neural network uses a unique two-stream approach, one based on raw red-green-blue channel (RGB) input and the other consisting of velocity information. In this manner, both spatial and temporal ultrasound features can be captured. Statistical Analysis Used: A 5-fold cross-validation approach was utilized for the evaluation. Cohen's kappa and Gwet's AC1 metrics are calculated to measure the agreement with the human rater for the three categories. Cases are also divided into interstitial abnormalities (B-lines) and other (A-lines and confluent abnormalities) and precision-recall and receiver operating curve curves created. Results: This study demonstrated robustness in determining interstitial abnormalities, with a high F1 score of 0.86. For the human rater agreement for interstitial abnormalities versus the rest, the proposed method obtained a Gwet's AC1 metric of 0.88. Conclusions: The study demonstrates the use of a deep learning approach to classify artifacts contained in lung ultrasound videos in a robust manner.

POSA190 Socio-Economic Influences on County-Level Differences in COVID-19 Incidence During the Second Wave in Germany

Objectives: During fall and winter 2020/2021, before vaccine availability, Germany experienced a severe second wave of the COVID-19 pandemic. Daily cases grew exponentially in October/November (5Oct to 5Nov 2020, phase I), plateaued in November (6Nov to 6Dec 2020, phase II), peaked in late December, and declined in January/February (12Jan to 12Feb 2021, phase III). We investigated whether socio-economic characteristics (population density (inhabitants/km2;2019;“popDens”), household size (average number of persons/household;2011;“hhSize”), average living space (m2/inhabitant;2011;“livSpace”), education level (percentage of inhabitants with university-entrance qualification;2019;“Abitur”) and disposable income (EUR/inhabitant;2018;“income”)) predicted regional differences in incidence (cases per 100,000 inhabitants;“cases/100k”) in each of the three phases. We expected counties with greater living space, education level and disposable income to report lower incidence. Methods: County-level daily COVID-19 cases were extracted from RKI databases. County-level predictor variables were retrieved from public sources. For each phase, we computed a robust linear regression model with popDens, hhSize, livSpace, Abitur, and income as predictor variables for county-level cases/100k. Analyses were performed using statistical software R. Results: For phase I, cases/100k significantly increased with popDens (beta=0.17, p<0.001), hhSize (beta=0.41, p<0.001), and income (beta=0.20, p<0.001), and decreased with livSpace (beta=-0.22, p<0.001), R2=0.32. For phase II, cases/100k significantly increased with popDens (beta=0.12, p<0.001), hhSize (beta=0.39, p<0.001), and income (beta=0.15, p<0.001) and decreased with Abitur (beta=-0.15, p=0.002) and livSpace (beta=-0.24, p<0.001), R2=0.26. For phase III, cases/100k were significantly decreasing with popDens (beta=-0.06, p<0.001), hhSize (beta=-0.14, p=0.031), income (beta=-0.16, p<0.001), and livSpace (beta=-0.21, p<0.001), R2=0.08. Conclusions: Socio-economic regional differences influenced COVID-19 incidence in Germany´s second wave, however the explanatory power was low. The inclusion of influence factors was limited by data availability. The relevant factors differed between phase I/II and phase III. Our research does not claim a causative relationship between variables.

PIN78 County-level differences in COVID-19 cases and fatalities in Germany

Objectives: On January 27, 2020 the first COVID-19 case in Germany was confirmed. By June 22, 2020, the Robert Koch-Institute (RKI) published 190,359 confirmed cases (fatal: 8,885;recovered: 175,300). Objective was to analyse if the large regional differences in the cases per 100,000 inhabitants (casesp100k, range 33.9–1,566.8) are correlated with the number of physicians per 100,000 inhabitants (physiciansp100k) and / or the gross domestic product per capita (GDPpc). Methods: The number of cases and fatalities per county were extracted from the official source at the RKI website. These data were supplemented by the 2019 population, the 2017 GDPpc and the 2019 physiciansp100k for each county. We used a linear regression model with physiciansp100k, GDPpc and squared GDPpc (GDPpcsq) as explanatory variables for casesp100k. For fatalities per 100,000 (fatalitiesp100k), casesp100k were the explanatory variable. Calculations were performed with statistical software R. Results: Casesp100k per county were found to be significantly decreasing with physiciansp100k (coefficient = −0.4784;p-value = 0.0172) and a significant positive, non-linear relationship with GDPpC, (coefficient = −0.0109, p-value < 0.001;GDPpcsq, coefficient < 0.0000, p-value < 0.001). This means, that 10 additional physicians translate into 4.78 additional cases and an increase of 1,000 € at the average GDP per capita of 37,158 € to 6.34 additional cases (reducing to 1.85 at 2-times the average GDP). Fatalities per 100,000, were fully explained by casesp100k as the only explanatory variable (coefficient = 1.0000;p-value < 0.001). Conclusions: This research analysis the potential influence of socio-economic differences in German regions on COVID-19 cases/fatalities. Due to limited data availability on the county level it was not possible to analyse potential influence factors. In interpreting of these results, it needs to be kept in mind that our analysis captures correlation between variables and does not claim a causative relationship between the variables.