COVID-19 feature detection with deep neural networks trained on simulated lung ultrasound B-mode images

Deep learning has been implemented to detect COVID-19 features in lung ultrasound B-mode images. However, previous work primarily relied on in vivo images as the training data, which suffers from limited access to required manual labeling of thousands of training image examples. To avoid this manual labeling, which is tedious and time consuming, we propose the detection of in vivo COVID-19 features (i.e., A-line, B-line, consolidation) with deep neural networks (DNNs) trained on simulated B-mode images. The simulation-trained DNNs were tested on in vivo B-mode images from healthy subjects and COVID-19 patients. With data augmentation included during the training process, Dice similarity coefficients (DSCs) between ground truth and DNN predictions were maximized, producing mean ± standard deviatio values as high as 0.48 ± 0.29, 0.45 ± 0.25, and 0.46 ± 0.35 when segmenting in vivo A-line, B-line, and consolidation features, respectively. Results demonstrate that simulation-trained DNNs are a promising alternative to training with real patient data when segmenting in vivo COVID-19 features. © 2022 IEEE.

Do lung ultrasound abnormalities change during hospitalization for COVID-19?

Background. While point-of-care ultrasound (POCUS) has been used to track disease resolution, temporal trends in lung ultrasound (LUS) findings among hospitalized patients with COVID-19 is not well-characterized. Methods. We studied 413 LUS scans in 244 participants >= 18 years of age hospitalized for COVID-19 pneumonia within 28 days of symptom onset from April, 2020 until September, 2021 at the Johns Hopkins Hospital, Baltimore Maryland. All patients were scanned using a 12-lung zone protocol and repeat scans were obtained in 3 days (N=114), 7 days (N=53), and weekly (N=9) from the initial scan. Participants were followed to determine clinical outcomes until hospital discharge and vital status at 28-days. Ultrasounds were independently reviewed for lung artifacts, and the composite mean LUS score (ranging from 0 to 3) across lung zones was determined. Trends of mean LUS scores and%lung fields with A-lines (indicating proportion of normal lung fields) were plotted by peak severity (mild, moderate, and severe defined by the World Health Organization Ordinal Scale) over time from symptom onset. Differences in mean LUS score or % A-lines changes over time between peak severity levels were evaluated using a Kruskal-Wallis test and linear mixed-effected models with an exchangeable correlation structure. Results. Among 244 patients in our cohort (mean age of 58.2 (SD 15.0) years, and 55.7% female) (Table 1), there was no change in average mean LUS scores between the first two visits by severity groups (Figure 1;Kruskal-Wallis p=0.63). Mean LUS scores were elevated by 0.22 (p< 0.001) in a dose-response manner regardless of duration of illness, but there was no change over time associated with peak severity (p=0.73). Similarly, percentage of A-lines were in 13.9% less lung fields for each increase in peak severity (p< 0.001;Figure 2) regardless of duration of illness. However, a change in mean LUS score did not differ significantly among peak severity levels (p=0.36). Conclusion. Mean LUS scores correlated with clinical severity among hospitalized adults when assessed cross-sectionally, however mean LUS score did not change or differ between peak severity levels over the time course of hospitalization. These results do not support serial LUS scans to monitor disease progression.

Corrigendum: Take a "selfie": Examining how leaders emerge from leader self-awareness, self-leadership, and self-efficacy.

[This corrects the article DOI: 10.3389/fpsyg.2021.635085.].

PROVIDER COMMUNICATION SURROUNDING ALCOHOL USE DISORDER IN HOSPITALIZED ADULTS WITH ALCOHOL-ASSOCIATED LIVER DISEASE

Background/Significance: Alcohol-associated liver disease (ALD) is now the leading indication for liver transplantation (LT). Disease burden, as well as hospitalizations, have risen during the SARS-CoV2 pandemic. Alcohol use disorder (AUD) treatments are underutilized in patients with ALD, which is an important focus for quality improvement at LT centers. Our aim was to describe current communication practices surrounding AUD care for hospitalized adults with ALD at a large-volume single LT center without integrated specialty addiction services. Methods: We performed semi-structured interviews with healthcare professionals providing care to hospitalized patients on an inpatient LT service from April to June 2021 using a videoconferencing platform. Interview guides focused on current processes for discussing and connecting hospitalized patients to AUD treatment;these were pilot-tested prior to use. Audio-files were professionally transcribed and imported to NVivo 12 (QSR International). Two qualitative researchers developed a codebook corresponding to 4 major domains of AUD care (discussing AUD, pharmacotherapy, behavioral therapy, and referrals to specialty addiction services) and assigned codes to all transcripts, with regular meetings to resolve discrepancies. A combination of inductive and deductive approaches was used to generate non-overlapping themes. Results: We interviewed 17 providers. Six main themes were generated (Table 1). When discussing AUD, most providers, other than social workers, rarely assessed patients' insight and motivations for drinking. Most providers were uncomfortable with the topic and encouraged behavior change by promoting guilt and fear of the consequences of continued alcohol use. Discussions about pharmacotherapy were rare and limited to few providers offering baclofen. Alcoholics Anonymous (AA) was offered as standard form of non-pharmacologic therapy, but alternative options were rarely presented. Discussions about referrals to specialty addiction services were emphasized most consistently to post-LT patients only;in general, providers expressed it was the patient's responsibility to set up appointments. Conclusions: Communication about AUD from the LT team was rare and focused on narrow topics, including negative consequences of drinking, AA participation, and importance to self-navigate specialty addiction referrals. Opportunities for improving communication include framing and exploring addiction in less stigmatizing ways, discussing the full range of evidence-based pharmacologic and non-pharmacologic treatments for AUD, and providing more direction for setting up specialty addiction referrals. These factors, including advocating for reduced insurance-related barriers for specialty addiction services, should be considered by LT centers when performing needs assessments for improving AUD care. (Table Presented) Table 1: Themes and Quotes from Providers Regarding Current Practices Surrounding AUD

INCIDENCE AND OUTCOMES OF COVID-19 IN POST-LIVER TRANSPLANT RECIPIENTS AFTER COVID-19 VACCINATION IN A LARGE INTEGRATED HEALTHCARE SYSTEM

Background There is limited data on the efficacy of SARS-CoV-2 vaccinations on the immunosuppressed population—especially in the liver transplant (LT) population. A study in Israel found that only 47% of LT recipients developed adequate antibodies against the virus while another study in Baltimore, MD found an immune response of 81%. Moreover, early studies in San Diego, CA and Miami, FL on the outcomes of COVID-19 disease among solid organ transplant recipients showed reductions in symptomatic disease of 75 to 80%. We aim to identify the incidence and outcomes of COVID-19 disease in fully vaccinated LT recipients in a large cohort of LT recipients. Methods In a large integrated healthcare system in Southern California with a population of 4.3 million active members aged 18 or older, data was extracted from the electronic health record (EHR) and transplant registry. COVID-19 disease was identified by a positive polymerase chain reaction (PCR) test for SARS-CoV-2. We defined fully vaccinated as 14 days after the 2nd dose of Pfizer or Moderna vaccines or after the 1st dose of Johnson and Johnson vaccine. Chi square analysis was used to compare the difference between 2 groups. Results We identified 1271 active members who had received a LT as of 12/1/2021. Among LT recipients, 90.6% (1152/1271) had received at least one dose of a COVID-19 vaccination, 89.1% (1132/1271) were fully vaccinated, and 58.6% (745/1271) had received booster vaccinations. Between 3/1/20 and 11/30/21, 172 (13.5%) LT recipients had been infected with COVID-19 disease, of which, 37 (3.3%) were infected after being fully vaccinated. Of those infected after being fully vaccinated, 38.9% (15/37) were female. The mean age was 58.7 ± 9.8. 62.2% had diabetes, 73.0% had hypertension, and the mean body mass index (BMI) was 29.3 ± 5.8. 24.3% (9/37) were hospitalized for COVID-19-related illness. The case fatality of COVID-19 was 2.7% (1/37) in post-vaccinated LT recipients compared with 7.4% (10/135) who were unvaccinated prior to infection (OR 0.35, 95% CI 0.04-2.80, p = 0.27). The patient who passed away from COVID-19 after vaccination (diagnosed 4 months after receiving the second dose) had chronic kidney disease and obesity (BMI 42). Conclusions In our large cohort of LT recipients, a significant proportion were fully vaccinated, and the majority had received a booster vaccination dose. A small proportion of LT recipients were infected with COVID-19 disease after being fully vaccinated for COVID-19. The case fatality rate, although not statistically significant, of patients infected post-vaccination was lower compared to unvaccinated patients. More research is needed on the long-term outcomes of COVID-19 and vaccine efficacy in this high-risk population.

Multi-stage investigation of deep neural networks for COVID-19 B-line feature detection in simulated and in vivo ultrasound images

COVID-19 is a highly infectious disease with high morbidity and mortality, requiring tools to support rapid triage and risk stratification. In response, deep learning has demonstrated great potential to quicklyand autonomously detect COVID-19 features in lung ultrasound B-mode images. However, no previous work considers the application of these deep learning models to signal processing stages that occur prior to traditional ultrasound B-mode image formation. Considering the multiple signal processing stages required to achieve ultrasound B-mode images, our research objective is to investigate the most appropriate stage for our deep learning approach to COVID-19 B-line feature detection, starting with raw channel data received by an ultrasound transducer. Results demonstrate that for our given training and testing configuration, the maximum Dice similarity coefficient (DSC) was produced by B-mode images (DSC = 0.996) when compared with three alternative image formation stages that can serve as network inputs: (1) raw in-phase and quadrature (IQ) data before beamforming, (2) beamformed IQ data, (3) envelope detected IQ data. The best-performing simulation-trained network was tested on in vivo B-mode images of COVID-19 patients, ultimately achieving 76% accuracy to detect the same (82% of cases) or more (18% of cases) B-line features when compared to B-line feature detection by human observers interpreting B-mode images. Results are promising to proceed with future COVID-19 B-line feature detection using ultrasound B-mode images as the input to deep learning models. © 2022 SPIE.

An Innovative Approach to Modeling VIPER Rover Software Life Cycle Cost

NASAs Volatiles Investigating Polar Exploration Rover (VIPER) will be the first robotic mission to prospect for water ice near the south pole of the Moon in late 2023 on a 100 Earth-day mission. The information that the VIPER rover provides will help improve understanding of the composition, distribution, and accessibility of Lunar polar volatiles and will help determine howthe Moons resources can support future human space exploration. VIPER, however, represents a radical departure from the way that NASA has traditionally developed planetary robotic missions. A key consequence of these differences is that estimating the cst of VIPERs rover software is challenging and complex. For example, VIPER is being developed using management procedures typically applied to NASA research and technology projects, rather than space flight programs. In addition, key portions of the rovers software are being designed as ground software to run on mission control computers (rather than on board the rover as flight software as with prior planetary missions) taking advantage of continuous, interactive data communications between the Moon and Earth and higher performance computing available on the ground. Moreover, the rovers software is being engineered using Agile software development practices and incorporates a significant amount of open-source code rather than following traditional (spir al, waterfall, etc.) development methods and mouse code. In this paper, we present an innovative process to estimate the life cycle cost of VIPERs rover software. We first describe how we modeled the architecture and code counts for three software elements: Rover Flight Software (RFSW), Rover Ground Software (RGSW), and Rover Simulation Software (RSIM). We then discuss key challenges and unique aspects of our approach, such as the lack of Lunar rover analogies, the need to integrate and test large opersource software, and the strategie developed to account for use of nonspace flight management practices and the impact of the COVID-19 pandemic We conclude with a summary of our results, including cumulative distribution, nearest neighbors and clusteanalysis, as well as heuristics used to confirm the reasonableness of the cost estimate.

An Innovative Approach to Modeling VIPER Rover Software Life Cycle Cost

NASA's 'Volatiles Investigating Polar Exploration Rover' (VIPER) will be the first robotic mission to prospect for water ice near the south pole of the Moon in late 2023 on a 100-Earth-day mission. The information that the VIPER rover provides will help improve understanding of the composition, distribution, and accessibility of Lunar polar volatiles and will help determine how the Moon's resources can support future human space exploration. VIPER, however, represents a radical departure from the way that NASA has traditionally developed planetary robotic missions. A key consequence of these differences is that estimating the cost of VIPER's rover software is challenging and complex. For example, VIPER is being developed using management procedures typically applied to NASA research and technology projects, rather than space flight programs. In addition, key portions of the rover's software are being designed as ground software to run on mission control computers (rather than onboard the rover as flight software as with prior planetary missions) taking advantage of continuous, interactive data communications between the Moon and Earth and higher performance computing available on the ground. Moreover, the rover's software is being engineered using Agile software development practices and incorporates a significant amount of open-source code, rather than following traditional (spiral, waterfall, etc.) development methods and in-house code. In this paper, we present an innovative process to estimate the life cycle cost of VIPER's rover software. We first describe how we modeled the architecture and code counts for three software elements: Rover Flight Software (RFSW), Rover Ground Software (RGSW), and Rover Simulation Software (RSIM). We then discuss key challenges and unique aspects of our approach, such as the lack of Lunar rover analogies, the need to integrate and test large open source software, and the strategies developed to account for use of non-space flight management practices and the impact of the COVID-19 pandemic. We conclude with a summary of our results, including cumulative distribution, nearest neighbors and cluster analysis, as well as heuristics used to confirm the reasonableness of the cost estimate. © 2021 IEEE.