Unsupervised Multi-Latent Space RL Framework for Video Summarization in Ultrasound Imaging.

The COVID-19 pandemic has highlighted the need for a tool to speed up triage in ultrasound scans and provide clinicians with fast access to relevant information. To this end, we propose a new unsupervised reinforcement learning (RL) framework with novel rewards to facilitate unsupervised learning by avoiding tedious and impractical manual labelling for summarizing ultrasound videos. The proposed framework is capable of delivering video summaries with classification labels and segmentations of key landmarks which enhances its utility as a triage tool in the emergency department (ED) and for use in telemedicine. Using an attention ensemble of encoders, the high dimensional image is projected into a low dimensional latent space in terms of: a) reduced distance with a normal or abnormal class (classifier encoder), b) following a topology of landmarks (segmentation encoder), and c) the distance or topology agnostic latent representation (autoencoders). The summarization network is implemented using a bi-directional long short term memory (Bi-LSTM) which utilizes the latent space representation from the encoder. Validation is performed on lung ultrasound (LUS), that typically represent potential use cases in telemedicine and ED triage acquired from different medical centers across geographies (India and Spain). The proposed approach trained and tested on 126 LUS videos showed high agreement with the ground truth with an average precision of over 80% and average F1 score of well over 44 ±1.7 %. The approach resulted in an average reduction in storage space of 77% which can ease bandwidth and storage requirements in telemedicine.

Gastrointestinal organoids in the study of viral infections.

Viruses are among the most prevalent enteric pathogens. While virologists historically relied on cell lines and animal models, human intestinal organoids (HIOs) continue to grow in popularity. HIOs are non-transformed, stem cell derived, ex vivo cell cultures that maintain the cell type diversity of the intestinal epithelium. They offer higher throughput than standard animal models while more accurately mimicking the native tissue of infection than transformed cell lines. Here, we review recent literature that highlights virological advances facilitated by HIOs. We discuss the variations and limitations of HIOs, but also how HIOs have allowed for the cultivation of previously uncultivatable viruses and how they have offered insight into tropism, entry, replication kinetics, and host-pathogen interactions. In each case, we discuss exemplary viruses and archetypal studies. We discuss how the speed and flexibility of HIO-based studies contributed to our knowledge of SARS-CoV-2 and anti-viral therapeutics. Finally, we discuss current limitations of HIOs and future directions to overcome these.

Gastrointestinal organoids in the study of viral infections.

Viruses are among the most prevalent enteric pathogens. While virologists historically relied on cell lines and animal models, human intestinal organoids (HIOs) continue to grow in popularity. HIOs are non-transformed, stem cell derived, ex vivo cell cultures that maintain the cell type diversity of the intestinal epithelium. They offer higher throughput than standard animal models while more accurately mimicking the native tissue of infection than transformed cell lines. Here, we review recent literature that highlights virological advances facilitated by HIOs. We discuss the variations and limitations of HIOs, but also how HIOs have allowed for the cultivation of previously uncultivatable viruses and how they have offered insight into tropism, entry, replication kinetics, and host-pathogen interactions. In each case, we discuss exemplary viruses and archetypal studies. We discuss how the speed and flexibility of HIO-based studies contributed to our knowledge of SARS-CoV-2 and anti-viral therapeutics. Finally, we discuss current limitations of HIOs and future directions to overcome these.

Unsupervised Multi-Latent Space RL Framework for Video Summarization in Ultrasound Imaging.

The COVID-19 pandemic has highlighted the need for a tool to speed up triage in ultrasound scans and provide clinicians with fast access to relevant information. To this end, we propose a new unsupervised reinforcement learning (RL) framework with novel rewards to facilitate unsupervised learning by avoiding tedious and impractical manual labelling for summarizing ultrasound videos. The proposed framework is capable of delivering video summaries with classification labels and segmentations of key landmarks which enhances its utility as a triage tool in the emergency department (ED) and for use in telemedicine. Using an attention ensemble of encoders, the high dimensional image is projected into a low dimensional latent space in terms of: a) reduced distance with a normal or abnormal class (classifier encoder), b) following a topology of landmarks (segmentation encoder), and c) the distance or topology agnostic latent representation (autoencoders). The summarization network is implemented using a bi-directional long short term memory (Bi-LSTM) which utilizes the latent space representation from the encoder. Validation is performed on lung ultrasound (LUS), that typically represent potential use cases in telemedicine and ED triage acquired from different medical centers across geographies (India and Spain). The proposed approach trained and tested on 126 LUS videos showed high agreement with the ground truth with an average precision of over 80% and average F1 score of well over 44 ±1.7 %. The approach resulted in an average reduction in storage space of 77% which can ease bandwidth and storage requirements in telemedicine.

COVID-19 susceptibility and clinical outcomes in autoimmune inflammatory rheumatic diseases (AIRDs): a systematic review and meta-analysis.

OBJECTIVE: This meta-analysis aims to assess the susceptibility to and clinical outcomes of COVID-19 in autoimmune inflammatory rheumatic disease (AIRD) and following AIRD drug use. MATERIALS AND METHODS: We included observational and case-controlled studies assessing susceptibility and clinical outcomes of COVID-19 in patients with AIRD as well as the clinical outcomes of COVID-19 with or without use of steroids and conventional synthetic disease-modifying antirheumatic drugs (csDMARDs). RESULTS: Meta-analysis including three studies showed that patients with AIRD are not more susceptible to COVID-19 compared to patients without AIRD or the general population (OR: 1.11, 95% CI: 0.58 to 2.14). Incidence of severe outcomes of COVID-19 (OR: 1.34, 95% CI: 0.76 to 2.35) and COVID-19 related death (OR: 1.21, 95% CI: 0.68 to 2.16) also did not show significant difference. The clinical outcomes of COVID-19 among AIRD patients with and without csDMARD or steroid showed that both use of steroid (OR: 1.69, 95% CI: 0.96 to 2.98) or csDMARD (OR: 1.35, 95% CI: 0.63 to 3.08) had no effect on clinical outcomes of COVID-19. CONCLUSIONS: AIRD does not increase susceptibility to COVID-19, not affecting the clinical outcome of COVID-19. Similarly, the use of steroids or csDMARDs for AIRD does not worsen the clinical outcome.

Clinical characteristics and treatments of multi-system inflammatory syndrome in children: a systematic review.

OBJECTIVE: Multisystem inflammatory syndrome in children (MIS-C) can occur in association with coronavirus disease 2019 (COVID-19). It is not easy to differentiate MIS-C from severe COVID-19 or Kawasaki disease based on symptoms. The aim of this study was to describe the clinical and laboratory characteristics of MIS-C. PATIENTS AND METHODS: We searched PubMed/Medline for case series and reports of MIS-C published until June 20, 2020. From a total of nine articles involving 45 cases, various clinical and laboratory data were extracted. Each target case was evaluated by using different diagnostic criteria. RESULTS: The average age at onset of MIS-C was 8.6 years. In 80% of cases, the age of patients ranged from 5 to 15 years. Fever (100%) and shock (82%) were the most common presenting symptoms. Sixty percent of cases met the diagnostic criteria for typical or atypical Kawasaki disease. Biomarkers indicative of inflammation, coagulopathy, or cardiac injury were characteristically elevated as follows: ferritin (mean: 1,061 ng/mL), CRP (217 mg/L), ESR (69 mm/hr), IL-6 (214.8 pg/mL), TNFα (63.4 pg/mL), D-dimer (3,220 ng/mL), PT (15.5 s), troponin I (1,006 ng/L), and BNP (12,150 pg/mL). Intravenous immunoglobulin was administered in all target cases, and inotropic agents were commonly used as well. No case of death was observed. CONCLUSIONS: This study demonstrated that MIS-C is a serious condition that presents with fever, rash, as well as cardiovascular and gastrointestinal symptoms. Although it is challenging to differentiate MIS-C from Kawasaki disease or severe COVID-19, initiation of appropriate treatments through early diagnosis is warranted.

Associations between SARS-CoV-2 variants and risk of COVID-19 hospitalization among confirmed cases in Washington State: a retrospective cohort study (preprint)

BackgroundThe COVID-19 pandemic is now dominated by variant lineages; the resulting impact on disease severity remains unclear. Using a retrospective cohort study, we assessed the risk of hospitalization following infection with nine variants of concern or interest (VOC/VOI). MethodsOur study includes individuals with positive SARS-CoV-2 RT-PCR in the Washington Disease Reporting System and with available viral genome data, from December 1, 2020 to July 30, 2021. The main analysis was restricted to cases with specimens collected through sentinel surveillance. Using a Cox proportional hazards model with mixed effects, we estimated hazard ratios (HR) for the risk of hospitalization following infection with a VOC/VOI, adjusting for age, sex, and vaccination status. FindingsOf the 27,814 cases, 23,170 (83.3%) were sequenced through sentinel surveillance, of which 726 (3.1%) were hospitalized due to COVID-19. Higher hospitalization risk was found for infections with Gamma (HR 3.17, 95% CI 2.15-4.67), Beta (HR: 2.97, 95% CI 1.65-5.35), Delta (HR: 2.30, 95% CI 1.69-3.15), and Alpha (HR 1.59, 95% CI 1.26-1.99) compared to infections with an ancestral lineage. Following VOC infection, unvaccinated patients show a similar higher hospitalization risk, while vaccinated patients show no significant difference in risk, both when compared to unvaccinated, ancestral lineage cases. InterpretationInfection with a VOC results in a higher hospitalization risk, with an active vaccination attenuating that risk. Our findings support promoting hospital preparedness, vaccination, and robust genomic surveillance.

Unsupervised multi-latent space reinforcement learning framework for video summarization in ultrasound imaging (preprint)

The COVID-19 pandemic has highlighted the need for a tool to speed up triage in ultrasound scans and provide clinicians with fast access to relevant information. The proposed video-summarization technique is a step in this direction that provides clinicians access to relevant key-frames from a given ultrasound scan (such as lung ultrasound) while reducing resource, storage and bandwidth requirements. We propose a new unsupervised reinforcement learning (RL) framework with novel rewards that facilitates unsupervised learning avoiding tedious and impractical manual labelling for summarizing ultrasound videos to enhance its utility as a triage tool in the emergency department (ED) and for use in telemedicine. Using an attention ensemble of encoders, the high dimensional image is projected into a low dimensional latent space in terms of: a) reduced distance with a normal or abnormal class (classifier encoder), b) following a topology of landmarks (segmentation encoder), and c) the distance or topology agnostic latent representation (convolutional autoencoders). The decoder is implemented using a bi-directional long-short term memory (Bi-LSTM) which utilizes the latent space representation from the encoder. Our new paradigm for video summarization is capable of delivering classification labels and segmentation of key landmarks for each of the summarized keyframes. Validation is performed on lung ultrasound (LUS) dataset, that typically represent potential use cases in telemedicine and ED triage acquired from different medical centers across geographies (India, Spain and Canada).

Homelessness in Transit Environments Volume II: Transit Agency Strategies and Responses

Transit settings represent sites of visible homelessness, especially since the advent of COVID-19, for many of the over 500,000 Americans unhoused each night. This report seeks to understand the scale of homelessness on transit and how transit agencies are responding to the problem. Part I describes the extent of homelessness on transit in several areas by using count data and synthesizing prior research. The authors find that transit serves as shelter for a high, though quite variable, share of unsheltered individuals, who are more likely than their unhoused peers elsewhere to be chronically unhoused and structurally disadvantaged. Part II provides detailed case studies of strategies taken by transit agencies around the country: hub of services, mobile outreach, discounted fares, and transportation to shelters. The authors summarize each strategy’s scope, implementation, impact, challenges, and lessons learned. Reviewing these strategies, they find value in collecting data more systematically, fostering external partnerships, keeping law enforcement distinct from routine homeless outreach, educating the public, and training transit staff—all in the context of a broader need for more housing and services.

Homelessness in Transit Environments Volume I: Findings from a Survey of Public Transit Operators

More than half a million individuals experience homelessness every night in the U.S. With the scale of the crisis often surpassing the capacities of existing safety nets—all the more so since the onset of the COVID-19 pandemic—many turn to transit vehicles, stops, and stations for shelter. Many also use transit to reach destinations such as workplaces, shelters, and community service centers. This report investigates the intersections of the pandemic, transit, and homelessness, presenting the results of a survey of 115 transit operators on issues of homelessness on their systems. The authors find that homelessness is broadly present across transit systems, though concentrated on larger operators and central hotspots, and has reportedly worsened on transit during the pandemic. The perceived challenges of homelessness are deepening, and data, dedicated funding, and staff are rare. However, a number of responses, including external partnerships and outreach and service provision, are growing, and agencies are adapting quickly to the pandemic. All told, centering the mobility and wellbeing of unhoused riders fits within transit’s social service role and is important to improving outcomes for them and for all riders.

Learning the Imaging Landmarks: Unsupervised Key point Detection in Lung Ultrasound Videos (preprint)

Lung ultrasound (LUS) is an increasingly popular diagnostic imaging modality for continuous and periodic monitoring of lung infection, given its advantages of non-invasiveness, non-ionizing nature, portability and easy disinfection. The major landmarks assessed by clinicians for triaging using LUS are pleura, A and B lines. There have been many efforts for the automatic detection of these landmarks. However, restricting to a few pre-defined landmarks may not reveal the actual imaging biomarkers particularly in case of new pathologies like COVID-19. Rather, the identification of key landmarks should be driven by data given the availability of a plethora of neural network algorithms. This work is a first of its kind attempt towards unsupervised detection of the key LUS landmarks in LUS videos of COVID-19 subjects during various stages of infection. We adapted the relatively newer approach of transporter neural networks to automatically mark and track pleura, A and B lines based on their periodic motion and relatively stable appearance in the videos. Initial results on unsupervised pleura detection show an accuracy of 91.8% employing 1081 LUS video frames.

Novel ELISA Protocol Links Pre-Existing SARS-CoV-2 Reactive Antibodies With Endemic Coronavirus Immunity and Age and Reveals Improved Serologic Identification of Acute COVID-19 via Multi-Parameter Detection.

The COVID-19 pandemic has drastically impacted work, economy, and way of life. Sensitive measurement of SARS-CoV-2 specific antibodies would provide new insight into pre-existing immunity, virus transmission dynamics, and the nuances of SARS-CoV-2 pathogenesis. To date, existing SARS-CoV-2 serology tests have limited utility due to insufficient reliable detection of antibody levels lower than what is typically present after several days of symptoms. To measure lower quantities of SARS-CoV-2 IgM, IgG, and IgA with higher resolution than existing assays, we developed a new ELISA protocol with a distinct plate washing procedure and timed plate development via use of a standard curve. Very low optical densities from samples added to buffer coated wells at as low as a 1:5 dilution are reported using this 'BU ELISA' method. Use of this method revealed circulating SARS-CoV-2 receptor binding domain (RBD) and nucleocapsid protein (N) reactive antibodies (IgG, IgM, and/or IgA) in 44 and 100 percent of pre-pandemic subjects, respectively, and the magnitude of these antibodies tracked with antibody levels of analogous viral proteins from endemic coronavirus (eCoV) strains. The disease status (HIV, SLE) of unexposed subjects was not linked with SARS-CoV-2 reactive antibody levels; however, quantities were significantly lower in subjects over 70 years of age compared with younger counterparts. Also, we measured SARS-CoV-2 RBD- and N- specific IgM, IgG, and IgA antibodies from 29 SARS-CoV-2 infected individuals at varying disease states, including 10 acute COVID-19 hospitalized subjects with negative serology results by the EUA approved Abbott IgG chemiluminescent microparticle immunoassay. Measurements of SARS-CoV-2 RBD- and N- specific IgM, IgG, IgA levels measured by the BU ELISA revealed higher signal from 9 of the 10 Abbott test negative COVID-19 subjects than all pre-pandemic samples for at least one antibody specificity/isotype, implicating improved serologic identification of SARS-CoV-2 infection via multi-parameter, high sensitive antibody detection. We propose that this improved ELISA protocol, which is straightforward to perform, low cost, and uses readily available commercial reagents, is a useful tool to elucidate new information about SARS-CoV-2 infection and immunity and has promising implications for improved detection of all analytes measurable by this platform.

Multidose evaluation of 6,710 drug repurposing library identifies potent SARS-CoV-2 infection inhibitors In Vitro and In Vivo. (preprint)

The SARS-CoV-2 pandemic has caused widespread illness, loss of life, and socioeconomic disruption that is unlikely to resolve until vaccines are widely adopted, and effective therapeutic treatments become established. Here, a well curated and annotated library of 6710 clinical and preclinical molecules, covering diverse chemical scaffolds and known host targets was evaluated for inhibition of SARS-CoV-2 infection in multiple infection models. Multi-concentration, high-content immunocytofluorescence-based screening identified 172 strongly active small molecules, including 52 with submicromolar potencies. The active molecules were extensively triaged by in vitro mechanistic assays, including human primary cell models of infection and the most promising, obatoclax, was tested for in vivo efficacy. Structural and mechanistic classification of compounds revealed known and novel chemotypes and potential host targets involved in each step of the virus replication cycle including BET proteins, microtubule function, mTOR, ER kinases, protein synthesis and ion channel function. In the mouse disease model obatoclax effectively reduced lung virus load by 10-fold. Overall, this work provides an important, publicly accessible, foundation for development of novel treatments for COVID-19, establishes human primary cell-based pharmacological models for evaluation of therapeutics and identifies new insights into SARS-CoV-2 infection mechanisms.

The Impact of COVID-19-Related Shutdown Measures on the Training Habits and Perceptions of Athletes in the United States: A Brief Research Report.

The purpose of the current study was to examine the impact of COVID-19 government-enforced shutdown measures on the training habits and perceptions of athletes. A web-based electronic survey was developed and distributed online to athletes. The survey contained questions regarding currently available resources, changes in weekly training habits, and perceptions of training such as intensity, motivation, and enjoyment. A total of 105 (males: n = 31; females: n = 74) athletes completed the survey (mean ± SD age = 19.86 ± 2.13 years). Ninety-nine (94.3%) athletes continued to receive guidance from their primary sport coach or strength training staff. There was a significant (p < 0.001) decrease (mean ± SD) in self-reported participation time for strength training (-1.65 ± 4.32 h. week-1), endurance (-1.47 ± 3.93 h. week-1), and mobility (-1.09 ± 2.24 h. week-1), with the largest reduction coming from participation time in sport-specific activities (-6.44 ± 6.28 h. week-1) pre- to post-shutdown. When asked to rate their current state of emotional well-being using a visual analog scale of 0-100, with 100 being exceptional, the mean score was 51.6 ± 19.6 AU. Athletes experienced notable reductions in training frequency and time spent completing various training related activities. In the future, practitioners should have preparations in place in the event of another lockdown period or future pandemic to avoid or minimize significant disruptions in training. Special considerations may be needed when athletes are allowed to return to sport in the event of significant levels of detraining that may have occurred.

Risk factors of COVID-19 mortality: a systematic review of current literature and lessons from recent retracted articles.

OBJECTIVE: Recently, two influential articles that reported the association of (hydroxy)chloroquine or angiotensin converting enzyme (ACE) inhibitors and coronavirus disease 2019 (COVID-19) mortality were retracted due to significant methodological issues. Therefore, we aimed to analyze the same clinical issues through an improved research method and to find out the differences from the retracted papers. We systematically reviewed pre-existing literature, and compared the results with those of the retracted papers to gain a novel insight. MATERIALS AND METHODS: We extracted common risk factors identified in two retracted papers, and conducted relevant publication search until June 26, 2020 in PubMed. Then, we analyzed the risk factors for COVID-19 mortality and compared them to those of the retracted papers. RESULTS: Our systematic review demonstrated that most demographic and clinical risk factors for COVID-19 mortality were similar to those of the retracted papers. However, while the retracted paper indicated that both (hydroxy)chloroquine monotherapy and combination therapy with macrolide were associated with higher risk of mortality, our study showed that only combination therapy of hydroxychloroquine and macrolide was associated with higher risk of mortality (odds ratio 2.33; 95% confidence interval 1.63-3.34). In addition, our study demonstrated that use of ACE inhibitors or angiotensin receptor blockers (ARBs) was associated with reduced risk of mortality (0.77; 0.65-0.91). CONCLUSIONS: When analyzing the same clinical issues with the two retracted papers through a systematic review of randomized controlled trials and relevant cohort studies, we found out that (hydroxy)chloroquine monotherapy was not associated with higher risk of mortality, and that the use of ACE inhibitors or ARBs was associated with reduced risk of mortality in COVID-19 patients.

Clinical characteristics and mortality of patients with hematologic malignancies and COVID-19: a systematic review.

OBJECTIVE: Hematologic cancer patients with Coronavirus Disease 2019 (COVID-19) tend to have a more serious disease course than observed in the general population. Herein, we comprehensively reviewed existing literature and analyzed clinical characteristics and mortality of patients with hematologic malignancies and COVID-19. MATERIALS AND METHODS: Through searching PubMed until June 03, 2020, we identified 16 relevant case studies (33 cases) from a total of 45 studies that have reported on patients with COVID-19 and hematologic malignancies. We investigated the clinical and laboratory characteristics including type of hematologic malignancies, initial symptoms, laboratory findings, and clinical outcomes. Then, we compared those characteristics and outcomes of patients with hematologic malignancies and COVID-19 to the general population infected with COVID-19. RESULTS: The median age was 66-year-old. Chronic lymphocytic leukemia was the most common type of hematologic malignancy (39.4%). Fever was the most common symptom (75.9%). Most patients had normal leukocyte counts (55.6%), lymphocytosis (45.4%), and normal platelet counts (68.8%). In comparison to patients with COVID-19 without underlying hematologic malignancies, dyspnea was more prevalent (45.0 vs. 24.9%, p=0.025). Leukocytosis (38.9 vs. 9.8%, p=0.001), lymphocytosis (45.4 vs. 8.2%, p=0.001), and thrombocytopenia (31.3 vs. 11.4%, p=0.036) were significantly more prevalent and lymphopenia (18.2 vs. 57.4%, p=0.012) less prevalent in patients with hematologic malignancies. There were no clinical and laboratory characteristics predicting mortality in patients with hematologic malignancies. Mortality was much higher in patients with hematologic malignancies compared to those without this condition (40.0 vs. 3.6%, p<0.001). CONCLUSIONS: Co-occurrence of hematologic malignancies and COVID-19 is rare. However, due to the high mortality rate from COVID-19 in this vulnerable population, further investigation on tailored treatment and management is required.

Genomic and phenotypic analysis of COVID-19-associated pulmonary aspergillosis isolates of Aspergillus fumigatus (preprint)

The ongoing global pandemic caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is responsible for the coronavirus disease 2019 (COVID-19) first described from Wuhan, China. A subset of COVID-19 patients has been reported to have acquired secondary infections by microbial pathogens, such as fungal opportunistic pathogens from the genus Aspergillus. To gain insight into COVID-19 associated pulmonary aspergillosis (CAPA), we analyzed the genomes and characterized the phenotypic profiles of four CAPA isolates of Aspergillus fumigatus obtained from patients treated in the area of North Rhine-Westphalia, Germany. By examining the mutational spectrum of single nucleotide polymorphisms, insertion-deletion polymorphisms, and copy number variants among 206 genes known to modulate A. fumigatus virulence, we found that CAPA isolate genomes do not exhibit major differences from the genome of the Af293 reference strain. By examining virulence in an invertebrate moth model, growth in the presence of osmotic, cell wall, and oxidative stressors, and the minimum inhibitory concentration of antifungal drugs, we found that CAPA isolates were generally, but not always, similar to A. fumigatus reference strains Af293 and CEA17. Notably, CAPA isolate D had more putative loss of function mutations in genes known to increase virulence when deleted (e.g., in the FLEA gene, which encodes a lectin recognized by macrophages). Moreover, CAPA isolate D was significantly more virulent than the other three CAPA isolates and the A. fumigatus reference strains tested. These findings expand our understanding of the genomic and phenotypic characteristics of isolates that cause CAPA.