Deep Learning Tool for Automatic B-Line Detection and Localization in Lung Point-of-Care Ultrasound

Study Objectives: Point-of-care ultrasound (POCUS) offers real-time data to guide clinical decision-making and patient care. Despite having advantages over alternative imaging studies such as computed tomography or magnetic resonance imaging, performing POCUS requires technical expertise for image acquisition and interpretation, thereby limiting its use for many clinicians. Deep learning technologies can provide automated interpretation of POCUS images thus making POCUS accessible to even novice users. B-lines are sonographic artifacts seen on lung POCUS which are diagnostic for pulmonary diseases such as pneumonia, COVID-19, or decompensated heart failure. In this work we aim to develop a deep learning tool to automatically detect and localize B-lines on lung POCUS clips. Methods: Using a 12-point scanning protocol, we prospectively collected lung POCUS clips from 25 patients presenting to the emergency department with shortness of breath and/or flu-like symptoms. Sub-sampled frames from 500 POCUS clips were annotated for B-lines by 3 physicians with expertise in POCUS acquisition and interpretation. A 2D U-Net deep neural network was trained on frames annotated from 15 patients, with frames from the remaining 10 patients being set aside for validation studies. Transformations from polar to rectangular coordinates were performed as part of pre-processing the data. Frame-level predictions were aggregated to predict the presence or abscence of B-lines over an entire clip. Experiments are currently underway for determining the impact of weakly supervised vs. fully supervised training. Results: Initial results show an AUC score (95% CI) of 0.82 (0.74-0.89) for clip-level B-line detection based on a 5- fold cross-validation for the 15 patient subset. Additionally, by first segmenting B-lines, our approach for localization is substantially more specific than common alternatives, such as class-activation mapping. Conclusion: Here we generated a deep learning model that can detect the presence of B-lines on POCUS clips with significant accuracy. This model was developed from a limited training subset, thus we predict that with more integrated data, our model can be further refined to identify and ideally quantify B-lines on POCUS clips collected from an array of machines and from users with variable image acquisition experience. Ideally, this tool may enable clinicians with minimal prior training in POCUS to integrate this powerful imaging tool into patient care.

Dysregulated Innate and Adaptive Immune Responses Discriminate Disease Severity in COVID-19.

The clinical spectrum of COVID-19 varies and the differences in host response characterizing this variation have not been fully elucidated. COVID-19 disease severity correlates with an excessive proinflammatory immune response and profound lymphopenia. Inflammatory responses according to disease severity were explored by plasma cytokine measurements and proteomics analysis in 147 COVID-19 patients. Furthermore, peripheral blood mononuclear cell cytokine production assays and whole blood flow cytometry were performed. Results confirm a hyperinflammatory innate immune state, while highlighting hepatocyte growth factor and stem cell factor as potential biomarkers for disease severity. Clustering analysis revealed no specific inflammatory endotypes in COVID-19 patients. Functional assays revealed abrogated adaptive cytokine production (interferon-γ, interleukin-17, and interleukin-22) and prominent T-cell exhaustion in critically ill patients, whereas innate immune responses were intact or hyperresponsive. Collectively, this extensive analysis provides a comprehensive insight into the pathobiology of severe to critical COVID-19 and highlights potential biomarkers of disease severity.

A systems approach to inflammation identifies therapeutic targets in SARS-CoV-2 infection (preprint)

Background Infection with SARS-CoV-2 manifests itself as a mild respiratory tract infection in the majority of individuals, which progresses to a severe pneumonia and acute respiratory distress syndrome (ARDS) in 10-15% of patients. Inflammation plays a crucial role in the pathogenesis of ARDS, with immune dysregulation in severe COVID-19 leading to a hyperinflammatory response. A comprehensive understanding of the inflammatory process in COVID-19 is lacking. Methods In this prospective, multicenter observational study, patients with PCR-proven or clinically presumed COVID-19 admitted to the intensive care unit (ICU) or clinical wards were included. Demographic and clinical data were obtained and plasma was serially collected. Concentrations of IL-6, TNF-, complement components C3a, C3c and the terminal complement complex (TCC) were determined in plasma by ELISA. Additionally, 269 circulating biomarkers were assessed using targeted proteomics. Results were compared between ICU and non ICU patients. Findings A total of 119 (38 ICU and 91 non ICU) patients were included. IL-6 plasma concentrations were elevated in COVID-19 (ICU vs. non ICU, median 174.5 pg/ml [IQR 94.5-376.3 vs. 40.0 pg/ml [16.5-81.0]), whereas TNF- concentrations were relatively low and not different between ICU and non ICU patients (median 24.0 pg/ml [IQR 16.5-33.5] and 21.5 pg/ml [IQR 16.0-33.5], respectively). C3a and terminal complement complex (TCC) concentrations were significantly higher in ICU vs. non ICU patients (median 556.0 ng/ml [IQR 333.3-712.5]) vs. 266.5 ng/ml [IQR 191.5-384.0 for C3a and 4506 mAU/ml [IQR 3661-6595 vs. 3582 mAU/ml [IQR 2947-4300] for TCC) on the first day of blood sampling. Targeted proteomics demonstrated that IL-6 (logFC 2.2), several chemokines and hepatocyte growth factor (logFC 1.4) were significantly upregulated in ICU vs. non ICU patients. In contrast, stem cell factor was significantly downregulated (logFC -1.3) in ICU vs. non ICU patients, as were DPP4 (logFC -0.4) and protein C inhibitor (log FC -1.0), the latter two factors also being involved in the regulation of the kinin-kallikrein pathway. Unsupervised clustering pointed towards a homogeneous pathogenetic mechanism in the majority of patients infected with SARS-CoV-2, with patient clustering mainly based on disease severity. Interpretation We identified important pathways involved in dysregulation of inflammation in patients with severe COVID-19, including the IL-6, complement system and kinin-kallikrein pathways. Our findings may aid the development of new approaches to host-directed therapy.