A scalable federated learning solution for secondary care using low-cost microcomputing: privacy-preserving development and evaluation of a COVID-19 screening test in UK hospitals.

BACKGROUND: Multicentre training could reduce biases in medical artificial intelligence (AI); however, ethical, legal, and technical considerations can constrain the ability of hospitals to share data. Federated learning enables institutions to participate in algorithm development while retaining custody of their data but uptake in hospitals has been limited, possibly as deployment requires specialist software and technical expertise at each site. We previously developed an artificial intelligence-driven screening test for COVID-19 in emergency departments, known as CURIAL-Lab, which uses vital signs and blood tests that are routinely available within 1 h of a patient's arrival. Here we aimed to federate our COVID-19 screening test by developing an easy-to-use embedded system-which we introduce as full-stack federated learning-to train and evaluate machine learning models across four UK hospital groups without centralising patient data. METHODS: We supplied a Raspberry Pi 4 Model B preloaded with our federated learning software pipeline to four National Health Service (NHS) hospital groups in the UK: Oxford University Hospitals NHS Foundation Trust (OUH; through the locally linked research University, University of Oxford), University Hospitals Birmingham NHS Foundation Trust (UHB), Bedfordshire Hospitals NHS Foundation Trust (BH), and Portsmouth Hospitals University NHS Trust (PUH). OUH, PUH, and UHB participated in federated training, training a deep neural network and logistic regressor over 150 rounds to form and calibrate a global model to predict COVID-19 status, using clinical data from patients admitted before the pandemic (COVID-19-negative) and testing positive for COVID-19 during the first wave of the pandemic. We conducted a federated evaluation of the global model for admissions during the second wave of the pandemic at OUH, PUH, and externally at BH. For OUH and PUH, we additionally performed local fine-tuning of the global model using the sites' individual training data, forming a site-tuned model, and evaluated the resultant model for admissions during the second wave of the pandemic. This study included data collected between Dec 1, 2018, and March 1, 2021; the exact date ranges used varied by site. The primary outcome was overall model performance, measured as the area under the receiver operating characteristic curve (AUROC). Removable micro secure digital (microSD) storage was destroyed on study completion. FINDINGS: Clinical data from 130 941 patients (1772 COVID-19-positive), routinely collected across three hospital groups (OUH, PUH, and UHB), were included in federated training. The evaluation step included data from 32 986 patients (3549 COVID-19-positive) attending OUH, PUH, or BH during the second wave of the pandemic. Federated training of a global deep neural network classifier improved upon performance of models trained locally in terms of AUROC by a mean of 27·6% (SD 2·2): AUROC increased from 0·574 (95% CI 0·560-0·589) at OUH and 0·622 (0·608-0·637) at PUH using the locally trained models to 0·872 (0·862-0·882) at OUH and 0·876 (0·865-0·886) at PUH using the federated global model. Performance improvement was smaller for a logistic regression model, with a mean increase in AUROC of 13·9% (0·5%). During federated external evaluation at BH, AUROC for the global deep neural network model was 0·917 (0·893-0·942), with 89·7% sensitivity (83·6-93·6) and 76·6% specificity (73·9-79·1). Site-specific tuning of the global model did not significantly improve performance (change in AUROC <0·01). INTERPRETATION: We developed an embedded system for federated learning, using microcomputing to optimise for ease of deployment. We deployed full-stack federated learning across four UK hospital groups to develop a COVID-19 screening test without centralising patient data. Federation improved model performance, and the resultant global models were generalisable. Full-stack federated learning could enable hospitals to contribute to AI development at low cost and without specialist technical expertise at each site. FUNDING: The Wellcome Trust, University of Oxford Medical and Life Sciences Translational Fund.

Real-time assessment of neutrophil metabolism and oxidative burst using extracellular flux analysis.

Neutrophil responses are critical during inflammatory and infective events, and neutrophil dysregulation has been associated with poor patient outcomes. Immunometabolism is a rapidly growing field that has provided insights into cellular functions in health and disease. Neutrophils are highly glycolytic when activated, with inhibition of glycolysis associated with functional deficits. There is currently very limited data available assessing metabolism in neutrophils. Extracellular flux (XF) analysis assesses real time oxygen consumption and the rate of proton efflux in cells. This technology allows for the automated addition of inhibitors and stimulants to visualise the effect on metabolism. We describe optimised protocols for an XFe96 XF Analyser to (i) probe glycolysis in neutrophils under basal and stimulated conditions, (ii) probe phorbol 12-myristate 13-acetate induced oxidative burst, and (iii) highlight challenges of using XF technology to examine mitochondrial function in neutrophils. We provide an overview of how to analyze XF data and identify pitfalls of probing neutrophil metabolism with XF analysis. In summary we describe robust methods for assessing glycolysis and oxidative burst in human neutrophils and discuss the challenges around using this technique to assess mitochondrial respiration. XF technology is a powerful platform with a user-friendly interface and data analysis templates, however we suggest caution when assessing neutrophil mitochondrial respiration.

Assessment of Alveolar Macrophage Dysfunction Using an in vitro Model of Acute Respiratory Distress Syndrome.

Background: Impaired alveolar macrophage (AM) efferocytosis may contribute to acute respiratory distress syndrome (ARDS) pathogenesis; however, studies are limited by the difficulty in obtaining primary AMs from patients with ARDS. Our objective was to determine whether an in vitro model of ARDS can recapitulate the same AM functional defect observed in vivo and be used to further investigate pathophysiological mechanisms. Methods: AMs were isolated from the lung tissue of patients undergoing lobectomy and then treated with pooled bronchoalveolar lavage (BAL) fluid previously collected from patients with ARDS. AM phenotype and effector functions (efferocytosis and phagocytosis) were assessed by flow cytometry. Rac1 gene expression was assessed using quantitative real-time PCR. Results: ARDS BAL treatment of AMs decreased efferocytosis (p = 0.0006) and Rac1 gene expression (p = 0.016); however, bacterial phagocytosis was preserved. Expression of AM efferocytosis receptors MerTK (p = 0.015) and CD206 (p = 0.006) increased, whereas expression of the antiefferocytosis receptor SIRPα decreased following ARDS BAL treatment (p = 0.036). Rho-associated kinase (ROCK) inhibition partially restored AM efferocytosis in an in vitro model of ARDS (p = 0.009). Conclusions: Treatment of lung resection tissue AMs with ARDS BAL fluid induces impairment in efferocytosis similar to that observed in patients with ARDS. However, AM phagocytosis is preserved following ARDS BAL treatment. This specific impairment in AM efferocytosis can be partially restored by inhibition of ROCK. This in vitro model of ARDS is a useful tool to investigate the mechanisms by which the inflammatory alveolar microenvironment of ARDS induces AM dysfunction.

Using a non-invasive assessment of lung injury in a murine model of acute lung injury.

Arterial oxygen saturation has not been assessed sequentially in conscious mice as a direct consequence of an in vivo murine model of acute lung injury. Here, we report daily changes in arterial oxygen saturation and other cardiopulmonary parameters by using infrared pulse oximetry following intratracheal lipopolysaccharide (IT-LPS) for up to 9 days, and following IT-phosphate buffered saline up to 72 h as a control. We show that arterial oxygen saturation decreases, with maximal decline at 96 h post IT-LPS. Blood oxygen levels negatively correlate with 7 of 10 quantitative markers of murine lung injury, including neutrophilia and interleukin-6 expression. This identifies infrared pulse oximetry as a method to non-invasively monitor arterial oxygen saturation following direct LPS instillations.

Improving care for patients with idiopathic pulmonary fibrosis (IPF) in the UK: a round table discussion.

This article is the result of a round table discussion held at BMA House. Its purpose is to outline the challenges facing both patients and doctors in managing idiopathic pulmonary fibrosis. A number of recommendations for service development and research have been identified.