In a recent study published in The Lancet Digital Health, a group of researchers developed and evaluated a scalable, privacy-preserving federated learning solution using low-cost microcomputing for coronavirus disease 2019 (COVID-19) screening in United Kingdom (UK) hospitals.
The use of patient knowledge in medical synthetic intelligence (AI) studies faces ethical, legal, and technical challenges, adding dangers of misuse and privacy violations. Federated learning offers a privacy-friendly approach by enabling the progression of AI models without sharing knowledge outside of organizations. It allows the formation of local knowledge, unlike the classic centralized training.
This method, especially client-server federated learning, involves sharing model weights, not patient data, for global model development. Real-world hospital implementations are rare, often requiring technical expertise and data separation from clinical systems.
Further studies are needed to refine and validate the federated learning technique in healthcare settings and to address demanding implementation scenarios for wider adoption in real-world clinical settings.
The study provided focused on a detailed procedure for scaling up and testing a federated learning solution for COVID-19 monitoring in UK hospitals. The researchers opted for four NHS hospital groups: University Hospitals Oxford (OUH), University Hospitals Birmingham (UHB). ), Bedfordshire Hospitals (BH) and Portsmouth University of Hospitals (PUH) and used four Raspberry Pi Model B devices for their full use. Stack federated learning. This setup allowed each hospital to train, calibrate, and compare AI models with locally anonymized patient data, ensuring privacy.
Inclusion and exclusion criteria were provided to NHS trusts for the extraction of data from electronic fitness records. Data de-identification was rigorously carried out through clinical groups or NHS IT specialists. The study used a pre-pandemic cohort and a COVID-19 positive cohort for training, with insights adding important signs, demographics, and blood test results. Data extracts were uploaded to consumer devices for federated training, calibration, and evaluation.
Federated education used logistic regression and deep neural network classifiers. Features were preprocessed in a common format and, in the absence of knowledge, local median values were imputed. The FedAvg ruleset facilitated education among hospital groups, and consumers passed style parameters to headquarters. Server for aggregation. Calibration of local styles to an explained sensitivity threshold, with evaluation effects added via the server.
The federated evaluation concerned the use of prospective cohorts from various hospitals. The methods of calibration and imputation were diverse depending on whether the sites referred to education and evaluation or only to evaluation. Site-specific styling tested the adaptability of the overall style, and a centralized server-side assessment verified the consistency of the federated assessment. The study also tested the effect of individual characteristics on style predictions.
The statistical research aimed to compare the functionality of style in other educational settings and methods, measures such as AUROC, sensitivity, and specificity.
In the study, the comparison revealed a notable increase in the AUROC of the logistic regression model. For example, the OUH saw an increase in the AUROC from 0. 685 to 0. 829, and the PUH saw an increase from 0. 731 to 0. 865. Similarly, deep neural network models showed even more significant improvements, with AUROC values expanding from 0. 574 to 0. 872 at OUH and from 0. 622 to 0. 876 at PUH.
Three NHS trusts (OUH, UBB, and PUH) participated in this federated training, offering insights from a giant cohort of patients. The federated evaluation included knowledge about patients admitted in the second wave of the pandemic, with different COVID-19 prevalence rates and mean ages. in attractive places.
When the final global styles were externally evaluated, the logistic regression and deep neural network styles demonstrated superior classification performance. Federated calibration achieved impressive sensitivities, with logistic regression style at 83. 4% and deep neural network style at 89. 7%.
The functionality of those models remained robust across all assessment sites. The deep neural network model, in particular, showed a more marked improvement through federation than the logistic regression model, achieving a plateau of functionality after about 75 to 100 rounds.
Site-specific tuning of the global models resulted in a slight improvement in the deep neural network model at PUH. Still, no significant improvement was observed for the logistic regression model. This suggested a high level of generalizability of the global models and minimal shifts in predictor distributions between sites.
The global logistic regression style analysis highlighted several key predictors, such as granulocyte count and albumin concentrations, which is consistent with previous studies emphasizing their role in the inflammatory response. Analysis of deep neural network style Shapley’s additive explanations revealed that eosinophil counts were a highly influential predictor.
Written by
Use one of the following to cite this article in your essay, article, or report:
APA
Kumar Malesu, Vijay. (2024, January 31). The researchers present a federated, scalable, privacy-focused, affordable microcomputing learning formula for COVID-19 testing in hospitals. Retrieved February 1, 2024, https://www. news-medical. net/news/20240131/Researchers-unveil-scalable -federated-learning-formula-privacy-focused–affordable-microcomputing-for-COVID-19-screening-in-hospitals. aspx.
Deputy
Kumar Malesu, Vijay. ” Researchers reveal a scalable, privacy-focused, federated learning formula, affordable microcomputing for COVID-19 testing in hospitals. “News-Medical. 01 February 2024.
Chicago
Kumar Malesu, Vijay. ” Researchers reveal a scalable, privacy-focused, federated learning formula, affordable microcomputing for COVID-19 detection in hospitals. “https://www. news-medical. net/news/20240131/Researchers-unveil-scalable-privacy-focused-federated-learning-formula–affordable-microcomputing-for-COVID-19-screening-in-hospitals. aspx. (accessed February 1, 2024).
Harvard
Kumar Malesu, Vijay. 2024. Researchers reveal a federated, scalable, privacy-focused learning formula, affordable microcomputing for COVID-19 testing in hospitals. News-Medical, accessed February 1, 2024, https://www. news-medical. net/ news/20240131/Researchers-unveil-scalable-privacy-focused-federated-learning-formula–affordable-microcomputing-for-COVID -19-screening-in-hospitals. aspx.
News-Medical. net – An AZoNetwork site
Owned and operated through AZoNetwork, © 2000-2024