A Robust Big Data Analytic Framework for IoMT Leveraged by Federated Learning

Abstract

With the fast development of the Internet of Medical Things, continuous health monitoring using interconnected smart-medical devices became possible, which produces large amounts of heterogeneous data. Nevertheless, the centralization of the processing of such data also presents significant issues concerning the scalability, latency, and privacy of data. This paper will offer a new model to overcome these shortcomings by proposing a new model known as FL-BDA-IoMT (Federated Learning-driven Big Data Analytics IoMT). The suggested algorithm combines distributed Federated Learning with scalable big data analytics that would allow privacy-sensitive, decentralized model training on various healthcare nodes. With the FL-BDA-IoMT, IoMT devices process and train models on sensitive medical data locally, and only model parameters are exchanged with a central aggregator, this way removing the sharing of raw data. The hierarchical structure with edge (fog) computing and cloud layers is used to decrease the overhead and latency in the communication. Adaptive aggregation, resource-conscious client selection are also proposed as part of the suggested approach to address the issue of data heterogeneity and client device limitations. The findings of the experiment demonstrate that the suggested solution improves predictive performance, minimizes the use of network bandwidth and achieves privacy requirements in contrast to conventional centralized machine learning methods. Moreover, with the incorporation of big data analytics, it is now possible to effectively manage the high-velocity and high-volume medical streams, which are critical to real-time decision-making within the healthcare system. The FL-BDA-IoMT algorithm offers an efficient, scalable and privacy-conscious system to next-generation intelligent healthcare analytics, which covers key issues in the IoMT setting.