Federated Learning toward Data Preprocessing: COVID-19 Context

Abstract

During this last decade, in the digital era, online and real-time data management becomes essential and primordial in several scenarios. In the health domain, and especially for remote healthcare monitoring systems, the management of data in real time becomes a requirement. Indeed, care providers need to access the vital information and sometimes the processed data in order to take timely the adequate decision. However, many issues arise in the data processing regarding real-time aspect, computational complexity and patient mobility. Therefore, Federated Learning (FL) engages promising technologies such as the fog and the cloud computing as well as machine learning (ML) and deep learning (DL) to address the aforementioned issues. Fog computing enables data preprocessing in proximity of medical users, exploiting patients mobile phones or personal digital assistant (PDA) or small-scale distributed servers. In this paper, we pinpoint the important role of the FL-based system within Internet of Medical Things (IoMT) to combat COVID-19 pandemic. Specifically, we first introduce the architecture highlighting the fog layer within the smart healthcare system. We then discuss the preprocesing tasks that could be implemented at the fog layer with a particular focus on ML and DL tasks. After that, an investigation related to the FL against several COVID-19 contexts is provided. Finally, this paper explores open issues and future directions regarding the FL potentialities in pandemic situation.