Federated learning in healthcare applications

Abstract

Federated learning (FL), also referred to as collaborative learning, uses a number of dispersed edge devices or servers to run the training algorithms, without exchanging local data samples. It differs from previous approaches in that it does not make the assumption that local data samples are evenly distributed, as is the case with more conventional decentralized systems. It also differs from traditional centralized machine learning techniques, which demand that all local datasets be uploaded to a single server. FL helps in handling crucial data-related challenges like managing and handling heterogeneous data, enabling privacy, access rights, security, etc. The major challenges and design considerations in FL related to the healthcare field are highlighted in this chapter. Because of the privacy and secrecy concerns of the medical data of the patients, sharing or exchange of diagnostic data across different entities is prohibitive. Moreover, there are multiple formats for collecting the medical data of the patients. This results in insufficient and imbalanced data, making model building and training a challenging task. Further, the collected medical diagnostic data are generally heterogeneous in terms of their statistical properties. This reduces the model's capability to generalize well in the medical domain. FL provides fusion-based secure, robust, cost-effective, and privacy-preserving solutions to all these challenges where knowledge obtained from different decentralized sources of data is fused to build a strong classification model. This chapter contains a detailed discussion of the above issues along with a possible scope for future research. It highlights the preliminaries, training algorithms, and the impact of FL on its stakeholders. It also describes the core applications of FL in the healthcare domain, which include electronic health record mining, remote health monitoring, medical imaging, disease prediction, etc. The challenges and considerations in FL, which include domain generalization, data and model heterogeneity, privacy and security, system architecture and resource sharing, lack of proper datasets and training methods in those cases, etc., are also discussed along with possible solutions for each of them. © 2024 Elsevier Inc. All rights are reserved including those for text and data mining AI training and similar technologies.