Machine Learning for Intelligent Healthcare Across Decentralized Patient Databases Using Federated Learning

Federated learning is a novel machine learning technique that allows model training on decentralized data. Multiple users train their own ML models privately on their own private data using this strategy. Only the model parameters and not raw data are shared to nodes, ensuring privacy and addressing data sovereignty issues. First, this chapter provides a quick overview of machine learning and deep learning, while focusing on the various issues one might encounter in implementing the above in healthcare scenarios. It then explains the concept of federated learning, how it works, and how it ensures privacy and security with various methods like secure multiparty computation or differential privacy. Finally, it focuses mainly on the examples of the plethora of applications of federated learning in healthcare where privacy, data security, and confidentiality are important, and detail the benefits and challenges of using federated learning for intelligent healthcare applications. These applications include, but are not limited to, neurological disorder diagnosis with EEG signal classifications, lung cancer predication, COVID-19 diagnosis, and smart electronic health records. We may use federated learning to create ML models collectively and acquire insights from aggregated global patterns while keeping data decentralized. It will also shed light on the several ethical issues involved in such models for healthcare applications such as bias and fairness.