"Researchers Develop Easy-To-Deploy Federated Learning System That Safeguards Patient Data"
Researchers at the University of Oxford have developed a new, user-friendly method for hospitals to contribute to developing Artificial Intelligence (AI) models while protecting patient data. The technique builds on recent advancements in decentralized Machine Learning (ML) and uses low-cost pre-programmed microcomputers, making it easy to implement in hospitals and inexpensive to scale. As patient privacy is critical, hospitals are often limited in sharing data to support AI algorithm development. Once the data is shared, it can be difficult to ensure confidentiality. Federated learning was developed in 2017 as a method of training AI algorithms without moving data, and researchers have been collaborating with major technology companies to explore how it can be used in healthcare systems. However, federated learning has seen limited adoption in hospitals, partly because its implementation usually requires specialist expertise at each hospital involved in AI development. Oxford researchers developed and tested a new technique called full-stack federated learning, in which the software is pre-bundled with low-cost microcomputing hardware to create a plug-and-play system that hospitals can easily implement. This article continues to discuss the researchers' federated learning system that protects patient data.
Submitted by grigby1