Privacy-first health research with federated learning

Abstract

Privacy protection is paramount in conducting health research. However, studies often rely on data stored in a centralized repository, where analysis is done with full access to the sensitive underlying content. Recent advances in federated learning enable building complex machine-learned models that are trained in a distributed fashion. These techniques facilitate the calculation of research study endpoints such that private data never leaves a given device or healthcare system. We show—on a diverse set of single and multi-site health studies—that federated models can achieve similar accuracy, precision, and generalizability, and lead to the same interpretation as standard centralized statistical models while achieving considerably stronger privacy protections and without significantly raising computational costs. This work is the first to apply modern and general federated learning methods that explicitly incorporate differential privacy to clinical and epidemiological research—across a spectrum of units of federation, model architectures, complexity of learning tasks and diseases. As a result, it enables health research participants to remain in control of their data and still contribute to advancing science—aspects that used to be at odds with each other.

Operationalizing and Implementing Pretrained, Large Artificial Intelligence Linguistic Models in the US Health Care System: Outlook of Generative Pretrained Transformer 3 (GPT-3) as a Service Model

“…In future iterations, once the value of using GPT-3 in the health care setting is assured, the responsibility of accessibility could be delegated to health care and government agencies. Such agencies may distribute the “GPT-3-as-a-Service” platform through secure cloud platforms and establish a federated learning mechanism to run decentralized training services while collaboratively contributing to the GPT-3 model [ 28 ]. This would also reduce the burden on individual health systems when it comes to building, training, and deploying their own GPT-3 platforms and reduce costs.…”

Applying Machine Learning in Distributed Data Networks for Pharmacoepidemiologic and Pharmacovigilance Studies: Opportunities, Challenges, and Considerations

“…Such countermeasures may be neither desirable nor needed in the presence of proper governance and trusted collaborative relationships between data partners in a DDN. Despite these challenges, recent studies using federated learning with different ML architectures have shown that it is possible to achieve levels of performance comparable to models trained using a centralized approach [ 112 ] and better than locally trained models. For example, in a multi-site study [ 113 ] using structured EHR data (e.g., laboratory data and vital signs) and chest X-ray images to predict the future oxygen requirements of symptomatic patients with COVID-19, the global federated deep learning model, trained using data from 20 clinical sites around the world, outperformed all local models that were trained at a single site using that site’s data (Table 2 ).…”

Privacy-preserving federated neural network learning for disease-associated cell classification

“…11 This approach has enabled collaborative learning for several medical applications, and it has been shown that FL performs comparably with centralized training on medical datasets. 12 , 13 , 14 Recently, the concept of swarm learning (SL) has been proposed; it enables decentralized machine learning for precision medicine. The seminal work of SL 15 is based on edge computing and permissioned blockchains and removes the need for a central server in the FL approach.…”