A Federated Approach for Learning from Electronic Health Records
Sadaf Md Halim, Latifur Khan, Kevin W. Hamlen, Bhavani M. Thuraisingham, Md Delwar Hossain · 2022
The free sharing and exchange of medical information is difficult because medical data is by its nature sensitive. However, medical data is often found across various sources. Each of these sources contain very useful information, and we investigate the possibility of aggregating this information and utilizing this data without its direct transfer. Instead, we explore the use of Federated Learning to transfer the knowledge found in sensitive patient information through gradient updates and model parameters in order to better inform a variety of learning tasks. We also explore specific types of models such as graph networks to model the information in Electronic Health Records (EHRs) collected at hospitals, and we investigate how we might use federated learning to aggregate information such as this across hospitals. Lastly, we show how we can mitigate some of the security risks that are inherent in such federated systems.