Federated health ML systems survey

Data privacy concerns and inability to centralize health data are roadblocks to successful research collaborations. Federated research networks can help with cross-site research in such cases. Existing systems vary in their offerings, providing different levels of privacy, machine learning algorithms, etc. Some systems are designed for particular domains, such as neuroimaging or genomic data. Ideally, a healthcare research network should be able to handle data in varying formats, support a variety of ML algorithms, offer intuitive interfaces, and provide advanced privacy guarantess through mechanisms like differential privacy.

In this project we will explore how LEAP compares to existing solutions. We will survey several dimensions, such as the mechanism through which privacy is offered, distribution architecture, query language, data source assumptions, popularity in healthcare or ML settings, interaction methods, maintenance status, deployability, and target user demographic. We will use a multi-dimensional approach to visualize the comparison between different systems through BCG matrix plots.