Breaking Barriers, Not Privacy - Real-World Split Learning across Healthcare Systems

Learn how to implement Split Learning technology to enable collaborative AI training across independent healthcare systems without compromising patient data privacy in this 17-minute conference presentation from PEPR '25. Discover the technical workflow including partial model hosting and secure exchange of activations that allows three separate Health Information Exchanges (HIEs) to jointly train deep learning models while maintaining regulatory compliance. Explore real-world implementation challenges including data integration and quality issues, network security considerations, and regulatory compliance requirements that must be navigated when deploying privacy-preserving machine learning in healthcare environments. Examine preliminary results demonstrating robust model performance and seamless interoperability across participating sites, providing a practical blueprint for scaling privacy-preserving ML solutions in healthcare settings where regulatory constraints and data silos typically prevent collaborative AI development.