Securing patient data in large-scale AI healthcare development: AI Centre and deepc launch FLIP

Developed, governed and maintained within the NHS, FLIP provides multi-institutional access to the large patient data sets needed for clinically-robust AI model training while keeping it protected, in one place, under secure local institutional control.

This unique capability means AI models can learn from a wider range of data - improving patient care through increased diagnostic accuracy, supporting earlier disease detection and helping bring new innovations into care faster - while maintaining high standards of privacy and data protection.

Organisations using FLIP retain full control over their data, approvals and ethics processes and can deploy FLIP locally to develop AI models using only their own data, or can choose to participate in federated collaboration across sites. This approach reduces duplication of effort, mitigates AI research vendor lock-in risk and reinforces transparency around data ownership and reuse.

FLIP's development reflects a long-term strategic collaboration between the AI Centre for Value-Based Healthcare, Guy's and St Thomas' NHS Foundation Trust and deepc to build open infrastructure to accelerate AI innovation as well as its clinical use.

NVIDIA's NVFlare team and AWS cloud infrastructure support have also contributed to the project, enabling scalable orchestration across distributed environments.

Initial programmes running on FLIP are demonstrating federated AI capabilities across diverse clinical applications including radiology, multimodal inflammatory diseases and the development of digital biomarkers.

For example, in partnership with St John's Institute of Dermatology, researchers are developing digital biomarkers for inflammatory skin disease using multimodal data. Another programme aims to build a biomarker for muscle mass and sarcopenia to inform cancer treatment decisions and improve outcomes prediction in critically ill patients.

These projects combine imaging with structured and unstructured clinical data, illustrating how federated infrastructure can enable more comprehensive, real-world model development across institutions.

Released under an Apache 2.0 open-source license and maintained by the AI Centre for Value-Based Healthcare, FLIP is now available for research on multi-modal data and meets real-world NHS requirements for information governance and cybersecurity from the outset. Healthcare providers, researchers and innovators worldwide are invited to adopt, contribute to and build upon the platform.