DataTools4Heart Final Consortium Meeting in Crete: Four Years of Innovation, From Vision to a Working Ecosystem for Trustworthy Cardiovascular Data Sharing

Four years after the launch of DataTools4Heart, the consortium met over two days, on 17 and 18 September 2026, in Crete for its final consortium meeting, bringing together partners from across Europe to review the project’s achievements and look ahead to the future of data-driven cardiovascular research.

What began as an ambitious vision to enable federated, privacy-preserving AI for cardiovascular research has developed into a working ecosystem combining clinical AI, federated learning, multilingual NLP, synthetic data and privacy-preserving technologies.

DataTools4Heart has transformed real clinical challenges into research use cases and validated modelling workflows. More than 3,000 patients were analysed for personalised Heart Failure and Chronic Kidney Disease risk prediction, while a second use case developed prognostic models for acute heart failure using data from eight partner hospitals.

Across the platform, 13 hospital data sources now feed a shared Common Data Model, supported by automated data-ingestion, transformation and quality-control pipelines.

The project has also created a growing ecosystem of multilingual clinical NLP resources, including the CardioCCC dataset, which now includes 508 cardiology documents across seven languages, with more than 400,000 annotations covering diseases, medications, procedures and symptoms. The resources are already being used beyond the consortium, with more than 9,800 Zenodo views, 2,200 downloads and 1 TB of data transferred.

Through CardioSynth, multiple synthetic-data generation approaches have been systematically evaluated for fidelity, utility and privacy. Testing across 43 datasets found no verbatim training rows, while downstream evaluations showed that selected models trained on synthetic data can achieve performance close to models trained on real data.

At Amsterdam UMC, a real-world use case involving more than 16,000 extracted patients demonstrated approximately 85–90% of the utility of the full dataset with a hybrid differentially private synthesis approach.

At the same time, blockchain and Secure Multi-Party Computation (SMPC) provide the infrastructure needed to support federated collaboration while keeping raw clinical data within participating hospitals.

Technology alone cannot enable responsible health-data sharing. DataTools4Heart has therefore also addressed the legal, ethical and governance dimensions of federated research.

The project has contributed to an approach that brings together scientific research requirements, anonymisation, consent, secure access and accountability, while aligning platform access with emerging European Health Data Space (EHDS) principles.

From validated clinical models and multilingual AI to synthetic data, federated analytics and privacy-preserving infrastructure, DataTools4Heart has built the foundations for more collaborative, responsible and scalable cardiovascular research across Europe.

The end of the DataTools4Heart project marks the beginning of the next chapter for the tools, data, methodologies and knowledge created by the consortium. The project may be coming to an end, but the research ecosystem it has created is designed to keep moving forward.

A sincere thank you to all partners, researchers, clinicians, technical experts, legal and ethics specialists and the coordination team who contributed to this journey.