THIS CONTENT IS BROUGHT TO YOU BY the Norwegian centre for E-health research - read more

Five Nordic and Baltic countries take a major step toward the future of health research

Researchers will work together without ever transferring patient data outside the hospital.

The joint Nordic-Baltic research project focuses on securely sharing and using health data.
Hasse Berntsen
COMMUNICATIONS ADVISER
Presented by: Norwegian centre for E-health research
Published

A new Nordic-Baltic research project is building a pan-European platform for health data analysis. The project is led by the Norwegian Centre for E-health Research.

Researchers are using a method called federated learning (FL). This allows them to train artificial intelligence (AI) models locally, without moving raw data between hospitals, institutions, or across national borders. 

The project is supported by multiple European institutions. It is among the first to apply FL to real-world health data. 

This safeguards patient privacy while also allowing researchers to share valuable insights. The result is faster diagnoses, more precise treatments, and improved prevention.

How does FL work?

“When we avoid moving raw data, the risk of data breaches is reduced,” says project manager Taridzo Chomutare.

Traditionally, researchers gather all data in a single location. This process can be both expensive and cumbersome. FL keeps patient data where it’s originally stored. Only updates to the AI models are sent across the network.

This allows hospitals, institutions, and countries to collaborate without sharing sensitive patient information. 

Researchers from Norway, Sweden, Denmark, Finland, and Estonia can thus tap into a much larger virtual dataset – without having to move raw data across borders.

“By avoiding the transfer of raw data, we minimise the risk of privacy breaches while still allowing analyses across large geographic regions,” says project manager Taridzo Chomutare from the Norwegian Centre for E-health Research.

Advantages of FL

  1. Better privacy: No raw data leaves the hospitals.
  2. Less need for harmonisation: Different systems and data formats pose fewer obstacles because the models train locally.
  3. Flexibility: Variations in data quality or network connectivity have minimal impact on the overall learning process.

Legal hurdles

“Complying with both national and international laws can be challenging. Privacy is always our main focus,” says Chomutare. 

The technology works, but the regulations are complex. There is no uniform European framework for the so-called secondary use of health data. Each country has its own requirements. 

The fact that the Nordic-Baltic network can operate at all suggests that it is possible to navigate bureaucratic challenges in health technology.

Ready for the future

The hope is to extend FL to other countries beyond the Nordic and Baltic region. 

If successful, vast amounts of data can be analysed safely and efficiently.

This could accelerate the pace of research.

Some potential benefits:

  • Earlier diagnoses: AI models detect diseases sooner.
  • Personalised medicine: Large, varied datasets enable more precise treatments.
  • International collaboration: FL tools and methods can be adapted to new areas.

A borderless collaboration project

The project demonstrates that experts from different fields – technology, medicine, and law – can collaborate to balance security and efficiency. 

While they still face technical and legal barriers, initial results are encouraging.

Experts from six institutions across five countries are collaborating to develop a federated health data network. The aim is to foster Nordic-Baltic cooperation in the secondary use of health data.

FL has the potential to transform everything from prevention to highly specialised treatment while ensuring privacy protection. 

If the Nordic-Baltic cooperation continues to grow, a new research paradigm may emerge.

This means improved patient care and great opportunities for researchers.

What Is federated learning?

Federated learning (FL) is a way to train AI models without sending raw data elsewhere. Models are trained locally, and only the updates are shared.

Key benefits:

  • Strong privacy protection: No raw data is moved.
  • Scalability: Suitable for large datasets.
  • Efficiency: Reduced need for data transfers.

Examples of use:

  • Healthcare: Developing diagnostic tools.
  • Mobile technology: Speech recognition, predictive text.
  • Finance: Fraud detection.

Reference:

Chomutare et al. 'Implementing a Nordic-Baltic Federated Health Data Network: a case report', Computer Science, 2024. DOI: 10.48550/arXiv.2409.17865 (Abstract)

———

Read the Norwegian version of this article on forskning.no

More content from the Norwegian centre for E-health research:

e-health-research artificial intelligence international relations partner health service health
Powered by Labrador CMS