AVATAR-Transfer builds on the research findings of the preceding AVATAR project. The focus is now on practical testing. The central question is how methods for anonymizing and protecting personal data can be tailored to specific medical datasets and usage scenarios. With strong roots in Thuringia, the AVATAR-Transfer consortium brings together technological development, medical practice and transfer into commercial applications.

Combining data protection and usability

Not every anonymization method is suitable for every dataset and intended use. Information is removed or altered to prevent conclusions from being drawn about individual people. At the same time, the data required for the intended analysis must be retained. Selecting and configuring appropriate privacy-enhancing technologies can therefore quickly become complex. This is where Fraunhofer IDMT’s “Model Driven Privacy” approach comes in. It demonstrates how the data, anonymization components, and the planned use case can be described in a model. This makes the selection and implementation of protective measures automatically executable, documented, and verifiable.

This is relevant, for example, when health data is used to train AI models or develop medical devices. Data protection also plays an important role in monitoring such devices after they have been placed on the market. Fraunhofer IDMT supports companies and medical institutions in assessing whether their use cases comply with the requirements of the General Data Protection Regulation. The researchers assess whether the datasets provide sufficient anonymity for the intended purpose and help select suitable anonymization tools. These tools are then configured and applied to the respective use case.

“When it comes to health data, many requirements need to be considered together. These include protecting patients as well as addressing technical, legal and ethical issues. There are also requirements for medical devices and services. Our consortium brings together extensive expertise from medicine, research, industry and law. The annual congress demonstrated just how important this exchange is. The next step is to test the methods developed on specific datasets and usage scenarios together with partners from practice,” explains Thomas Köllmer, data protection expert at Fraunhofer IDMT.

Protecting EEG data and tracing its origin

The congress presentation on the EEG signal analysis use case across different sleep stages demonstrated how anonymization, controlled data sharing and traceability can work together. In cooperation with medical technology manufacturer eemagine Medical Imaging Solutions GmbH and TU Ilmenau, Fraunhofer IDMT is investigating how such sensitive biosignals can be anonymized while remaining usable for further analysis. In addition, the institute is researching digital watermarks that allow information to be embedded directly in the EEG signal. This makes it possible to trace the origin of the data, document processing steps and label synthetic data.

Initial tests showed that the watermark under investigation is robust against various signal modifications without affecting the sleep-stage detection examined.

From research to pilot projects

Over the next two years, the methods investigated in the AVATAR-Transfer project are to be gradually integrated into a platform. The platform is intended to enable health data to be processed in anonymized form and securely provided for secondary use in research and development. The methods will then be tested in specific applications. In addition to clinical biosignal data, these include hearing health, ophthalmological rehabilitation, post-market surveillance of medical devices and genomic data.

The German Federal Ministry of Research, Technology and Space is funding AVATAR-Transfer from January 1, 2026, to December 31, 2028. Companies and medical institutions wishing to assess and test anonymization methods for a specific dataset or usage scenario are welcome to contact Fraunhofer IDMT.