European test infrastructure for AI in healthcare

Robots assist in the operating room, monitoring systems help to detect diseases earlier and medically monitor chronic diseases: Artificial Intelligence is increasingly being used in hospitals, doctors’ surgeries and laboratories. It supports workflows in the healthcare sector and makes care more efficient. But how secure and trustworthy are such AI-based applications? There are currently hardly any standardised processes and infrastructures for testing and validating such products adequately under real-life conditions. However, this is now also mandatory at European level for the health sector. This is where TEF-Health comes in, a cross-border “Test and Experimentation Facility for Health AI and Robotics.” Prof. Dr. Petra Ritter, Head of the “Brain Simulation Section” at the Berlin Institute of Health (BIH) at Charité – Universitätsmedizin Berlin, coordinates the consortium. In the Brain City interview, she gives an insight into the TEF-Health project.

Prof. Dr. Ritter: the TEF-Health project was launched in early 2023. Why right now? 

There is an important reason for this: the AI Act came into force across Europe in 2024. The regulation is intended as a comprehensive legal framework to ensure that Artificial Intelligence systems are developed and deployed in Europe in a secure and trustworthy manner. Of course, this also applies to the medical sector. Specifically, this means: from 2026, all medical devices that contain AI systems – i.e. that create more complex mathematical calculations based on patient data – must be tested and certified for safety and trustworthiness in accordance with the AI Act. There is currently still a period of grace. However, these requirements pose a major challenge for startups and SMEs in particular. Certifiers also have to adapt to the new AI Act first. Standards, norms and test protocols must be defined. Companies, on the other hand, need secure test and experimentation infrastructures in which they can test their AI systems and which also provide access to health data. This is where TEF-Health comes into play.

What exactly does TEF-Health offer? 

With TEF-Health, we primarily want to test new AI approaches in realistic scenarios. This includes new software used in areas such as patient care and diagnostics, as well as surgical and nursing robots. We will examine how we can facilitate market access and acceptance of these smart technologies. This requires cloud infrastructures, high-performance computers and physical environments such as operating theatres or intensive care units. As the European “Testing Experimentation Facility for Health AI and Robotics”, TEF-Health is a consortium of 52 organisations, half private and half public, under the leadership of Charité - Universitätsmedizin Berlin. Together, we currently offer a repertoire of services for SMEs and startups. In the longer term, we would like to address the entire healthcare industry. This requires many different services, which we bring together on the platform.

And how does the platform bring such service providers together with companies?

The TEF-Health database currently contains around 340 services from providers in eight European countries. These can be filtered and viewed by category via a search on the website. If you click through, you will quickly find very detailed information with price examples for each individual service. Companies can then use the “Apply for Discount” button to apply for a service. During the first five years in which the project is still being subsidised, we can still offer these services at reduced prices. The application is then checked and evaluated by us for its suitability. If the result is positive, a matchmaking process then takes place between the company and the service provider. These matches can take place across the borders of the member countries. This means that an SME from Germany, for example, can utilise services from Portugal or Sweden or vice versa. This leads to much greater networking and also to the utilisation of synergies.

Does the project only offer infrastructure – or are the services also linked to research?

Our target group are providers of innovative products that have at least a Technology Readiness Level (TRL) of six. This is the prototype phase, in which validation and certification take place and clinical trials take place. We don’t actually accommodate the research and development phases below TRL six. We occasionally make exceptions. In principle, however, we do not offer research services, but only validation, testing and support to make product innovations ready for the market.

How are you involved in the TEF-Health project? Does your specialisation in brain simulation play a role? 

I coordinate the European test infrastructure. Due to the requirements of brain simulations, for many years our working group has been developing cloud infrastructures with high-performance computing capacities for personal and health data – i.e., for digital twins of brains. Initially within the framework of the European Open Science Cloud (EOSC), but also in follow-up projects that were funded with many millions. We have developed virtual, secure research environments that form an important basis for such test infrastructures. And of course, brain simulation and AI are closely linked, because Artificial Intelligence is inspired by information processing in neural networks. I found the task of coordinating the project exciting from the very beginning. TEF-Health has many aspects that we can serve and enrich with our developments and expertise.

The TEF-Health project was launched in early 2023. How did more than 50 partners come together within two years?

I originally received the information that the European Commission was planning a corresponding call for proposals via Berlin Partner. This was communicated through various channels at the time, and a key requirement was that the member states of the project together provided 50 per cent of the total funding amount of 60 million euros. I then activated or made new contacts at a large number of meetings and built up the consortium. This was already in its current form by the application deadline in May 2022.

How is the consortium composed?

The composition of the partners goes far beyond pure research expertise. In addition to hospitals, universities and clinical institutions, standardisation authorities from member countries are also represented in the consortium. For example, the Physikalisch-Technische Bundesanstalt (PTB) or its French counterpart, the Laboratoire national de métrologie et d'essais (LNE). Organisations of certifiers such as the TÜV Association and the TÜV AI.Lab, which was founded from the TÜV Association, are also taking part. In addition to Charité and the Berlin Institute of Health (BIH), partners from the Berlin area include the KI Park, the Fraunhofer Heinrich Hertz Institute (HHI) and, of course, Berlin Partner for Business and Technology. 

And how long will the TEF-Health project run?

There is no end to it. The project is to continue as a working network after the end of the initial five-year funding from the European Community and the member states. This was stipulated in the funding decision. The project is intended to generate income, but this is not enough to finance the further development work of TEF-Health with all its facets, as stipulated in the funding decision. That is why we are currently obtaining financial commitments from the member states. The fact that European countries are obliged under the AI Act to provide companies with so-called “regulatory sandboxes” for their AI systems helps us here. Real-world laboratories in which companies can utilise the same infrastructure and support structures that we have developed with TEF. The individual EU member states must provide permanent resources for this. This can also take place across national borders in the form of consortia. European Digital Infrastructure Consortia (EDICs) are ideal for this purpose. The advantage is that the EU Commission also makes a financial contribution here. TEF-Health is therefore to become a permanent institution. This will not only benefit the European healthcare system, but all of us. With TEF-Health, we want to help ensure that the digital transformation in the healthcare sector is secure and transparent.”

Interview: Ernestine von der Osten-Sacken