Technological advances in AI and robotics have been unprecedented and have captured the imagination of companies, start-ups and researchers alike, with no signs of slowing down. Healthcare is no exception, but new medical devices and procedures must be proven safe and useful before they can be used on patients. And while the European Union has high quality requirements, researchers and entrepreneurs are faced with insufficient testing infrastructure to develop standards, test innovations and certify new products.
The new EU-funded project “Testing and Experimentation Facility for Health AI and Robotics” (TEF-Health), with a total budget of around €60 million, aims to facilitate and accelerate the validation and certification of AI and robotics in medical devices. In conversation with project lead Prof. Dr. Petra Ritter, Director of the Berlin Institute of Health (BIH)and the Brain Simulation Section at the Department of Neurology with Experimental Neurology at Charité and project partner Dr. Dirk Schlesinger Head of the TÜV AI Lab we learned more about their innovative approach, the importance of data accessibility and being at the heart of Berlin's AI landscape.
We are so happy to have you here today to talk about this ambitious project. Could you briefly explain the project, the consortium and what the TEF-Health is all about?
Prof. Dr. Petra Ritter: Thank you very much for having us. The new project is called: “Testing and Experimentation Facility for Health AI and Robotics” (TEF). It's a big consortium with 51 partners, 9 participating countries and several European institutions. The goal of the project is both to bring innovations in the field of AI and robotics to market and to support our future customers, that are startups and SMEs, in bringing their innovations to market.
Dr. Dirk Schlesinger: If I had to summarize the whole project, it would be about three things: First, it's about data, because without data there is no AI, and medical data is sensitive, at the same time rare and needs to be shared. How can you resolve this contradiction? Second, it's about infrastructure, we're talking about learning systems; so we need an infrastructure of supercomputers, labs, and basically places where use cases can be executed. And third, it's about certification. It's about the trust that we need to create as well as the process of how we can bring this innovation to the market. And we need to do it quickly, without consuming too many resources. That's the goal of the project.
Can you explain the idea behind the project and how you plan to proceed?
PR: TEF-Health is organized into nine nodes-each country forms its own node. Based on use cases, we work with startups and SMEs to guide and support the process of certification and validation of applications.
DS: That's a very important point, because with this project we want to pave the way for what we call - for lack of a better term- "agile certification". In today's process, you build something, it's ready, then someone comes and audits it, then it gets certified, and overall it takes forever. We can't do that anymore in AI and especially in medical innovation, it has to be much faster. So we want to develop a process that can be parallelized. Now it gets complex, because there is the research, the company that makes medical devices, the auditor, and finally the certifier, who has to be an independent third party. That's why we need the ecosystem in TEF Health. We need to bring all these players together in one place and then make sure that they have the use cases that Petra was talking about, because that's what process development does practically. This is not an exercise on paper.
Then let's get down to business: Can you elaborate on any of these use cases?
PR: Let's take the use case of the brain. In the event of damage to the brain, for example due to a stroke, patients sometimes have problems with the coordination of their extremities. Or if the spinal cord is damaged, and in this case the robotic exoskeleton is used and developed for patients to enable movement. This would be a use case where digital twins of the brain could be directly connected to such a robotic device to assist the patients.
What is the AI component of the data and the technology behind it?
PR: The brain contains very complex data that provide information about the processes taking place. Comprehensive individual information is needed to develop health applications for the brain that are useful for patients. It is not expedient to anonymize the data of the patients, as this would result in the loss of important individual information. You can't take the complexity or the biometric information - which is as revealing as a fingerprint - out of the data, because then they would basically be useless for most questions. So we need ways to integrate the personal data in a privacy-compliant way to create digital twins of the brain and enable, for example, mechanistic simulations that can be usefully combined with machine learning and artificial intelligence.
DS: If I may go into "geek land" for a moment: We're going to work on "federated learning", which means that it's not the data that moves from hospital to hospital, but the model, so we have more data to train the AI. We are talking about homomorphic encryption to ensure privacy of data. There are many different approaches that help ensure that where we think we are at a disadvantage to China and the U.S., we can remedy that because we have the technology to be able to safely and securely provide data for AI training.
Data is a sensitive topic, especially when it comes to health data. Why is data so important and how can we protect it?
PR: Yes, health data is indeed a very sensitive issue. They contain personal information about individuals and therefore need to be protected, and we have laws in the European Union, such as the GDPR, that ensure the protection. Our goal is to provide access to sensitive data in a secure way, so that patients get the protection they need and at the same time researchers and industry get the access they need to validate their AI and robotics applications. Of course, the GDPR sets the bar high for implementation, but it can also be an advantage because it's a way to gain patient trust.
DS: Exactly, it can be a competitive advantage if we do it right. Right means fast and easy and still safe, which is certainly a challenge for the project. If you look at the innovation cycle in AI, it's much faster than regulation and legislation. So how can we - and this will be an additional work package - speed up the process and still be compliant with legislation? Certifications according to the Medical Devices Regulation, which ensure that the product on the market is of high quality, are mandatory and we will not and do not want to shake that.
Can you talk a little bit about that and the intricacy of the data, what are you hoping to accomplish? What are the goals for the patient?
PR: Ultimately, the goal is to bring the health-innovations made in Europe to the marketplace. And that requires much more than just research. It needs validation, it needs building trust, reliability, explainability, generalizability. These are all key words that need to be put into practice, which means that the application or the innovations need to be tested, because the applications, they can decide about the health of an individual, they can decide about the well-being of people. So we need a lot of testing and validation before these innovations can really be brought to the patient.
DS: Right, and it has to be fast and at the same time thorough. To be very honest, if you look at the EU-certification process for medical devices, we actually have a shortage of certifiers right now. So another challenge for us is not only to make it faster, but also to make it resource light so that: A) the capacity of the certifiers is sufficient and B) that the company that certifies their product doesn't have to spend enormous amounts of resources and money to get their product to market.
Why is this project in Berlin and how are you managing it?
DS: There are several reasons for that. First of all, why Berlin? We have the largest ecosystem of AI companies here in this city. Only if we tackle the problems together can we come up with a solution that really helps the patient. Secondly, Petra is based in Berlin and leads the project on a European level. She basically directs the project.
PR: Charité is one of the largest university hospitals in Europe, and as Dirk said, we have a very vibrant research environment. There are many developments in artificial intelligence, both in academia and in industry. This combination of leading medical research and a team to support it, together with this environment and the ecosystem that is being created in Berlin, we are very well equipped to coordinate this major initiative.
DS: It's all about talent, and there is talent here in Berlin. To be honest, it's also about funding, and the Berlin Senate has been kind enough to co-fund the project, and Petra has taken on the responsibility of leading the project from the beginning. So it's a very natural place for everything to come together.
PR: Indeed, we have been very well supported. It is very important to note that 50% of the funding comes from the Member States and in our case the Berlin Senate is very helpful in providing co-funding in addition to the Berlin Institute of Health (BIH), which is also supporting the initiative, and we have other supporters like Berlin Partner for example, which is also a very relevant contributor. We also work with institutions like the Physikalisch-Technische Bundesanstalt (PTB) – Germany’s national metrology authority and very important other stakeholders.
What are the next steps for TEF-Health?
PR: The project has started its work. We had a kickoff event in January, and the partners now meet in weekly coordination meetings. This will ensure that all strings come together so that in the end we create a consistent infrastructure and make optimal use of the synergies of the individual partners. In June, an interdisciplinary launch event is planned in Copenhagen - together with the three other funded TEFs in the areas of smart cities, agriculture and crafts.
Thank you both very much for your insights and all the best for the future of TEF-Health.
*This interview was conducted by Amira Gutmann-Trieb, Cluster Manager at ICT, Media and Creative Industries at Berlin Partner.