EHS Public Debate: experts weigh in on EU’s AI Act in healthcare

Six experts from different fields gathered for the European Health Summit Taskforce Public Debate last 24 February 2022 to discuss AI in Healthcare.  i~HD President Prof. Dipak Kalra moderated the debate with participants Yiannos Tolias (Legal Officer, DG SANTÉ, European Commission), Elke Grooten (Head EU Relations, Novartis), Michele Calabrò (Senior Manager for Digital Health Policy, Digital Europe), Sara Roda (EU Senior Policy Adviser, Standing Committee of European Doctors), Cornelia Kutterer (Senior Director, Rule of Law & Responsible Tech, European Government Affairs, Microsoft) and Birgit Bauer (Digital Health & Social Media Entrepreneur and Consultant, Patient Expert, Manufaktur für Antworten UG). 

Exciting use-examples 

The digital discussion kicked off when Prof. Dipak asked the participants about health use-examples of AI that personally excites them the most. 

For Digital Europe’s Calabro, it is the potential that AI has on all levels of healthcare starting from the individual level. “It also has strong potential to improve population health management and operation and in strengthening health innovation.” 

Meanwhile, it is about the possibility to revolutionise healthcare in a positive way for the Standing Committee of European Doctors. “One would be increasing the accuracy of diagnosis and the efficiency of treatments,” said Roda. 

Novartis emphasised that AI is already part of their business and research processes. According to Grooten, there are currently around 100 use cases of AI running in their pharma company. “I just want to give one example where AI is used: for the detection of diabetes with patients through the scanning of their eyes,” added Grooten. 

Kutterer found it fascinating how AI can augment health caregivers to make better decisions, and as a result, augment patients’ lives. She discussed “Project Tokyo”, a research case study undertaken by Microsoft’s research lab in Cambridge, UK. “A system is built on a whole new lens augmented by facial recognition to help blind people in guiding students to circulate more freely in classrooms… This augmentation enabled [visually-impaired] people to do what seeing people do.” 

Bauer related her personal experience to AI in healthcare. “As a person living with MS [multiple sclerosis], I see AI growing, analyzing scans for example, to find out if a disease is progressing or not, and also to have a faster diagnosis, also better recommendations for a treatment or not, or also what to do next.” 

AI and data quality 

Coming from his experience in working on the AI Act of the European Commission, Tolias explained that to create public trust and acceptance of AI, the AI systems should be performing well. In the healthcare sector, human-AI interactions must be clear and appropriate. “The AI Act… makes it a bit stricter because it says that the user, i.e. the physician, should be using the AI in accordance with the instructions provided.” He also highlighted the importance of data quality when rolling out AI. According to him, data quality is actually one of the main core provisions of the AI Act in general.

Moderator Prof. Kalra also summarised one key challenge in developing AI systems. “It’s not only going to be about creating AI that works in the lab, but when you put it in the real world with all the mess that there is, with temporary staff coming in who have not been trained, or some system or network connection goes down. or some data is not available for the AI to reason on. The real world is really messy. It’s not anything like as clean as where the labs are.” 

Multisectoral reactions (and collaboration) 

Experts from the tech and digital sectors all welcome the AI Act as formulated by the European Commission. However, they also mentioned the need to continually monitor developments in the technology and the evolution of related regulations. 

“It can bring a lot of trust and assurance to all stakeholders , but it needs to be done in a clear way, and actually provide the right explanation because this is going to be very, very important,” opined Calabro. 

Kutterer agreed. “That is why we have the office for responsible AI. They have developed an internal standard. And because our ethical principles are fairly consistent with other global principles, including the high-level group on AI principles that have been developed, there is some consistency with the technical requirements that I’ve mentioned before in the AI Act.” 

Bauer meanwhile proposed that patients should have a more active role in the development of AI systems in healthcare. 

“Of course, patients cannot design AI. But they can have that voice and they can say what their needs are, and this would be the AI of my dreams. I would work with these people and I want to learn more about that and I think this is important then because we can learn from each other. The developers can learn from us and we can learn from them and this also grows, as we said before, the trust, the acceptance, and the understanding, and I think this is what we need. We need to come together,” concluded Bauer. 

A replay of the debate, as well as more information on the European Health Task Force, is available via https://ehtaskforce.eu/ 

Related news items: