The EU risks being left behind in AI, as it was in the social media revolution. This suggestion was made by Rony Medaglia, professor at Copenhagen Business School, at a workshop yesterday in Dubai (16 March), which showcased leading European healthcare AI projects as well as the EU’s approach to AI.

The block missed out on the innovations of social media and failed to play a leading role, a position held by North America and Asia, Medaglia argued, saying that now is the time for the EU to leverage its value-based approach.

“Europe is always the one that comes and gives fines to the ones who break trust,” he said. But it needs to be a player. “If you want to win the Champions’ League you cannot just be a referee.”

Other speakers included Karim Lekadir, professor at the University of Barcelona and project leader of EUCanImage, Federico Álvarez-Garcia, professor at Polytechnic University of Madrid and project leader of Genomed4all, Fosca Giannotti, from the University of Pisa, Isabell Tributsh, project manager at the University of Barcelona and Malavika Jayaram, executive director of the Digital Asia Hub. Gianluca Misuraca moderated the discussion, with Signe Daugbjerg, both from IntouchAI.eu.

Key themes that emerged were how to translate AI research into clinical practice and embed trustworthiness into these developments, winning the confidence of patients and clinicians.

Challenges with AI implementation?

AI presents a new and unique challenge, according to Daugbjerg, who is involved in a project using AI to diagnose dementia. Daugbjerg’s background is in epidemiology, healthcare systems, and health technology assessment, but she found that AI—where algorithms continually evolve and update—differed from both telemedicine and ICT. The multiple questions it engenders are entirely new, she said. “We have to develop a framework that is tailored to artificial intelligence.”

Giannotti said that AI is not yet used much in health systems. One potential hurdle is a lack of integration with clinical guidelines. Innovations are sometimes overlooked because they do not integrate with clinical guidelines, according to Lekadir. “We need to work with professional societies,” he said. “That’s how you go from something that is in research—and has been proven for trustworthiness—to something that would actually be used.”

A role for legislation

A tension between protecting patient privacy and the need to give researchers access to data is common in discussions about AI in healthcare. But Jayaram felt that the dualism is flawed. Both are possible, she argued. “Seeing it [privacy] as an either-or is very very problematic,” she said. “You can have privacy and trust, you can have adoption and security.”

The panelists agreed that earning trust in this rapidly evolving sphere is a complex and vital task. For Jayaram, legislation will play a central role. Health data are so uniquely sensitive that the risks of harm are high. “Having strong data protection laws and regulations is essential to building trust,” she said.