From niche solutions to thousands of approved products
AI in Medtech is nothing new, but its development has accelerated dramatically. In the U.S., the FDA has approved more than 2,300 AI- or machine learning-based medical devices since the mid-1990s—nearly 1,000 of which have been approved in the last five years. Radiology dominates, accounting for three-quarters of all solutions, followed by cardiovascular and neurological applications.
Why radiology, specifically? The answer lies in the availability of enormous amounts of data and a global shortage of radiologists. AI has become a powerful tool for analyzing images and reducing the workload on specialists. But even though the technology is advanced, it is not autonomous. Most CE-marked products in Europe are Class IIa, which means they serve as decision-support tools, not as substitutes for clinical judgments.
Regulations That Shape Development
Technology is advancing rapidly, but regulations are moving slowly. In Europe, the transition to the MDR has led to stricter requirements for software, which has slowed development in certain areas. At the same time, new regulations such as the AI Act are being introduced, which apply in parallel with sector-specific regulations. The AI Act classifies many medical device AI products as high-risk and imposes requirements regarding, among other things, cybersecurity, transparency, and quality management systems.
Henrik Fält Padoan, a consultant in Quality and Healthcare, puts it this way: “Regulations are updated very slowly, and technology is moving faster. But we can only hope that the regulations will keep up as closely as is practically possible.”
It is a balancing act between patient safety and innovation. Regulations that are too strict risk stifling development and creating a shortage of new solutions. Regulations that are too lax jeopardize safety. The European Commission is currently working to evaluate the MDR and IVDR to find the right balance—an effort that could be crucial for future competitiveness.
Interoperability—the major bottleneck
Technology depends not only on regulations, but also on data. A fragmented landscape with heterogeneous systems makes it difficult to implement AI effectively. This is where the European Health Data Space (EHDS) could be a game-changer. By facilitating the secondary use of data, it opens up new opportunities for training AI models, particularly for rare diseases.
But interoperability is more than just technology—it’s a matter of collaboration, standardization, and trust. Without a common foundation, AI risks becoming an isolated solution rather than an integrated part of the care pathway.
Opportunities for the Future and the Next Steps
The needs are clear. According to the EU, there was a shortage of 1.2 million healthcare professionals in 2022. AI cannot solve the underlying problem, but it can streamline administrative tasks, improve patient flow, and free up time for what really matters—such as the interaction between healthcare staff and patients.
At the same time, new questions are just around the corner. What constitutes a safety component? How do we handle changes in AI models over time? And how do we classify borderline cases, such as administrative systems that sometimes become medical devices? Expert groups, regulatory agencies, and standards organizations are working intensively on guidelines, but the technology continues to outpace them.
Henrik sums it up well: “The important thing here is to ensure that patient safety remains the focus, without hindering progress or neglecting the major overarching societal issues.”
Are we ready for the next step?
AI in healthcare is no longer a vision of the future—it’s here and now. But unlocking its full potential requires more than just algorithms. It requires collaboration, clear regulations, and a shared commitment to balancing innovation with responsibility.
Want to know more?
Listen to our podcast episode here or contact us today to find out how we can support your organization.
