AI medical device software has the potential to greatly improve healthcare. This entails the compliance to regulatory requirements to safeguard patient safety and ensure the benefit-risk of new technologies. Read our latest white paper for recommendations regarding automation bias and performance conformity assessment in a changing European regulatory landscape for the medical devices industry.
The digitalization of healthcare generates a massive amount of data that exceeds the analytical capabilities of individual clinicians. Artificial intelligence (AI) has the potential to automate the analysis of these data, carrying the promise of better and more sustainable healthcare. However, doing this requires AI medical device software to be as safe and performant as any other trustworthy medical device.
Significant progress has been made in artificial intelligence (AI) over the past few years. However, substantial work still needs to be done to understand the technical and clinical aspects that make AI software systems different from established medical device software and the impact they have on their regulatory conformity assessment in Europe. Without explicit regulatory guidance, some conformity assessment aspects of AI technologies remain unclear.
This paper reviews some key aspects of safety and performance assessments of AI medical device software according to the existing regulation (EU) 2017/745 (MDR) ahead of the upcoming regulatory updates and standards publication.
Discover more about DNV:
DNV is an independent foundation with over 150 years of history in assurance and risk management. We use our knowledge to advance safety and performance, set industry benchmarks, and inspire and invent solutions to tackle global transformations. DNV’s Healthcare research programme works with partners through large-scale public-private research and innovation projects (see BigMed, the Nordic Alliance for Clinical Genomics, REALMENT and AI-Mind to name a few) to assess trustworthy AI adoption, enable data sharing opportunities through federated data networks, meet patient autonomy needs with dynamic consent, explore quality driven benchmarking approaches in clinical genomics and assure the safe bringing-to-market of new AI medical device software solutions.
