The new Team NB Position Paper, introducing the Questionnaire: ARTIFICIAL INTELLIGENCE IN MEDICAL DEVICES, has been published. The document aims to provide guidance for manufacturers of AI-based medical devices under EU regulations.
As outlined in the document, medical devices incorporating or using AI systems may be placed on the EU market, provided that appropriate conformity assessment procedures are conducted in accordance with the MDR/IVDR and the horizontal requirements of the AI Regulation as they become applicable. However, practice has shown that manufacturers face significant challenges in demonstrating conformity for AI devices, particularly those relying on in-field self-learning mechanisms. Currently, notified bodies do not consider such devices “certifiable” unless manufacturers implement robust measures to ensure safe operation within the validation scope outlined in the technical documentation.
This guideline offers AI-specific action steps to assist manufacturers in demonstrating conformity. Most notably, the questionnaire emphasizes that the safety of AI-based medical devices can only be achieved through a process-oriented approach, requiring consideration of all relevant processes and life cycle phases. Rather than focusing on specific product requirements, the guideline prioritizes robust processes.
While the document does not claim to be exhaustive or mandate application, it serves as a practical tool for manufacturers to verify that all critical aspects are addressed, ensuring the systematic safety of AI-based medical devices.
The document should be understood within the context of the MDR, IVDR, and MDCG guidance on Medical Device Software (MDSW). While the specific requirements of the Artificial Intelligence Regulation (AI Regulation (EU) 2024/1689) are not yet included, as explained in the position paper, these will be addressed in a future comprehensive revision. Nevertheless, there is considerable overlap between the AI Regulation and the existing regulatory and standards framework for software and AI, as noted in the position paper.
