Skip to main content

Software & AI as Medical Device

Overview​

This page covers advanced considerations for software medical devices and AI/ML-based devices. For foundational qualification and IEC 62304 requirements, see Software as a Medical Device and Cybersecurity.

AI/ML Device Lifecycle Considerations​

AI/ML-based medical devices present unique lifecycle challenges because the algorithm may continue to learn or be updated post-market:

  • Predetermined Change Control Plan (PCCP) — describes planned algorithm changes and the associated risk management and performance monitoring framework; reduces the need for new conformity assessment for every update
  • Algorithm transparency — training data sources, data preprocessing, model architecture, and validation methodology must be documented in technical documentation
  • Performance monitoring post-market — real-world performance monitoring is integral to the PMS plan for AI/ML devices; distribution shift and model drift must be detected and managed

Swissmedic Expectations​

Swissmedic aligns with MDCG guidance and is monitoring the EU AI Act's implications for high-risk AI systems. For AI/ML medical devices, Swissmedic expects: comprehensive intended use and intended user specification; full documentation of the training/validation/test data split; an ongoing performance monitoring plan integrated with PMS.

Disclaimer

AI-assisted content for navigation only. Always verify against official Swissmedic and Fedlex sources. Not legal or regulatory advice.