Skip to main content

AI/ML Medical Device Guidance

Overview​

Artificial intelligence (AI) and machine learning (ML)-based medical devices are regulated as medical devices in Switzerland where they meet the MedDO intended purpose criteria. Switzerland follows EU MDR/IVDR and MDCG guidance (particularly MDCG 2020-1) for AI/ML medical devices, with no separate Swiss AI/ML regulatory framework.

AI/ML Device Qualification​

An AI/ML system is regulated as a medical device if its output is intended to drive clinical decisions (diagnosis, treatment planning, patient monitoring). An AI/ML system used purely for administrative or operational purposes (e.g. scheduling, billing optimisation) is not a medical device. The qualification analysis follows the same MDCG 2019-11 framework as other SaMD.

AI/ML-Specific Technical Documentation Requirements​

Training data documentation The technical documentation must include: description of the datasets used for training, validation, and testing; data preprocessing steps; demographic and clinical representativeness of the training data; data quality assurance measures.

Model description and performance Required: model architecture; performance metrics on the test dataset; performance stratified by relevant subgroups (where applicable); statistical uncertainty in performance estimates.

Algorithm change management AI/ML algorithms may change over time (learning, retraining, updates). The technical documentation must address: how algorithm changes are managed and validated before deployment; what constitutes a "significant change" requiring new conformity assessment; how performance drift is detected and managed post-market.

Predetermined Change Control Plan (PCCP) A PCCP describes the types of algorithm changes the manufacturer anticipates post-market and the associated evidence standards that will be applied before implementing those changes. A well-developed PCCP can reduce the regulatory burden of iterative algorithm improvements by pre-agreeing the change management framework with the NB.

Post-Market Performance Monitoring​

AI/ML devices require specific post-market performance monitoring:

  • Real-world performance metrics collected from deployed systems
  • Monitoring for performance drift (degradation of algorithm performance over time due to population or data distribution shifts)
  • Processes for detecting and responding to unexpected algorithm outputs
  • Regular comparison of real-world performance against performance claims in the technical documentation

EU AI Act Implications​

The EU AI Act (entered into force August 2024) classifies high-risk AI systems, including AI used as or in medical devices. While the EU AI Act applies directly only within the EU, Swiss manufacturers or manufacturers accessing both EU and Swiss markets must consider EU AI Act compliance for medical device AI systems. Swissmedic is monitoring EU AI Act implementation for potential alignment implications.

Official Sources​

Disclaimer

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