Artificial Intelligence in Medical Devices - Notified Bodies' Point of View

In November 2024, "Team NB" published a 28-page questionnaire-style guideline on the topic of "Artificial Intelligence in Medical Devices".

Content of the Guideline

The document was written to provide guidance on the use of artificial intelligence in medical devices. The basic idea is that the safety of artificial intelligence (AI)-based medical devices can only be achieved through a process-oriented approach. All relevant processes and phases of the life cycle must be taken into account. Therefore, the guideline does not contain any specific requirements for products, only for the processes.

Interestingly, the authors point out in a preliminary note that specific requirements on artificial intelligence coming from the EU Regulation (2024/1689) were not considered in the guideline. They will be included in an update of the questionnaire at a later date. However, the authors do see some overlaps between the EU Regulation and other regulations relating to artificial intelligence (AI).

In principle, Annex II (4.) of both the Medical Devices Regulation and the In-Vitro Diagnostics Regulation must be observed when it comes to software for medical devices. This also applies to medical devices using AI. According to the guideline, validation is always necessary for such products. It is part of the technical documentation and will be reviewed by notified bodies.

Certification for "self-learning systems" is currently difficult unless appropriate validation testing regarding the safety of the product is included in the technical documentation, the guideline states.

A list recommends that manufacturers consider the following aspects, either in instructions or in plans, to ensure the safety of the product:

• Development
• Risk management
• Data management (property of the customer)
• Verification or validation (if not part of the development)
• Surveillance and vigilance after market launch
• Software lifecycle, including maintenance, installation and decommissioning
• Configuration management
• Training and qualification
• Customer communication
• Procurement, suppliers and supply chain control
• Management review

Subsequently, the document provides a list of necessary questions to consider when using AI. This is in the style of a questionnaire and lists the questions in great detail with references to regulations, standards, etc.

Conclusion

Artificial intelligence in medical devices is certainly the way of the future. Considering that, the NB team's questionnaire offers valuable input.

The document is available in full on the Team NB website.