Software in Medical Devices, by MD101 Consulting

This article is the first occurence of AI/ML on this blog. AI/ML in medical devices has been overlooked, mainly by lack of qualifications on this subject. Things are changing in 2025, AI/ML in medical devices is now a major subject. It's time to deal with it.

We will have a series of articles on AI, exploring the different facets of AI/ML in medical devices:

• The brand new draft FDA guidance on AI-enabled device software functions,
• The FDA guidance on Predetermined Change Control Plan (PCCP),
• On the EU side, the checklist on AI published by the Team-NB,
• The AI Act (the AI Act makes consultants spinning on their chairs, seeing a great opportunity of business and revenue. Needless to say there's no such excitement from medical device manufacturers, who see the AI Act as a constraint and a source of costs),
• More technical: the AI model lifecycle and AI data management lifecycle,
• Standards about AI: it's the jungle out there and we'll need to clarify what can be relevant / applicable for medical devices.

Releasing a ML model with an appropriate performance is a major concern with AI. Performance validation or clinical validation is thus a major subject in ML model design files.
However, the editorial line of this blog is to exclude any subject about clinical validation. By lack of competence, of course.
Given the importance of clinical validation for ML models with data-driven design, the subject of clinical validation will be tackled. But only staying in general considerations on AI lifecycle, and never diving into details.


Important note: no AI software, no LLM were harmed, writing the articles published on this blog. And we will continue this way. This blog is and will remain handmade and LLM-free.