Dear all,
I would like to share two videos I discovered yesterday by Prof. Dr. Christian Johner, head of the Johner Institute and, in my experience working in medical technology, _the_ German thought leader on everything new in this field—from market trajectories and technological opportunities to regulatory issues.
You can find his videos here (German only, unfortunately):
– https://lnkd.in/eT8b984C
– https://lnkd.in/e37mDKqW
In these videos, Prof. Johner experiments with the AI-based design of a defibrillator and a diabetes management system. A defibrillator falls under the highest risk class of the Medical Device Regulation—so hardly an easy task.
Given the slow-moving nature of European regulations, I had assumed AI would not make inroads into medical technology anytime soon. While AI is already used to support practitioners with recommendations, it will not take an _autonomous_ role in treatment except in very narrow cases. Safety is the main reason: even a well-trained neural network can behave unpredictably in edge cases. And who wants to find out during treatment that they belong to the “edge-case group”. Active, autonomous treatment with AI will only be conceivable when it can be proven to outperform humans—e.g., with fewer errors and better outcomes than a surgeon.
A far more promising application is product development, both hardware and software—precisely what Prof. Johner explores. From my experience as a product owner (software) for a surgical device and my extensive work with AI, I can confirm that product design could benefit greatly—shortening innovation cycles and potentially improving quality. What remains unclear is how regulators will treat AI-generated code. Hopefully, Prof. Johner will also shed light here.
One thing is certain: what matters is not _how_ code is created, but that it is traceable to requirements, validated for quality, and risk-assessed. Requirements engineering, validation, and verification lie at the core of medical device development. With test-driven development (see my article on AI swarms for agile TDD in Heise / iX) and human oversight, quality can be assured—while saved time enables companies to reach new levels of performance.
To see what is possible, I strongly recommend watching Prof. Johner’s videos.
Medical device manufacturers will undoubtedly adopt these tools, whether openly or quietly. If they don’t, competitors will replace years of development with months. No company can afford to ignore this opportunity.
At the same time, ensuring the highest quality in new medical devices is in everyone’s interest. To maximize opportunities while minimizing risks, I encourage companies to share their experiences openly.
If you are considering a pilot project using AI in medical device software development, feel free to contact me. I am passionate about both AI and medical device software and available for consultation.
Have fun,
Rüdiger