New FDA Guidance Issued for AI-Powered Medical Devices
The FDA has recently published a guidance whitepaper that will eventually underpin a framework for the regulation of AI products in medicine.
Join the DZone community and get the full member experience.Join For Free
When you think of AI in healthcare, it's perhaps common to think of software products that can help provide the kind of earlier diagnoses promoted by companies like DeepMind and IBM. AI is increasingly being used in medical devices, however, so it's interesting to see that the FDA has recently published a guidance whitepaper that will eventually underpin a framework for the regulation of AI products in medicine.
It represents the clearest indicator from the FDA that regulation is coming to AI-driven products in healthcare. The paper highlights the various criteria the FDA believes will be used to determine just how much that medical products rely on AI, and therefore require FDA approval before being sold to customers.
"A new approach to these technologies would address the need for the algorithms to learn and adapt when used in the real world," the FDA says. "It would be a more tailored fit than our existing regulatory paradigm for software as a medical device."
The paper is interesting as it marks the latest in an ongoing race between regulators and industry to keep up with the rapid pace of change in the sector. It illustrates the direction of travel, however, and highlights the importance of AI-driven applications securing proof from clinical, real-world environments.
It's easy to believe that regulation is purely a restrictive force on the development of AI in healthcare, but if done properly, it can actually help support the ability of AI to improve and learn over time.
The whitepaper marks the latest step from the FDA to understand the space, and they already approve medical devices that utilize 'locked algorithms,' which don't change each time the algorithm is used but are instead updated periodically by the manufacturer as a result of updated training data.
It perhaps goes without saying that these locked algorithms present a relatively static target, and are therefore much easier to regulate. When algorithms start learning on their own, it becomes infinitely harder to ensure that the life-and-death decisions these algorithms are making are based on the best, most accurate information. Nonetheless, the agency accept that it's vital that they try, as the benefits from the technology can be considerable.
"Artificial intelligence has helped transform industries like finance and manufacturing, and I'm confident that these technologies will have a profound and positive impact on health care," they conclude. "I can envision a world where, one day, artificial intelligence can help detect and treat challenging health problems, for example by recognizing the signs of disease well in advance of what we can do today."
Published at DZone with permission of Adi Gaskell, DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.