The FDA is considering a total product lifecycle monitoring approach to emerging artificial intelligence devices, the agency said in a proposed rulemaking issued April 2.
Regulators may also consider what the agency calls a predetermined change control plan, with “detailed information … about the types of anticipated modifications based on the algorithm's re-training and update strategy,” Commissioner Scott Gottlieb said in announcing the proposed rule.
Gottlieb said the goal of the framework is to “assure that ongoing algorithm changes follow pre-specified performance objectives and change control plans” and that the process includes validation and real-world monitoring of performance once the device is on the market.
The FDA has already approved two AI-driven devices that use “locked” algorithms to perform the same function over and over again. Regulators are hopeful that sponsors can harness machine learning for devices that can adapt to changing conditions.