The US Food and Drug Administration (FDA) has proposed a new regulatory framework for artificial intelligent medical devices.
To ensure that the machines only work for their approved purpose, FDA-cleared algorithms are currently locked to prevent the system from continuing to adapt and learn each time the code is used. The algorithm is instead trained manually.
Suggested framework changes will allow these continuously learning algorithms to be trained using real-world user data while also maintaining the safety and efficacy of the device.
FDA commissioner Scott Gottlieb said: “With AI, because the device evolves based on what it learns while it is in real-world use, we’re working to develop an appropriate framework that allows the software to evolve in ways to improve its performance while ensuring that changes meet our gold standard for safety and effectiveness throughout the product’s lifecycle-from premarket design throughout the device’s use on the market.”
The regulatory agency has published a discussion paper outlining the new framework that considers an algorithm’s performance, the manufacturer’s modification plans and the maker’s ability to manage and control modification risks.
The FDA intends to publish draft guidance based on the feedback received from the discussion paper.
AI algorithms are currently being used to screen diseases and offer treatment recommendations.
In April last year, the FDA cleared IDx’s AI-based medical device to detect diabetic retinopathy.
The agency also granted marketing authorisation to VizAI’s clinical decision support software in February last year.