By Conor Hale, FierceBiotech | April 3, 2019

The FDA has begun to reconsider how it reviews and approves medical devices that employ artificial intelligence and to learn from data to adapt their care.

The agency hopes a tailored regulatory framework will help promote the development of machine-learning devices and programs, which FDA Commissioner Scott Gottlieb says “have the potential to fundamentally transform the delivery of healthcare.”

“As technology and science advance, we can expect to see earlier disease detection, more accurate diagnosis, more targeted therapies and significant improvements in personalized medicine,” Gottlieb said in an agency statement.

Biopharma is a fast-growing world where big ideas come along every day. Our subscribers rely on FierceBiotech as their must-read source for the latest news, analysis and data in the world of biotech and pharma R&D. Sign up today to get biotech news and updates delivered to your inbox and read on the go.

Currently, the FDA employs a risk-based approach to determine whether a new premarket submission is required each time a manufacturer makes substantial, iterative changes through a software update or makes other changes that would significantly affect the device’s performance.

But that’s not ideal when applied to AI and machine learning-based algorithms—medical devices that may continuously update themselves in response to real-world feedback. The agency found it needed to reimagine its approach to foster software that evolves over time to improve care, while still guaranteeing safety and effectiveness.

As a first step, the FDA has released a white paper exploring the proposed framework, which may allow some modifications without review. It also considers the agency’s Software Pre-Cert Program and applying a regulatory approach toward a product’s entire life cycle, including assessments of individual companies’ culture of quality and organization of their software development, testing and safety-monitoring activities.

Future goals include a new draft guidance on the topic, but in the meantime the agency will continue to apply its current authorities, including its regulations governing software as a medical device.

RELATED: FDA approves diabetic retinopathy-detecting AI algorithm

Last year, the FDA approved an AI algorithm for the early detection of diabetic retinopathy, which can lead to loss of vision, from pictures taken of the back of the eye. The system, developed by IDx, aims to help primary care clinics screen their diabetes patients without having to employ a specialist.

The agency also approved AI software from Viz.AI, designed to scan CT images, identify if a patient has suffered a stroke and take steps to alert neurovascular specialists.

RELATED: Stroke detection software developer Viz.ai brings in $21M in series A round

However, the FDA describes both of these devices as “locked” algorithms, or products that don’t continually adapt and are dependent on updates from the manufacturer, which can include training the algorithms with new data to improve their performance. Those changes would require manual validation and verification of the updates.

“But there’s a great deal of promise beyond locked algorithms that’s ripe for application in the health care space, and which requires careful oversight to ensure the benefits of these advanced technologies outweigh the risks to patients,” Gottlieb said.

The agency’s future approach may require scrutinizing manufacturers’ prespecified plans for modifications, including through algorithm retraining and updates, as well as their ability to manage and control the resulting risks.