“It is impracticable to halt the exponential advancement of technology.”

Ben Goertzel, artificial intelligence researcher

 

There is an exponential increase in the number and interest in artificial intelligence-enabled medical devices. It is not always clear, however, how these devices have been approved both in the US and Europe.

These authors from the University of Zurich, Switzerland, delineate the AI-enabled devices and their approval profiles in both the US and Europe in this exhaustively-researched manuscript. This is a very timely article given the very recent Google announcement of its AI-enabled Dermatology Assist tool

The number of such devices may be surprising to some with 222 and 240 such AI-enabled devices in the US and Europe, respectively (in the US, there were 9 approvals in 2015 but on a trajectory to be close to 100 in 2019). Most of these devices are in the domain of radiology (129/222 = 58%) as well as cardiology and neurology; very few considered high-risk devices.

The majority of these devices were approved first in Europe, possibly due to its less rigorous process of evaluation of medical devices. There is no centralized agency in Europe such as that of the FDA in the US, which had announced a recent review of the regulatory process for these AI-based medical devices. Europe has accredited private Notified Bodies for the higher risk devices while the lower risk devices are left up to the manufacturer for self-responsibility to assure safety.

While the FDA has regulatory processes such as premarket approval, 510(k) pathway, and de-novo premarket review for these devices, Europe seems to have less regulatory processes that are specific for AI-enabled devices. Such a comprehensive and transparent regulatory process is needed for AI-based medical devices in both US and Europe.

Such a regulatory pathway will then gain the trust of the public and clinicians as well as promote safety and quality necessary for applications. In addition, a publicly accessible database of such devices for both CE-marked as well as FDA-approved devices that utilize AI is also needed.

The full article can be read here