|
Trends from FDA-Approved AI/ML Devices
(Thursday, August 15, 2024) In the last 10 years, the FDA has approved about a thousand medical devices containing Artificial Intelligence/Machine Learning (AI/ML) algorithms. Although the devices cover diverse indications and types, there are some interesting regulatory trends to be considered for those developing a new AI/ML device. The number of AI/ML devices approved by the FDA has steadily increased over the years with more than 800 such devices cleared by the FDA in the last five years. The indications for these devices are as diverse as the devices themselves. However, the most common indication area for AI/ML devices is radiology with more than 70% (723) devices using radiology-related indications. This is not surprising since the most common application of AI and ML so far has been in image analysis applications which abound in radiology. Cardiovascular indications, with 98 devices, and neurology, with 34 AI/ML devices, account for more than 85% of the AI/ML devices together with radiology. In terms of the regulatory pathway, almost all the AI/ML devices (~98%) were approved under the traditional 510k pathway, with only about 12 Denovo 510ks in the last 5 years. From a strategic point of view, a radiology device has a much better chance of finding multiple predicate devices while a gynecology or physical medicine device would have to live with only one predicate in each category. The reason for radiology being a popular AI/ML target is because of the ease of use of such algorithms for imaging-based applications and vice-versa for those with fewer devices. So, how does this list help developers of new devices? For one, it gives the list of most of the AI/ML devices in one place. This list is based on the device descriptions and labeling used in the their market approval applications. If a device contained AI/ML tools but did not include its description in the device description or summary section of the market approval application, it would not be on this list. But still, this is the best place to start your search for a predicate. Second, this is the best evidence that the traditional 510k pathway is the preferred regulatory pathway for all AI/ML devices. There have been no PMAs on this list in the last 10 years. Third, the number of devices approved by the FDA has exponentially increased in the last 5 years indicating better FDA support for the developers of such devices leading to more applications. This is a free resource available to developers to find other similar devices and get rapid access to the summaries of the 510ks of predicate devices in one place. 
AUTHOR
Dr. Mukesh Kumar Founder & CEO, FDAMap Email: [email protected] Linkedin: Mukesh Kumar, PhD, RAC Instagram: mukeshkumarrac Twitter: @FDA_MAP Youtube: MukeshKumarFDAMap |
|
|
|
