In September 2021, FDA’s Digital Health Center of Excellence released a list of artificial intelligence/machine learning-based software as a medical device (“AI/ML-based SaMD”) that are marketed legally in the United States through the 510(k) process, De Novo authorization, or premarket approval (PMA). This list is not exhaustive. However, it does showcase FDA’s increased efforts to provide guidance on this burgeoning area of technology. This is especially noteworthy in light of FDA’s AI/ML-based SaMD action plan that was released in January 2021. This list does not limit itself to AI/ML-based SaMD. It encompasses all medical devices that are enabled by artificial intelligence and machine learning in some material way.
The list includes information such as the submission number, the device and company name, and the date of marketing authorization. It also notes the primary product code and the lead panel associated with the device, such as radiology, cardiovascular, gastroenterology-urology, etc. There is a link provided for each device which allows the user to access the FDA medical device database, wherein the marketing authorization is located. So far, the database contains a listing of 343 entries, and it illustrates the diversity of medical fields that have been incorporating artificial intelligence and machine learning into medical devices.
The majority of approved devices are diagnostic and monitoring medical devices. This list is very important for medical device manufacturers looking to submit marketing authorization requests for medical devices that incorporate AI/ML. It is also key for those in the design process to see the types of medical devices that have obtained marketing authorization successfully. Although AI/ML medical devices for diagnosis and management of illness have been most successful at gaining marketing authorization, this list will also be helpful for those devices that are made for treatment of illnesses using AI/ML. The primary concern for these devices is the issue of transparency with regards to software programs. This newly released list should be helpful in guiding medical device manufacturers on how to create systems that allow for greater reporting. It will be interesting to see how these medical devices accumulate real world data and how that will translate into increased precision and accuracy of diagnosis.