OBJECTIVES: The U.S. Food and Drug Administration (USFDA) is increasingly accepting artificial intelligence and machine learning (AI/ML)-enabled medical devices in healthcare. It is important to understand the regulatory requirements for approval and oversight of these devices. To characterize the AI/ML-enabled medical devices, we conducted a comprehensive analysis of devices approved by the USFDA.

METHODS: AI/ML-enabled medical devices approved by the USFDA from 1995 to July 2023 were extracted using the summary data from the USFDA electronic database. Data pertaining to the date of approval, indication type, approval type, device class, intended use, and type of algorithm employed was collected and analyzed using descriptive statistics.

RESULTS: A total of 692 AI/ML-enabled medical devices were approved till July 2023. The approvals increased from one in 1995 to 108 by July 2023. Most of these devices were employed in the field of radiology (76.7%) and cardiovascular (10.3%). Majority of the devices were approved under the 510(k) premarket notification pathway (96.5%). Almost all the devices (99.6%) were classified as class II medical devices and were intended for prescription use only (97.8%). Out of 692 devices, only 14% of medical devices had mobile applications operating on compatible smartphones. The most common applications were image reconstruction (63.4%) and diagnostic assistance (31.8%). The primary algorithm employed for these devices was deep learning (36.4%), of which convolutional neural network was most commonly used (26.2%).

CONCLUSIONS: Since their inception, the rapid growth of AI/ML-enabled medical devices can be attributed to their transformative potential in healthcare, markedly enhancing diagnostic accuracy and patient outcomes. Our analysis showed that most of these were class II devices approved under 510(k) pathway, with predominant application in radiology to assist image reconstruction. Notably, deep learning algorithms were commonly deployed in these devices due to their robust adaptability for advancing diagnostic capabilities.