As of 2025, radiology accounted for 76% of all AI or machine learning devices (ML) cleared by the U.S. Food and Drug Administration (FDA), according to a recent report.
The finding is from a longitudinal analysis of 1,430 AI/ML-enabled devices authorized by the FDA between September 1995 and December 2025 and confirms that the rapid growth in AI has been concentrated in image-rich diagnostic specialties, noted lead author Pouyan Golshani, MD, an interventional radiologist with Kaiser Permanente Los Angeles Medical Center.
“The cumulative regulatory record demonstrates rapid growth that has been concentrated in image-rich diagnostic specialties, with limited representation across many specialties that account for substantial clinical activity in the United States,” the authors wrote. The study was published July 13 in Cureus.
The FDA maintains a public list of AI-enabled medical devices, which it updates periodically and which has been the basis of multiple published analyses characterizing the regulatory landscape. No published analysis has characterized the full cumulative dataset through the end of calendar year 2025, however, a period during which authorization volume continued to accelerate, the authors noted.
To provide an update, the researchers classified every device on the FDA’s public AI/ML list by regulatory pathway, lead review panel, and manufacturer, using each submission number to identify how it was cleared. The vast majority of devices, 1,376 of 1,430, cleared through the streamlined 510(k) process rather than the more rigorous De Novo or premarket approval pathways.
Annual and cumulative FDA authorizations of AI/ML-enabled medical devices, 1995–2025. CureusAccording to the results, the annual authorization volume increased from a mean of 1.8 per year between 1995 and 2014 to 264 per year between 2023 and 2025, with 331 authorizations recorded in 2025 alone. Devices reviewed by the FDA’s Radiology panel accounted for 1,094 of 1,430 authorizations (76.5%), and the three most represented panels (radiology, cardiovascular, and neurology) accounted for 90.6% of all authorizations.
Several large clinical specialties were represented by very small numbers of authorized devices, including pathology (n = 9, 0.6%), microbiology (n = 6, 0.4%), and obstetrics and gynecology (n = 4, 0.3%). No authorizations were recorded under a psychiatry or behavioral health review panel, the authors noted.
Lastly, of 740 unique companies, 502 (67.8%) had a single authorized device, while 13 (1.8%) companies accounted for 247 (17.3%) devices.
“Our analysis extends prior published reports of this regulatory record by approximately two additional years and approximately 414 additional authorizations relative to the most recent published taxonomic analysis,” the group wrote.
The authors concluded that future work could examine why AI development has clustered so heavily around imaging and a small group of repeat manufacturers, and what would be needed to broaden AI’s reach into underrepresented specialties.
The full study can be found here.
Leave a Reply