Radiology Claims 76% Of FDA-Cleared Medical AI, Study Finds

Medical artificial intelligence cleared by US regulators remains heavily clustered in one specialty, according to findings announced on October 6, 2026 by GigHz, a Los Angeles software and research firm started by a physician.

The analysis, published in Cureus on July 13, 2026 and written by GigHz founder Pouyan Golshani and Mary S. Joseph, reviewed 1,430 entries on the FDA's public list of AI- and machine-learning-enabled devices authorized between September 1995 and December 2025. Radiology review accounted for 1,094 of those records, or 76.5%. Adding the cardiovascular and neurology panels brings the share to 90.6%. Pathology had nine records, microbiology six and obstetrics and gynecology four. No device was authorized under a psychiatry or behavioral health panel.

The pace has accelerated sharply: an average of 1.8 authorizations a year from 1995 through 2014 versus 264 a year from 2023 through 2025, including 331 in 2025.

The market is fragmented. Of 740 developers, 502 had just one authorized device, while 13 firms held 247 devices, or 17.3% of the total.

Golshani attributed radiology's lead to its existing digital imaging infrastructure, cautioning that the data measures authorizations, not clinical use or patient benefit. "Delaying something useful has a cost, too," he said, according to the company's announcement distributed by PR Newswire.

This story was first reported by PR Newswire.