JMIR Publications

ABSTRACT

Background:

The US Food and Drug Administration (FDA) has authorized artificial intelligence (AI)– and machine learning (ML)–enabled medical devices since 1995 and maintains a public registry of these authorizations. Prior analyses report that radiology dominates this landscape, but whether that concentration has persisted, intensified, or begun to reverse across three decades, particularly during the rapid growth in approvals since 2022, remains insufficiently characterized.

Objective:

This study aimed to (1) characterize the longitudinal growth of FDA-authorized AI/ML-enabled devices from 1995 to 2025; (2) quantify the temporal evolution of clinical specialty distribution across four eras; (3) identify emerging specialties; and (4) examine the association between manufacturer type and non-radiology authorization.

Methods:

All 1,430 devices listed in the FDA AI-Enabled Medical Devices registry as of December 2025 were analyzed. Devices were stratified by clinical specialty (FDA advisory committee panel) and four eras: Era 1 (1995–2015), Era 2 (2016–2019), Era 3 (2020–2022), and Era 4 (2023–2025). Concentration was quantified using the Herfindahl-Hirschman Index (HHI) with bootstrap confidence intervals (CIs); the Cochran-Armitage test assessed trends in specialty share. Multivariable logistic regression with manufacturer type and era as predictors estimated the odds of non-radiology authorization, with an era-by-manufacturer interaction term. Manufacturers were classified using FDA records, Crunchbase, PitchBook, and company websites.

Results:

Annual authorizations rose from a mean of 2.0 per year in Era 1 to 264 per year in Era 4, with 331 in 2025 alone; the 510(k) pathway accounted for 96.2%. Radiology led in every era but followed a nonmonotonic trajectory, rising from 35.7% (Era 1) to a peak of 85.5% (Era 3) before declining to 77.5% in Era 4, the first significant decline on record (P=.001). The HHI fell from 0.738 (Era 3) to 0.612 (Era 4; bootstrap P<.001), indicating measurable diversification. Specialty distribution was associated with era (χ²48=328.0; P<.001; Cramer V=0.28). Compared with incumbents, startups (odds ratio [OR] 5.09, 95% CI 3.33–7.79) and technology companies (OR 50.62, 95% CI 12.90–198.64) had substantially higher odds of non-radiology authorization, although the era-by-manufacturer interaction was nonsignificant (likelihood ratio test χ²8=11.16; P=.19), indicating a persistent rather than widening effect.

Conclusions:

Although radiology remains dominant, a measurable diversification across specialties began during 2023–2025. This expansion is associated with startup and technology-company activity and coincides with the maturation of clinical data infrastructure beyond imaging, with implications for health-system readiness, workforce training, and specialty-specific regulatory frameworks.