FDA-Cleared AI/ML Medical Device Software: Market Landscape
The FDA has authorized over 1,250 AI/ML-enabled medical devices for marketing in the United States as of 2025, with roughly 90% of all authorizations occurring since 2018. The 510(k) pathway accounts for the vast majority of clearances, with De Novo and PMA routes reserved for higher-risk or novel devices.
Specialty Distribution
| Clinical Specialty | Share of Clearances | Key Applications |
|---|---|---|
| Radiology | ~76% | CT triage, chest X-ray screening, mammography CAD |
| Cardiology | ~10% | ECG analysis, echocardiography guidance, plaque quantification |
| Neurology | ~5% | Stroke detection, EEG analysis, brain bleed quantification |
| Other (Ophthal., Pathology, etc.) | ~9% | Diabetic retinopathy screening, digital pathology |
Regulatory Pathways
Of the AI/ML devices authorized by the FDA, approximately 97% used the 510(k) pathway, with 22 De Novo authorizations and 4 PMA approvals. The median review time for 510(k) submissions in 2025 was 142 days, with one-quarter of devices cleared in under 90 days.
Evolving Regulatory Framework
In January 2025, the FDA published draft guidance on Artificial Intelligence-Enabled Device Software Functions, outlining lifecycle management and marketing submission recommendations. This framework addresses a key industry challenge: how to regulate algorithms that continuously learn and adapt post-deployment. Manufacturers must now consider a Predetermined Change Control Plan (PCCP) to document anticipated modifications to the AI/ML algorithm without requiring new submissions for each update.
Market Implications for Health Systems
For hospital CIOs evaluating AI software procurement, the 510(k) clearance status serves as a regulatory baseline but does not indicate clinical efficacy. Key considerations include:
- Integration capability
- DICOM/HL7 FHIR compatibility, PACS integration, and EHR interoperability
- Clinical validation
- Published peer-reviewed studies beyond the FDA submission data
- Vendor concentration risk
- A small number of vendors (GE HealthCare, Siemens Healthineers, Aidoc, Viz.ai) hold disproportionate shares of clearances