Guides · Guide

FDA AI Medical Devices: Five Datasets, 1995–2026

Five downloadable datasets from the FDA's AI-enabled device list: authorizations by year, panel, pathway and company, plus pathology. Limits stated.

We may earn a commission if you buy through our links. It never changes our scores. How we stay independent

The FDA publishes a list of medical devices it has authorized that use artificial intelligence. We downloaded it and turned it into five small datasets you can read here and download as CSV files. Nothing in them is estimated: every figure is a count of rows in the FDA file, or a percentage worked out from those counts.

Status: data from the FDA list, reviewed by Dr. Gupta. The file we used has 1,614 devices with decision dates from 29 September 1995 to 29 June 2026. We downloaded it on 17 October 2026. A later download will have more rows, and totals quoted elsewhere will differ, because they come from a different source and cut-off date.

How to read these numbers

  • The list is not complete. The FDA says it is not a comprehensive resource. Devices are included mainly because AI-related terms appear in their summary or classification. Counts here are counts of that list, not of every AI device on the US market.
  • An authorization is not a performance result. It says the FDA decided a device met a standard for a stated use. It does not show the tool will help your patients or save you time.
  • The file has no description of what each tool does. We cannot split devices into triage, detection or reporting, so we do not.
  • The last year is partial. 2026 runs to 29 June only.

Dataset 1: authorizations per year

The yearly count grew from 18 in 2016 to 335 in 2025. In 2025 the count was 43% higher than in 2024 (235). 42 devices on the list were decided before 2016. Download: united-states-fda-ai-devices-per-year.csv (all years from 1995).

YearDevices
201618
201727
201865
201980
2020114
2021129
2022162
2023226
2024235
2025335
2026 (to 29 June)181

Dataset 2: by FDA review panel

The FDA assigns each device a lead review panel. These are FDA panels, not clinical specialties, so a device can be listed under a panel that differs from the specialty that uses it. Radiology is 76.2% of the list. Download: united-states-fda-ai-devices-by-review-panel.csv (all 18 panels; the table shows the top eight).

FDA review panelDevicesShare
Radiology123076.2%
Cardiovascular1549.5%
Neurology734.5%
Anesthesiology301.9%
Gastroenterology-Urology271.7%
Hematology221.4%
Pathology100.6%
Ophthalmic100.6%

Dataset 3: authorization pathway by year

The submission number shows the route. K numbers are 510(k) clearances, D numbers (DEN) are De Novo authorizations and P numbers are premarket approvals (PMA). Of 1,614 devices, 1,553 are 510(k), 40 are De Novo and 21 are PMA. Download: united-states-fda-ai-devices-pathway-by-year.csv (all years).

Year510(k)De NovoPMA
20185870
20198000
202011310
202112261
202216011
202321961
202422591
202532924
202617641

Dataset 4: companies with the most authorizations

Companies are shown exactly as the FDA lists them, so one company can appear under several names (for example separate lines for Siemens Medical Solutions USA and Siemens Healthcare GmbH, or for different GE and Philips entities). We did not merge them. Only companies with eight or more devices are shown. Download: united-states-fda-ai-devices-top-companies.csv.

Company (as listed by FDA)Devices
Siemens Medical Solutions USA, Inc.56
Canon Medical Systems Corporation42
Aidoc Medical, Ltd.30
Shanghai United Imaging Healthcare Co., Ltd.29
Philips Ultrasound LLC14
Siemens Healthcare GmbH13
GE Medical Systems Ultrasound and Primary Care Diagnostics13
Samsung Medison Co., Ltd.13
GE Medical Systems, LLC13
Clarius Mobile Health Corp.12
Hyperfine, Inc.12
GE Medical Systems SCS12
Viz.ai, Inc.10
Subtle Medical, Inc.9
Brainlab AG9
GE Medical Systems Ultrasound and Primary Care Diagnostics,9
Brainomix Limited8
Philips Medical Systems Nederland B.V.8
Qure.ai Technologies8

Dataset 5: the FDA's "Pathology" panel

Only 10 devices have Pathology as the lead panel. This is a count of a panel label, and the list includes products that are not whole-slide AI: two 1990s automated Pap smear screening systems and genomic tests. Read the device names, not just the count. The Paige Prostate decision date matches the FDA's own announcement of 21 September 2021. Download: united-states-fda-ai-devices-pathology.csv.

Decision dateSubmissionDeviceCompanyPathway
09/29/1995P950009AUTOPAP(R) 300 QC AUTOMATIC PAP SCREENER/QC SYSTEMBD DIAGNOSTICSPMA
11/08/1995P940029PAPNET Testing SystemNeuromedical Systems, Inc.PMA
09/21/2021DEN200080Paige ProstatePaige.AIDe Novo
04/28/2023P210011xT CDxTempus Labs, Inc.PMA
01/31/2024DEN210035“Genius™ Digital Diagnostics System with the Genius™ Cervical AI algorithm”Hologic Inc.De Novo
01/24/2025K241232Galen™ Second Read™Ibex Medical Analytics Ltd.510(k)
07/31/2025DEN240068ArteraAI ProstateArtera Inc.De Novo
08/29/2025K250003GENESEEQPRIME NGS Tumor Profiling Assay (FFPE) (GS6005)Geneseeq Technology Inc.510(k)
12/09/2025K243449INFINITT DPSInfinitt Healthcare Co., Ltd.510(k)
05/04/2026K254115ArteraAI BreastArtera, Inc.510(k)

Pathology AI devices can also sit under other panels, such as Radiology or Clinical Chemistry, so this is not a full list of pathology AI.

Using this when you evaluate a tool

Look up the exact submission number on the FDA's own database and read the stated indication before you buy. Then ask the vendor for performance on patients like yours. The clinician stays responsible for every result.

Dr. Gupta's note. I have read this page. It counts entries on the FDA's list of AI-enabled devices. It does not show how well any device works. FDA authorisation means a device met the agency's requirements for sale. It is not proof that it improves patient care. The review panel is the FDA's own sorting, not a clinical specialty, so please read those counts as a guide only. Use this page to see where approvals are going, and check a specific product's evidence before you rely on it.

Outside the US

Other regulators publish different lists. Australia's TGA lists AI-enabled devices in its ARTG register and says the list does not capture every such device. Health Canada's licence listing (MDALL) covers Class II, III and IV devices and has no AI flag that we found. In the UK, the MHRA's registration database (PARD) searches by manufacturer and device type. An earlier third-party analysis by Hardian Health said AI devices cannot be identified there. The MHRA has since launched a new PARD site, and we have not checked whether it adds an AI filter. We plan separate datasets where the data supports them.

Sources

Why you can trust this

Reviewed by Dr. Ravi Shankar Gupta

DMRE, DCP, FAGS · Radiologist & Pathologist · 50+ years in practice. Every review is checked by him before it is published. See how we review

The monthly briefing

One email. The AI tools that actually matter.

New reviews, price changes and the one tool worth trying this month. Read in two minutes.

By subscribing you agree to receive our monthly briefing and to our privacy policy. Unsubscribe in one click. For doctors in the US, UK, Canada & Australia.