Diabetic Retinopathy Detection AI
Screens photographs of the back of the eye for diabetic damage, in some cases giving the result without a specialist reading it.
This is where autonomous AI genuinely arrived in American medicine, and it is worth understanding why it happened here first.
Diabetes damages the small blood vessels at the back of the eye. Caught early it is treatable. Caught late it blinds people. Everyone with diabetes is supposed to be screened every year, and huge numbers are not, because it means another appointment with another specialist.
The screening photograph, though, is standard and easy to take. So the question became whether software could read it without a specialist at all.
Some products in this group are cleared to do exactly that: a camera in a primary care clinic takes the picture, the software returns a result, and the patient is either reassured or referred, in one visit. No eye doctor in the loop for a normal result.
That is a real transfer of a clinical decision to software, done in the one place where the trade was clearly worth it.
Clearances per Year
What the FDA Actually Signed Off On
- FDA product code
- PIB
- FDA name for it
- Diabetic Retinopathy Detection Device
- Rule number
- 21 CFR 886.1100
- Risk class
- Class 2
- Medical field
- Ophthalmic
Before You Sign Anything
- Ask whether the clearance is autonomous or assistive. They are different products and different liability conversations.
- Ask what share of images the software refuses to grade. Ungradable images have to go somewhere, and that referral load is real.
- Ask which cameras it is cleared with. These clearances are usually tied to specific hardware.
- Ask how referrals are tracked. Screening only helps if the people who need care actually get there.
Who Holds the Most Clearances Here
| Company | Cleared products |
|---|---|
| Aeye Health, Inc. | 2 |
| Eyenuk, Inc. | 2 |
| Digital Diagnostics, Inc. | 2 |
| iHealthscreen, Inc. | 1 |
| Idx, LLC | 1 |
Every Cleared Product, Newest First
Each row links to the FDA's own record. The number in the last column is the clearance number, which is the single best thing to ask a vendor for.
| Product | Company | Cleared | FDA record |
|---|---|---|---|
| iPredict-DR | iHealthscreen, Inc. | July 2, 2026 | K253704 |
| AEYE-DS | Aeye Health, Inc. | April 23, 2024 | K240058 |
| EyeArt v2.2.0 | Eyenuk, Inc. | June 16, 2023 | K223357 |
| AEYE-DS | Aeye Health, Inc. | November 10, 2022 | K221183 |
| IDx-DR v2.3 | Digital Diagnostics, Inc. | June 17, 2022 | K213037 |
| IDx-DR | Digital Diagnostics, Inc. | June 10, 2021 | K203629 |
| EyeArt | Eyenuk, Inc. | August 3, 2020 | K200667 |
| IDx-DR | Idx, LLC | April 11, 2018 | DEN180001 |
Clearance data comes from the FDA's public 510(k) and device classification records through openFDA. A clearance means the FDA agreed the product is close enough to something already on the market, not that the FDA tested it in your hospital. It says nothing about accuracy in your patient mix, your scanners, or your workflow. Confirm the current status and the exact cleared indication on the FDA record before you buy.