The signal
On 11 April 2018, the FDA granted a De Novo classification request for IDx-DR, a healthcare software device, creating a new regulation, 21 CFR 886.1100, and product code for what FDA calls a “retinal diagnostic software device.” The De Novo summary for DEN180001 states IDx-DR “is indicated for use by health care providers to automatically detect more than mild diabetic retinopathy (mtmDR) in adults diagnosed with diabetes who have not been previously diagnosed with diabetic retinopathy,” using images from a specified Topcon camera. FDA's own De Novo database entry independently records the same 04/11/2018 decision date, applicant IDx, LLC, and a granted status.
The evidence
The De Novo pathway itself is informative: FDA uses it only when no legally marketed predicate device exists for comparison, so the grant is documentary evidence of a novel device category, not merely a press description calling it “first.” The clinical study behind the grant found that of 852 eligible participants, 819, or 96.1%, produced images of sufficient quality for the software to generate a result, and the summary states the system “met predetermined sensitivity and specificity standards for the autonomous detection of more than mild DR or DME in people with diabetes but no history of DR in primary care settings.” The word “autonomous” in FDA's own summary refers specifically to that tested claim: a screening output generated without a specialist interpreting the retinal image first, not a general claim about unsupervised diagnosis.
Timeframe and confidence
Confidence in the classification date and the tested indication is high, since both come directly from FDA's own summary and its device database. What the grant does not establish, an editorial distinction, is how far “autonomous” extends: the summary's own warnings state IDx-DR is not intended to detect other eye diseases such as glaucoma, does not treat retinopathy, and that a patient for whom it cannot generate a result may still need referral for vision-threatening disease. The autonomy applies to one narrow screening output inside a bounded clinical pathway, not to diagnosis in general.
What would change the reading
A second De Novo or 510(k) grant for a comparably autonomous device in a different specialty, opened and cited directly, would show the regulatory pathway generalising rather than remaining a single case. A subsequent FDA safety communication narrowing IDx-DR's authorised use, or a competitor device cleared under the same 886.1100 regulation, would each update how durable this milestone looks.
- Does “autonomous” in a regulatory clearance mean no clinician reviews the image, no clinician reviews the output, or both?
- What performance margin did the pivotal study report above FDA's specified sensitivity and specificity thresholds?
- Has regulation 21 CFR 886.1100 since been used to clear a device from a different manufacturer, and on what evidence?
The 2018 grant is best read narrowly: FDA created a category for one scoped autonomous task inside a defined clinical pathway, not a general finding about AI diagnosis, and its wider significance depends on whether later De Novo and 510(k) decisions build on the same pathway.
Source trail
- De Novo Classification Request for IDx-DR (DEN180001)www.accessdata.fda.gov · Source publication: 2018-04-11 · Retrieved 2026-09-16
FDA's own summary of the De Novo grant, the device's indication for autonomous detection, and the pivotal study's results.
- FDA De Novo Classification Database entry (DEN180001)www.accessdata.fda.gov · Source publication: not established · Retrieved 2026-09-16
FDA's database record confirming the 04/11/2018 decision date, applicant and granted status for DEN180001, as retrieved 16 September 2026.
- Event date
- 2018-04-11
- First source date
- 2018-04-11
- Source-record publication
- Not supplied — draft retained
- Preparation
- 2026-09-16