On April 11, 2018, the U.S. Food and Drug Administration granted a De Novo classification request for IDx-DR, an AI software device that analyzes images of the back of the eye to detect diabetic retinopathy. The FDA’s grant order, document DEN180001, states that “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.” What made the authorization a landmark was the word “automatically”: IDx-DR returns a screening result on its own, without requiring a clinician to also read the image. It was the first autonomous AI diagnostic system the FDA authorized for the U.S. market.
The De Novo pathway the FDA used is itself part of the story. Because there was no previously cleared device of this type to compare IDx-DR against, the agency created a new device category - “retinal diagnostic software device,” defined in the order as “a prescription software device that incorporates an adaptive algorithm to evaluate ophthalmic images for diagnostic screening to identify retinal diseases or conditions” - and classified it as Class II with special controls. The order’s risk table lists exactly the failure modes that matter for a system making a medical call on its own: false positives leading to unnecessary procedures, false negatives delaying treatment, and the software or algorithm simply failing - each paired with required mitigations such as clinical performance testing and labeling. The device was authorized for use only with a specific retinal camera and only for the narrow task of diabetic retinopathy screening; the FDA was explicit that it should not be used to detect any other condition.
This entry is deliberately the success-side counterweight to this library’s account of Watson Health. Watson won a quiz show and was marketed as the future of medicine, then was sold for parts when the clinical results never matched the hype. IDx-DR went the opposite way: a narrowly scoped, rigorously validated system that did one well-defined diagnostic job and earned regulatory authorization to do it autonomously. The contrast is the lesson. The path from impressive demonstration to deployable medical AI runs through narrow scope, real clinical validation, and a regulator’s risk analysis - not through broad ambition. IDx-DR mattered because it showed that path could actually be walked, and it opened the door for the autonomous diagnostic AI that followed.