FDA AI-enabled medical devices, one record each

FDA publishes a list of the AI-enabled devices it has authorized, with the submission number, the company, the lead panel and the product code. A script reads that list, joins each submission to openFDA's bulk 510(k), PMA and classification records, opens the public decision document and copies the Indications for Use statement word for word, and asks MAUDE how many adverse-event reports it holds for that product code. The flags under each record are pattern matches on the Indications text: a modality named, "adjunct", "triage", "not for primary diagnosis". They are facts about the text, not our view of the device.

This is a record, not medical advice, not a safety rating, and not a characterization of any device.MAUDE counts are by product code, which many devices share, and MAUDE reports are unverified by FDA. Read the decision document and the device's labeling before relying on any entry, and tell us at about if a record is wrong.

The reason the record exists: every device here is cleared for a stated use, and every clinical note, claim or report that cites one can be checked against that statement. What Apodicta does on a medical record is exactly that check, line by line, before anything renders.