FDA grants De Novo authorization to an AI model that reads ECGs for hidden heart attacks

The US Food and Drug Administration granted De Novo authorization, decision number DEN250044, to Powerful Medical Inc.’s “STEMI AI ECG Model” on September 3, 2026, according to the FDA’s own De Novo classification database. The device, marketed as Queen of Hearts, is an AI model that analyzes a patient’s electrocardiogram to identify ST-elevation myocardial infarction and its equivalents - the electrical signatures of a fully blocked coronary artery, the kind of heart attack that demands the fastest possible path to catheterization. The FDA’s record lists the request as received September 22, 2025, reviewed under the cardiovascular advisory committee, and the outcome as granted.

De Novo is the pathway the FDA uses for novel devices that have no prior cleared equivalent to piggyback on, so it demands more evidence than the 510(k) route that most AI-enabled devices take. Two details in the record are worth more than the headline. The product code is new (SHS), meaning the FDA created a classification rather than fitting the device into an existing one. And the record states that a Predetermined Change Control Plan is authorized, which is the mechanism letting the manufacturer ship defined model updates under a pre-agreed plan instead of returning for a new submission each time - the FDA’s answer to the problem that a cleared AI model is a moving target. Powerful Medical had arrived at this point already carrying FDA Breakthrough Device Designation and a Class IIb CE mark under EU MDR, and states on its own site that the technology has been used by more than 100,000 clinicians across more than 2.5 million ECGs and 150 hospital deployments outside the US.

What the De Novo grant changes in practice is legal marketing status inside the United States: an AI reading of an ECG that flags a STEMI-equivalent can now be sold and used as a cleared triage aid rather than an investigational tool, in an emergency-medicine context where minutes between an ECG and a catheterization lab decide how much heart muscle a patient keeps. The device’s target is specifically the subset of heart attacks that traditional STEMI criteria on a standard 12-lead reading can miss, which is precisely where a pattern-recognition model has something to add over a rule threshold.

The clearance is worth watching less for STEMI detection in isolation and more as a marker of where the De Novo pathway is willing to go for pattern-recognition AI in acute care: a new classification, awarded to a device making a time-critical triage call rather than a slower diagnostic read, is a different regulatory bar than the radiology and imaging AI that dominates the FDA’s AI device list. It clears one product for the US market; it does not, on its own, establish how emergency departments will integrate an AI triage flag into existing STEMI activation protocols, or what liability looks like when the model and a clinician disagree.