An ECG That Can’t Quite Diagnose You: Why Tempus AI’s Latest Clearance Is Both a Victory and a Trap
(SeaPRwire) – By: Oliver Hawthorne
Pulmonary hypertension is one of those conditions medicine loves to miss. It hides in plain sight, masked by fatigue and shortness of breath that patients and doctors alike dismiss as aging, anxiety, or deconditioning. By the time it’s confirmed through right heart catheterization — the invasive gold standard — the damage is often advanced. The condition affects roughly one percent of the general population, and that figure climbs to ten percent in adults over sixty-five. That’s a massive blind spot. The question Tempus AI is now asking the industry to answer is whether an AI reading a routine ECG can close it.
Tempus received FDA 510(k) clearance for its ECG-PH software. This is the third cardiovascular AI product in Tempus’s portfolio to clear the same regulatory threshold, joining tools already approved for detecting atrial fibrillation and low ejection fraction. The mechanism is straightforward enough. The software takes a standard twelve-lead resting ECG and looks for patterns that suggest elevated mean pulmonary artery pressure. It outputs a binary flag. That’s it. It does not diagnose. It does not replace right heart catheterization. It only works for patients aged forty and older who are presenting with cardiovascular symptoms such as dyspnea, fatigue, chest pain, or edema and have no prior PH diagnosis. It cannot be used for serial monitoring. It will not process ECGs from patients with paced rhythms. The clearance is limited to clinical diagnostic ECG recordings at healthcare facilities. None of this is an exaggeration or an overreach. The press release and regulatory pathway make every one of these boundaries explicit.
The subtext is where things get interesting. Tempus holds one of the largest multimodal health data libraries in the world. They have been methodically building a cardiovascular AI suite that uses the same input modality across three distinct conditions. A standard ECG. Three separate detection algorithms. The same clinical setting. This is not accidental. Brandon Fornwalt, Tempus’s Senior Vice President of Cardiology, framed the clearance as part of a broader push to surface cardiovascular risk earlier. The strategy is to make the ECG a universal screening gateway inside the hospital. If Tempus can prove that a single AI-native workflow can flag AFib, low ejection fraction, and now PH from the same tracing, they own a powerful positioning argument. But the binary output creates a real tension. A flag is not a diagnosis. Clinicians are expected to use the result alongside the full clinical picture. That means the tool will sit in a gray zone — useful enough to keep ordering, restrictive enough to never replace the existing diagnostic pathway. The 510(k) pathway itself confirms this positioning. It requires substantial equivalence to an already marketed device. It does not certify clinical superiority.
The commercial loop is where this story lands. Tempus is not selling a diagnostic replacement. They are selling a risk-surfacing layer that plugs into an existing clinical workflow. The value proposition is early identification, not definitive diagnosis. For hospitals, that means a tool that can flag suspected PH cases for further workup without displacing cardiologists or catheterization labs. For Tempus, it means recurring revenue from a software layer inside an established examination pathway. The endgame for the industry is clearer than the press release lets on. Tempus is building the AI infrastructure around the ECG — the most common, most routine, most underutilized diagnostic test in modern medicine. Once they control that workflow across multiple conditions, they control a significant portion of early cardiovascular risk detection. The question is not whether this model works. It is whether the binary, non-diagnostic nature of the tool limits adoption before the moat becomes defensible.
Author bio: Oliver Hawthorne is a Principal Correspondent permanently stationed at an international technology review, covering the intersection of AI, healthcare, and regulatory policy.