News
Pathway Labs' EchoNext gets FDA clearance for multicondition structural heart disease detection from routine ECGs
The NewYork-Presbyterian / Columbia-developed model reads standard 12-lead ECGs to flag six structural heart conditions — including right- and left-sided heart failure, valve disease, and pulmonary hypertension — that clinicians typically miss until patients are symptomatic. Trained on 700,000+ paired ECG/echo cases and validated across 20+ hospitals and 500,000 patients. Pathway is bundling access through OpenEvidence, which reaches roughly half of US clinicians.
- FDA
- cardiology
- ECG
- screening
- devices
Notable for two reasons. First, it is the first FDA clearance for a multicondition AI in cardiology — six indications on a single 510(k). Most cleared cardiology AI to date has been point-solution: a single arrhythmia, a single valve condition, a single risk score. Bundling six structural-heart conditions on one clearance is a bet that the underlying signal in a routine 12-lead ECG is richer than the point-solution frame has treated it as, and that regulators will accept the multicondition claim when the evidence supports it.
Second, this is an early example of AI screening built into an already-ubiquitous test rather than as a bespoke workflow. The 12-lead ECG runs on every patient who touches an emergency department, most primary-care visits with a cardiac question, and most preoperative workups. A well-behaved AI screen on that test — flagging patients with likely structural heart disease that clinicians typically miss until symptomatic — is the kind of “no new order needed” implementation that has been the theoretical dream of screening AI for a decade. Pathway’s evidence base — 700,000+ paired ECG/echo cases at NewYork-Presbyterian/Columbia, validated across 20+ hospitals and 500,000 patients — is one of the larger cardiology-AI training sets under a single clearance.
The distribution partnership with OpenEvidence — a decision-support platform that reaches roughly half of U.S. clinicians — is the operational question that will decide whether this clears the “cleared but never deployed” hurdle that plagues so many AI-enabled devices. A screening tool only produces clinical value if it gets in front of ordering clinicians in a workflow they already use.
Related coverage: medical imaging AI topic, FDA & devices topic.
Primary source: Read the full original on BusinessWire ↗