Dallas researchers say a low-cost heart test paired with artificial intelligence may give clinics without high-end imaging a way to catch heart failure earlier, before patients even know something is wrong. A UT Southwestern-led team plugged an AI algorithm into routine 12-lead electrocardiograms and found it could flag left ventricular systolic dysfunction, a common warning sign that heart failure is on the way. The findings land at a time when hospitals and public health programs are scrambling for cheaper ways to triage heart care in places where resources are tight.
The study enrolled nearly 6,000 adults receiving routine care at eight health facilities across Kenya, and 1,444 of them also underwent confirmatory echocardiograms, according to UT Southwestern. Echocardiography, essentially an ultrasound of the heart, is the current gold standard for diagnosing left ventricular dysfunction, but the machines, staff, and maintenance they require are often out of reach in much of sub-Saharan Africa. The project was backed by a research grant from AstraZeneca along with in-kind support from Medical AI and Tricog Health, according to the release notes.
In the cross-sectional analysis published in JAMA Cardiology, the algorithm picked up left ventricular systolic dysfunction in 14.1% of the patients who also had echocardiograms. When matched against the echo results, the AI-ECG delivered a sensitivity of 95.6%, a specificity of 79.4%, and a negative predictive value of 99.1%. In plain English, it almost never missed patients whose hearts were truly struggling to pump. The author list includes clinicians from Kenyan hospitals along with international collaborators, underscoring that this was not a single-center experiment.
How It Could Work Where Echos Are Scarce
Because standard ECG machines are far cheaper and far more common than echocardiography units, the team argues that AI-assisted ECGs could become a first-pass screen. The idea is to quickly rule out people who are unlikely to have the disease and to reserve limited ultrasound slots for those flagged as higher risk. The European Society of Cardiology spotlighted the study’s prospective, multi-clinic design as a meaningful step for implementation in low-resource health systems, according to the European Society of Cardiology. Clinicians involved say a tool like this could cut down on unnecessary referrals while speeding up diagnosis for patients who truly need treatment.
Next Steps And Limits
Lead author Ambarish Pandey, MD, called the algorithm “a practical, scalable screening tool” as teams work out how to fit it into everyday clinic routines, according to UT Southwestern. The next phase, investigators say, will need pragmatic trials and decision-support studies to see whether AI-driven screening actually reduces missed cases, trims wasted echocardiograms, and does all of that without baking in new forms of bias.
An editorial that accompanied the paper in JAMA cautioned against getting too far ahead of the data. Many AI-ECG models, it noted, are trained on populations from high-income countries, and “external validation in low-resource settings ... is generally lacking,” which means any rollout needs careful monitoring and ongoing evaluation. Even with those caveats, researchers frame the Kenyan work as a concrete step toward lowering the barriers to cardiac screening in places where echocardiography remains a rare commodity.
