Heart failure requires close monitoring, especially after a hospital admission. A key indicator is the ejection fraction (LVEF), which is usually measured with echocardiograms. Researchers from MIT and Harvard present PULSE-HF, a model that analyzes a simple electrocardiogram to predict if the LVEF will worsen. This would allow prioritizing the monitoring of higher-risk patients.
Deep learning interprets hidden signals in the ECG 💡
PULSE-HF uses convolutional neural networks trained with thousands of pairs of ECGs and echocardiograms. The model does not diagnose the disease, but identifies subtle patterns in the heart's electrical signal that precede a deterioration in pumping function. By processing a standard ECG, it generates a risk prediction. The approach seeks to be a triage tool, complementing more expensive methods.
Your heart has a history that the ECG doesn't forget 🫀
It seems that the electrocardiogram, that routine test that sometimes feels like a formality, was keeping secrets. While the doctor reviews the tracing, an algorithm might be whispering: this patient will need more attention. It's as if the heart's wiring left an advance error message, a technical premonition that turns a simple exam into a digital crystal ball. The machine no longer just sees the present; now it reads between the lines of the future.