'Superhuman' AI Reads ECGs to Flag Heart Disease in Seconds

Imperial College London's AI spots heart failure and valve disease from routine ECGs in under two seconds, trialling in NHS hospitals to cut long scan waits.

Doctors have developed a "superhuman" AI tool that can extract signs of heart disease from a routine electrocardiogram in less than two seconds. Standard ECGs, used for a century, cannot detect heart disease — that normally requires an echocardiogram ultrasound patients often wait months for. The results were presented to delegates at the European Society of Cardiology annual congress in Munich.

The tool is being described by its developers as "superhuman" because it detects disease signals hidden within a routine test performed about a billion times a year worldwide. In a US trial of 67,000 patients, it identified up to 81% of heart failure cases and up to 90% of heart valve disease cases, according to The Guardian. The Sun, which also covered the congress, reports it spotted 86% of reduced heart pumping function cases and up to 90% accuracy for aortic stenosis.

Nearly 400,000 people are on NHS waiting lists for heart tests, and The Sun reports the technology will be trialled in NHS hospitals in London and Bristol to filter high- and low-risk patients. Early diagnosis is vital for heart failure and valve disease, letting patients start lifesaving medicines before they become dangerously unwell. The tool cannot definitively diagnose or rule out conditions on its own, but it could fast-track those most likely to have abnormalities for echocardiograms instead of waiting months. Professor Fu Siong Ng of Imperial College told The Guardian it could also opportunistically flag uns suspected cases by running on all hospital ECGs.

Dr Ahmed El-Medany said the next challenge is designing handheld AI-led ECG readers for healthcare professionals, according to The Guardian. Dr Roy Jogiya of Heart Research UK told The Sun the real potential is not replacing cardiologists but acting as "an incredibly rapid extra pair of eyes." The Munich congress also showcased related advances: researchers at the University of Tokyo and the Institute of Science Tokyo presented AI analysis of five-second facial videos that could detect undiagnosed high blood pressure and type 2 diabetes, with The Sun reporting up to 90% accuracy for blood pressure detection.