Doctors have developed an AI tool that can spot signs of heart disease in less than two seconds. The tool reads a standard ECG and pulls out details that are hard for a human to catch.
An ECG records the electrical activity of the heart. It shows the heart's rate and rhythm and has been used for about a hundred years.
On its own, a normal ECG cannot detect heart disease. Doctors usually need an echocardiogram, which is an ultrasound scan of the heart, to confirm it.
The problem is that patients often wait months for an echocardiogram after being referred. That delay can slow down treatment for serious conditions.
How the AI Tool Works
The new AI system was trained on millions of ECGs. It looks for patterns linked to heart failure and heart valve disease, two of the most common types of heart disease.
The findings were shared at the European Society of Cardiology congress in Munich, the largest heart conference in the world.
In a trial of 67,000 patients in the US, the tool identified up to 81% of people with heart failure. It also identified up to 90% of people with heart valve disease.
Dr Sonya Babu-Narayan of the British Heart Foundation said the speed of the read-out felt like the blink of an eye. She said the tool will not catch every case, but it could help doctors fast-track the patients most likely to have a problem.
What This Means for Patients
The AI tool cannot diagnose heart disease by itself. It gives doctors a strong signal that someone may need further testing.
A patient flagged by the tool could be sent for an echocardiogram faster than the usual waiting list allows. That could lead to earlier treatment.
Prof Fu Siong Ng of Imperial College London said patients can wait several months for a heart ultrasound after a referral. He said the tool could help prioritize the people at highest risk.
Ng also said the AI could be used on ECGs taken for unrelated reasons. This means it might catch heart failure or valve disease in people who were not being checked for either condition.
He suggested the model could be run on every ECG done in a hospital. This would flag anyone at high risk of these two conditions for further review.
Dr Ahmed El-Medany, who led the analysis at Imperial College London, called the tool a superhuman AI. He said the next step is building handheld AI-powered ECG readers for health workers.
Other researchers at the Munich conference presented separate AI work using facial videos. Teams from the University of Tokyo and the Institute of Science Tokyo said a five-second facial video could help detect undiagnosed high blood pressure and type 2 diabetes.
Millions of people worldwide have high blood pressure or type 2 diabetes without knowing it. Researchers say tools like these could help identify those cases sooner.
