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

Imperial College London's AI flagged heart failure and heart valve disease from routine ECGs in under two seconds, detecting up to 90% of valve cases.

@sharedot

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

By @sharedot · · 8 pages

  • AI
  • Health Tech
  • Heart Disease
  • Medical AI

Imperial College London's AI flagged heart failure and heart valve disease from routine ECGs in under two seconds, detecting up to 90% of valve cases.

What happened at the Munich cardiology congress

Doctors at Imperial College London unveiled an AI tool that can extract signs of heart failure and heart valve disease from a routine electrocardiogram (ECG) in less than two seconds.

Details were presented at the European Society of Cardiology annual congress in Munich, the world's largest heart conference, where Dr Ahmed El-Medany, a British Heart Foundation clinical research fellow who led the analysis, described the technology as a 'superhuman AI'.

The model was trained on millions of patient ECG recordings, and The Guardian reports the trial it was tested in was funded by the British Heart Foundation.

Why this result is surprising

The ECG has been a vital diagnostic tool for a century, recording the heart's electrical rate and rhythm, but it cannot by itself detect structural heart disease — that normally requires an echocardiogram, an ultrasound scan patients can wait months for.

The AI works by extracting patterns from the electrical signal that are too subtle or complex for the human eye to see, turning one of medicine's oldest and cheapest tests into a powerful early-warning system.

Mugglehead Investment Magazine notes the result is the latest chapter in a longer Imperial programme rather than a sudden breakthrough.

The evidence from 67,000 patients

In a trial involving about 67,000 patients in the United States, the AI tool identified up to 81% of those who had heart failure and up to 90% of those with heart valve disease, according to The Guardian and Health and Me.

Health and Me explains that the model is not simply reading the obvious peaks and dips on the trace but hunting for hidden structural signals buried in the electrical pattern.

The tool cannot definitively diagnose or rule out either condition on its own, but it gives a strong indication that a patient needs urgent follow-up.

The stakes for patients and hospitals

Early diagnosis is vital for heart failure and heart valve disease, allowing patients to start lifesaving medicines before they become dangerously unwell.

Because around a billion ECGs are performed worldwide each year, the tool could fast-track high-risk patients to echocardiograms instead of months-long waits.

Prof Fu Siong Ng of Imperial College London said patients can often wait several months for a heart ultrasound scan after referral, so prioritising those most at risk could speed diagnosis and treatment.

Dr Sonya Babu-Narayan, clinical director of the British Heart Foundation, cautioned that the technology will not detect everyone with a heart condition.

Opportunistic screening beyond suspected cases

A further application could prove lifesaving: running the model automatically on every ECG performed in a hospital to flag people whose heart problems are not yet suspected, including patients whose scans were ordered for unrelated reasons such as chest pain or pre-operative checks.

Mugglehead Investment Magazine reports that the same Imperial group has previously shown AI can read ECGs to estimate the risk of future valve leaks, complete heart block and early death, work spun out into a company called cardiovolt.ai.

What comes next

Dr El-Medany said the next challenge is designing handheld AI-led ECG readers that healthcare professionals can carry, and Mugglehead reports Imperial is already running a smaller prospective NHS study in London and Bristol to test the technology in real-world settings.

The Munich congress also heard that AI analysis of five-second facial videos from University of Tokyo and Institute of Science Tokyo researchers could detect undiagnosed high blood pressure and type 2 diabetes, underscoring how quickly AI screening is spreading across routine medicine.

Sources

  1. theguardian.com › 'Superhuman' AI tool spots heart disease in less than 2 seconds
  2. mugglehead.com › AI flags heart disease from routine ECG test within seconds
  3. healthandme.com › New AI Tool Reads ECG In 2 Seconds, Detects Up To 90% Of Heart Valve Disease Cases
  4. convergence-now.com › 'Superhuman' AI ECG Tool Spots Heart Disease in Under 2 Seconds

More on AI