'Superhuman' AI Reads ECGs in Under 2 Seconds, Spotting Hidden Heart Disease

Doctors have developed a “superhuman” AI tool that can spot heart disease in less than two seconds — faster than most people can read this sentence. Presented this weekend at the European Society of Cardiology (ESC) annual congress in Munich, the technology extracts more diagnostic information from a routine electrocardiogram (ECG) than the human eye can typically see, and could fast-track high-risk patients toward lifesaving treatment.

The breakthrough, led by researchers at Imperial College London and funded by the British Heart Foundation (BHF), targets two of the most common forms of heart disease: heart failure and heart valve disease. Both are conditions where early diagnosis is vital, yet both routinely slip through the cracks of standard clinical workflows because the century-old ECG was never designed to find them.

What the tool actually does

The traditional ECG records the heart’s electrical activity — rate, rhythm, and the telltale patterns of a heart attack or arrhythmia. What it cannot do is detect heart failure or valve disease. Confirming those conditions requires an echocardiogram, an ultrasound scan of the heart, for which patients often wait months.

That is the gap this AI fills. Trained on millions of routine ECGs, the model learns subtle waveform signatures that correlate with structural heart problems — patterns, as Imperial researcher Dr Arunashis Sau put it when describing the broader program, that are “not things that clinicians can already do, but things that no cardiologist, no matter how expert, can do.” The result is a read-out in what BHF clinical director Dr Sonya Babu-Narayan called “the blink of an eye”: under two seconds from ECG to risk flag.

The numbers behind the claim

In a trial involving 67,000 patients in the US, funded by the BHF, the tool identified:

• Up to 81% of patients who had heart failure
• Up to 90% of those with heart valve disease

Those figures matter because of scale. ECGs are among the most common tests in medicine — roughly a billion are performed worldwide each year. A model that can re-read every one of those traces and flag hidden disease converts an inexpensive, ubiquitous, 10-second test into a first-line screening tool for conditions that currently demand a scarce imaging appointment to confirm.

Not a standalone diagnosis — a triage rocket

The researchers are careful about what the tool is not. It cannot definitively diagnose or rule out heart failure or valve disease on its own. What it provides is a strong indication that someone may have the conditions — strong enough to justify pulling them out of the standard months-long echocardiogram queue and scanning them urgently.

“It will not detect everyone with a heart condition,” Babu-Narayan cautioned. “But it could be a solution to help fast-track the patients who are most likely to have a heart abnormality. When it comes to the heart, earlier diagnosis and treatment saves and improves lives.”

Professor Fu Siong Ng, professor of cardiology at Imperial College London, pointed to the second, arguably more transformative application: opportunistic screening. Because the model runs on ECGs that are already being performed for unrelated reasons — pre-surgery checks, chest-pain workups, routine physicals — it could silently scan an entire hospital’s ECG output and flag patients whose heart failure or valve disease was never suspected. “The AI model could be run on all ECGs done in a hospital to flag those at highest risk of these diseases, so that they can be diagnosed earlier,” Ng said.

From lab to hospital: Cardiovolt.ai

The technology is closer to the clinic than most conference headlines. The Imperial team, led by Professor Ng’s group at the National Heart and Lung Institute, has already launched a spinout company, Cardiovolt.ai, to commercialize the AI ECG platform, with Ng serving as Chief Medical Officer and Dr Sau as Chief Scientific Officer. The spinout’s initial clinical focus is exactly what the Munich data addresses: detecting heart failure and valve disease that would otherwise go unnoticed.

Dr Ahmed El-Medany, the BHF clinical research fellow who led the Imperial analysis, described the tool as a “superhuman AI” and said the next challenge is designing handheld AI-led ECG readers for healthcare professionals to use at the point of care — a step that would push the technology out of hospital cardiology departments and into GP surgeries and community clinics.

A wider AI wave in cardiology

The ESC congress, held 28–31 August in Munich with AI as a spotlight theme, made clear this is one instance of a broader trend. Delegates also heard that AI-based analysis of five-second facial videos could rapidly detect undiagnosed high blood pressure and type 2 diabetes, in work from the University of Tokyo and the Institute of Science Tokyo. Cardiology — a specialty drowning in waveform and imaging data — is becoming one of the most productive proving grounds for applied medical AI.

Earlier BHF-supported work on the same 1.6-million-ECG dataset showed AI correctly identifying 77% of structural heart problems from a 10-second ECG, versus 64% for human cardiologists. The new results push both speed and disease coverage further.

Why it matters

Heart failure affects tens of millions of people worldwide, and valve disease is increasingly common in ageing populations. Both are far cheaper to manage — and far less deadly — when caught before symptoms become severe. A tool that turns the world’s most common cardiac test into a two-second screening gateway for both conditions is exactly the kind of quiet, deployable AI that changes outcomes at population scale, without a single new piece of hospital hardware.