Superpower Daily
Clipto Raises $15M at $250M Valuation to Turn Local Files Into...Nvidia Puts $3.5B Into MediaTek to Keep Custom AI Chips in Its...Unitree’s $4,017 Go2 Pro Makes Robot Dogs Cheaper, Not Yet Ever...Claims Adjusters’ AI Backlash Hits 98% Negative Reviews as Insu...Micron Trades at Just Above 6x Forward Earnings as Contracts Te...Replica Cyber Ships Isolated AI Agents With Runtime Controls fo...Clipto Raises $15M at $250M Valuation to Turn Local Files Into...Nvidia Puts $3.5B Into MediaTek to Keep Custom AI Chips in Its...Unitree’s $4,017 Go2 Pro Makes Robot Dogs Cheaper, Not Yet Ever...Claims Adjusters’ AI Backlash Hits 98% Negative Reviews as Insu...Micron Trades at Just Above 6x Forward Earnings as Contracts Te...Replica Cyber Ships Isolated AI Agents With Runtime Controls fo...
Toolspublished

AI ECG Tool Flags Heart Risk in Two Seconds, but Still Needs the Scan That Diagnoses It

The reported results suggest routine electrical heart traces could help prioritize scarce ultrasound appointments. The system’s stated role, however, stops short of confirming disease or ruling it out.

By 2 min read
AI ECG Tool Flags Heart Risk in Two Seconds, but Still Needs the Scan That Diagnoses It
AI ECG Tool Flags Heart Risk in Two Seconds, but Still Needs the Scan That Diagnoses It
Sign in to save and react

Listen to this story

The audio brief

About 1:27
0:001:27
Read transcript
An AI system can read a routine heart trace in less than two seconds and flag people who may need urgent follow-up for heart failure or valve disease. The important caveat is that it still needs the scan that actually confirms those conditions. The system was trained on millions of routine ECGs to detect signals that are not typically visible in the recordings. In a US trial involving 67,000 patients, it identified up to 81 percent of heart-failure cases and up to 90 percent of valve-disease cases. Those are case-finding results: they show how many cases the system flagged in that trial, not how many diagnoses it could make on its own. An ECG records electrical activity, including the heart’s rate and rhythm. But confirming heart failure or valve disease requires an echocardiogram, an ultrasound scan of the heart. Patients can wait several months after referral, so the proposed role here is triage. The AI could help hospitals prioritize the people most at risk for faster echocardiograms, and potentially scan hospital ECGs to surface disease that was not previously suspected. The analysis was led through Imperial College London, funded by the British Heart Foundation, and presented at the European Society of Cardiology’s annual congress in Munich. Ahmed El-Medany says the next hardware challenge is a handheld AI ECG reader for clinicians. The constraint remains clear: faster risk flags, followed by confirmatory scanning—not a standalone diagnostic device.

Story brief

3 key points

Researchers led through Imperial College London presented an AI ECG system that scans routine heart traces in under two seconds and flags patients who may need urgent follow-up. In a 67,000-patient US trial, it identified up to 81% of heart-failure cases and 90% of valve-disease cases, but those figures are case-finding results—not diagnoses. The clinical opportunity is triage: prioritizing echocardiograms, which...

  1. 01

    The system was trained on millions of routine ECGs to detect signals not typically visible in recordings.

  2. 02

    Echocardiography remains the required confirmation test; the AI cannot diagnose or rule out either condition.

  3. 03

    Researchers propose deploying it across hospital ECGs to surface previously unsuspected disease.

A newly presented AI ECG system can produce a risk read-out from a routine heart trace in less than two seconds. It may help move people with possible heart failure or valve disease toward faster ultrasound scans, but it cannot diagnose or exclude either condition by itself.

The findings were presented at the European Society of Cardiology annual congress in Munich. The British Heart Foundation funded the research, which was led through an Imperial College London analysis. Researchers trained the model on millions of routine ECGs to extract signals beyond those typically visible in the recordings.

Case finding is not a clinical verdict

In a US trial of 67,000 patients, the tool identified up to 81% of patients with heart failure and up to 90% of those with heart valve disease. Those results describe how many cases the tool identified in that trial; they do not make its output a definitive diagnosis.

Reported US trial results
Up to 81%Heart failure cases identified

The trial included 67,000 patients in the US.

Up to 90%Valve disease cases identified

The valve-disease result came from the same trial.

The proposed gain is speed to echocardiography

An ECG records the heart’s electrical activity, including its rate and rhythm. Diagnosis of heart failure or valve disease requires an echocardiogram, an ultrasound scan of the heart, and patients can wait several months after referral for that scan.

That makes the tool a routing layer rather than a replacement for clinical assessment. Researchers said it could identify the patients at greatest risk so they can be prioritized for faster, more urgent echocardiograms. They also proposed running it across hospital ECGs to flag people whose conditions were not previously suspected.

The next hardware question

Ahmed El-Medany, the British Heart Foundation clinical research fellow who led the Imperial analysis, said the next challenge is designing handheld AI-led ECG readers for healthcare professionals. For now, the evidence supports a fast risk flag followed by confirmatory scanning, rather than a standalone diagnostic device.

Sources

  1. theguardian.com‘Superhuman’ AI tool spots heart disease in less than 2 seconds

Related coverage

Continue reading