An AI algorithm interprets an ECG in under 2 seconds and detects heart disease

Researchers from Imperial College London presented an artificial intelligence-based algorithm at the European Society of Cardiology Congress in Munich, capable of interpreting an electrocardiogram (ECG) in under two seconds. The system was developed to rapidly flag early signs of heart failure and valvular heart disease.

How accurate is the algorithm

Unlike conventional interpretation carried out by a physician, the algorithm was trained on millions of ECG recordings and can identify patterns and subtle electrical changes that are difficult to detect with the human eye. Tested on more than 67,000 patients, the system achieved:

• 81% detection rate for heart failure cases
• 90% detection rate for severe valvular conditions

Dr Ahmed El-Medany, one of the researchers involved in the project, described the algorithm's performance as "superhuman artificial intelligence", owing to its ability to analyse minute electrical changes in cardiac activity.

What this means for patients

The technology does not replace investigations such as echocardiography; rather, it functions as a rapid triage tool. At present, many patients wait months for a confirmatory cardiac ultrasound. With this system, general practitioners and emergency physicians could identify far more quickly who requires priority further investigation.

A significant advantage is the potential for incidental diagnosis: the algorithm can detect an underlying cardiac problem even when a patient undergoes a routine ECG for an entirely different condition, before severe symptoms emerge.

In the next phase, the research team intends to integrate this software into portable, compact ECG devices, so that it can be used widely in GP surgeries and outpatient clinics.