AI detects hidden heart disease from routine ECG in just 2 seconds
Researchers at Imperial College London have developed an artificial intelligence system that analyzes routine ECGs in less than two seconds to identify patients with hidden heart failure or valve disease. Trained on 1.6 million ECGs from Brazil and millions of recordings from the US, the system achieves diagnostic accuracy of 81–90 percent and is intended primarily as a screening and prioritization tool for further cardiac investigations rather than as a replacement for conventional diagnosis.
Exactly the kind of technology that is incredibly useful. It takes what we already have, like ECG, and gives us something we didn't have before. I'm looking forward to more equally useful applications of AI in healthcare.
How does the AI system detect heart disease from ECG recordings?
The system analyzes electrical signals captured during an ECG and identifies subtle patterns associated with specific cardiovascular conditions. It was trained on millions of ECG records linked to patient diagnoses, learning to recognize features in the recording that would normally require additional testing to confirm.
What are the key limitations or risks of such a system?
The AI cannot independently diagnose heart failure or valve disease—those conditions require further clinical assessment and tests like echocardiography. The system may also miss some forms of heart disease and requires additional validation across different populations and healthcare settings before routine clinical use.
How do the researchers envision this technology being used in clinical practice?
The team sees the system primarily as a screening and prioritization tool to help hospitals identify patients more likely to have significant disease and expedite their referral for echocardiography. This could be particularly valuable in health systems where demand for echocardiography exceeds available capacity.
- An Alien Mind: OpenAI's Chief Scientist on Intelligence We Don't Fully Understand — openai.com 72 % match
- OpenAI's coding agents are accelerating AI development and now exceed human research capacity — openai.com 72 % match
- CenSegNet reveals hidden centrosome patterns in breast cancer with new diagnostic insights — research.uhs.nhs.uk 71 % match
- Imperial College London
- Cardiovolt.ai
- Dr Ahmed El-Medany
- Professor Fu Siong Ng
- Dr Sonya Babu-Narayan
- European Society of Cardiology
- British Heart Foundation