AI Enhances Detection of Heart Conditions in Under Two Seconds

A groundbreaking artificial intelligence system developed by researchers at Imperial College London promises to significantly enhance the detection of heart conditions such as heart failure and valvular disease. This innovation, which leverages electrocardiograms (ECGs), could enable healthcare professionals to identify issues previously undetectable by traditional methods.

For over a century, ECGs have been a staple in medical diagnostics due to their capacity to monitor the heart’s electrical activity. However, the technology has intrinsic limitations, chiefly in its ability to detect subtle patterns indicative of complex conditions. Researchers now assert that their AI solution can extract additional insights from ECG data within a matter of seconds.

This new system was introduced during the 2026 Congress of the European Society of Cardiology, hosted in Munich. It has been trained on millions of ECGs, intricately linked with patient medical records, allowing it to identify critical signals that human eyes might miss. Within a study encompassing 67,000 patients in the United States, the AI was able to correctly identify up to 81% of heart failure cases and 90% of patients suffering from valve disease.

Dr. Sonya Babu-Narayan, a prominent cardiologist and clinical director at the British Heart Foundation, remarked on the efficiency of the AI tool, stating that it provides valuable readings “in what seems like the blink of an eye.” While the technology shows great promise, it importantly serves as a preliminary screening tool rather than a definitive diagnostic instrument. Patients flagged by the AI as high-risk would still require a follow-up echocardiogram—a comprehensive imaging test that assesses heart structure and function—to confirm any diagnosis.

Professor Fu Siong Ng, also a cardiologist at Imperial College, noted that patients often experience delays of up to several months for cardiac ultrasounds. Having an automated tool that can promptly analyse ECGs allows for more efficient patient management, enabling healthcare providers to prioritise who needs further testing based on risk. The technology, known as Cardiovolt.ai and supported by the British Heart Foundation, is being gradually integrated into clinical practice. Despite changes in analytical capabilities, the traditional ECG remains unchanged.

Another notable advancement in cardiac diagnostics is the ECG-CLIP model, developed by scientists at Scripps Research. This AI tool improves heart disease detection using significantly less labelled data. Documented in the reputable journal Lancet Digital Health, ECG-CLIP can detect various heart diseases with greater adaptability than existing models.

This new model demonstrates remarkable efficiency, requiring merely a few confirmed ECG examples to make future diagnoses, akin to a clinician's learning process. Unlike traditional AI systems that demand substantial amounts of labeled training data, ECG-CLIP can achieve accuracy with limited inputs—proving particularly advantageous in settings with scant information.

In evaluating its performance, ECG-CLIP outperformed standard models across several key clinical tasks, including the detection of acute myocardial infarction, cardiac amyloidosis, and hypertrophic cardiomyopathy, while utilising around 91% less hand-labelled training data. The ability to utilise single-lead ECG data for detecting conditions such as myocardial infarctions also potentially extends this model’s applicability in resource-limited environments.

Furthermore, the tool exhibited superior performance in predicting adverse outcomes, demonstrating the potential for better patient management in emergency situations. Researchers have developed saliency maps to clarify which features of ECG data contribute to the model’s predictions, thereby increasing transparency and trust among healthcare professionals.

Moving forward, scientists aim to broaden the datasets available for both Cardiovolt.ai and ECG-CLIP, ultimately enhancing their decision-making capabilities and compatibility with various ECG recording technologies, including wearable devices. These advancements signify a substantial leap forward in cardiac care, encouraging a more proactive approach to heart disease detection and management. As cautioned by the research team, thorough validation through prospective clinical trials is essential before widespread adoption can occur.