A research team co-led by investigators from Mass General Brigham and the Broad Institute has developed and validated an artificial intelligence model, called ECG2Stroke, that can estimate a person’s risk of stroke up to 10 years in advance using a single 10-second electrocardiogram (ECG). The findings were published in the Journal of the American College of Cardiology.
Current tools used to identify patients at the highest risk of stroke often rely on complex clinical scoring systems that are difficult to scale and, as a result, are not widely used in routine care.
To address this challenge, study researchers created a deep learning model using patient data from Massachusetts General Hospital. The model detects subtle patterns in ECG waveforms. ECGs are inexpensive, non-invasive tests that measure the heart’s electrical activity through electrodes placed on the skin. The researchers then validated ECG2Stroke in patients from Brigham and Women’s Hospital and Beth Israel Deaconess Medical Center. Altogether, data from more than 200,000 patients were used to train and test the model.
The results showed that ECG2Stroke, using only ECG data along with a patient’s age and sex, was able to predict stroke risk over the next decade with accuracy comparable to an established clinical risk score across multiple hospitals and patient groups. Indicators of dysfunction in the atria, the heart’s upper chambers, were among the strongest contributors to the model’s predictions. The tool was especially effective at identifying the risk of cardioembolic stroke, which occurs when a blood clot forms in the heart and travels to the brain and can often be prevented with blood-thinning medications.
If confirmed in prospective, real-world studies, tools like this could help identify patients who should be prioritised for intensive stroke prevention strategies. It may also provide new insights into how abnormalities in the heart’s upper chambers contribute to stroke risk.
Source: Mass General Brigham
Image Credit: iStock
