Cardiac Anatomical and Mechanical Properties Prediction From Electrocardiography (ECG) Using Multi-Modal Representation Learning

What this Trial Is About

Researchers are evaluating the accuracy of a smartphone application called DigitHeart-2 combined with MERL-ECHO and other machine learning models to predict cardiac anatomical and mechanical properties from electrocardiograms (ECGs). This observational study involves patients with heart disease or related cardiac conditions who are scheduled for echocardiography, which serves as the gold standard for comparison. The study aims to improve non-invasive cardiac assessment using advanced technology and artificial intelligence. Participants will have a 12-lead ECG performed if one has not been done within the past six months. Research staff will capture a photo of the ECG using the DigitHeart-2 smartphone application. The machine learning models will analyze the ECG images to predict key heart parameters. Meanwhile, the patient's cardiologist, blinded to these predictions, will perform a standard echocardiography following established guidelines. This process allows researchers to compare the accuracy of the AI-based predictions against traditional echocardiography. Participants will be interviewed to explain the study and provide informed consent. Their medical history and demographics will be collected, and they will undergo the ECG and echocardiography procedures. The primary outcome measured is the accuracy of the models in predicting important echocardiographic parameters on the first day of assessment. Secondary outcomes include accuracy in predicting other cardiac anatomical and mechanical properties. The study is sponsored by The University of Hong Kong and is expected to complete in 2027.

DigitHeart Echo Study