Prospective evaluation of artificial intelligence-enhanced electrocardiography for diagnosis of structural heart disease

Undiagnosed structural heart disease (SHD) contributes to significant morbidity, hospital admissions, and early mortality, conferring a huge financial burden. Our group has been at the forefront of developing explainable and actionable artificial intelligence (AI) enhanced electrocardiogram (AI-ECG) models offering a promising approach to early diagnosis, risk detection, and disease management using routine 12-lead ECGs, and through wearable and portable devices that enables cheap, non-invasive health monitoring. Although our models have been extensively externally validated through retrospective analysis of large datasets, prospective validation in real-world settings is essential to establish the efficacy of these technologies in clinical practice and instil clinician and patient confidence. We hypothesise that our AI-ECG platform can accurately identify SHD, including reduced left ventricular ejection fraction, moderate to severe valvular heart disease, and pulmonary hypertension, from ‘real world’ 12-lead ECG data, and ECG data derived from consumer and medical portable and wearable devices. In addition, we will compare and combine our AI-ECG detections with established biomarkers such as NT-proBNP in their ability to diagnose SHD. Results will support the application of AI-ECG models using 12-lead ECGs, and ECG data using wearable and portable medical devices, to facilitate their role in providing inexpensive and non-invasive point-of-care health assessments.