AI-enabled ECG to reduce demand for echocardiography and shorten diagnostic waiting lists

Background and Rationale Echocardiography, a cornerstone of cardiovascular diagnostics, is essential for the assessment of heart structure and function in the diagnosis and monitoring of heart disease. Over 1.8 million echocardiograms were performed in England in 2024-25, yet demand continues to exceed capacity. As of February 2025, 147,648 patients were awaiting echocardiography, with 24% waiting over 6 weeks and 11% beyond 13 weeks. However, only one-quarter of echocardiograms yield clinically actionable findings, highlighting inefficiencies in the referral pathway. The need for precision is particularly acute in suspected heart failure, which makes up 44% of referrals. Heart failure affects 1-2% of the UK population and consumes 2% of the NHS budget. Its prevalence is increasing due to ageing demographics and improved survival from other cardiac conditions. Early detection and treatment improve outcomes, but diagnosis is often delayed by bottlenecks in imaging. AI-ECG Solution The Intelligent Heart Evaluation Framework (iHeF) is an AI-based triage approach using standard 12-lead electrocardiograms (ECGs). Our models, trained on ~42,000 ECGs paired with the gold-standard cardiac MRIs, predict key measurements of heart structure and function including left ventricular ejection fraction (LVEF), global longitudinal strain, filling pressure, volume, and mass. These predictions support early detection of abnormal cardiac function, and can be used to triage patients for cardiac imaging. The models show strong diagnostic performance, with area under receiver operating characteristic curve (AUROC) values between 0.78 and 0.94 across imaging markers, and a mean absolute error of 5.6% for LVEF – outperforming typical human variability. Objectives The project aims to validate our AI-ECG models in real-world clinical settings, assess user and patient acceptability, evaluate economic benefits, develop market readiness, and prepare for regulatory approval. Outputs will include clinical performance metrics, usability insights, economic modelling, and a Quality Management System (QMS) supporting CE/UKCA and US FDA certification. Methodology The project includes six work packages (WPs): • WP1: Retrospective validation using Barts Health NHS Trust data, assessing accuracy via AUROC, sensitivity, and specificity • WP2: Usability studies with patients and clinicians, gathering in-situ feedback during clinical use • WP3: Prospective validation at multiple NHS sites, enrolling 500 patients to assess reductions in unnecessary echocardiography and ensure diagnostic safety • WP4: Health economic analysis quantifying cost savings, resource optimisation, and efficiency gains • WP5: Commercial strategy development, including competitor mapping and a go-to-market roadmap • WP6: Regulatory preparation, including QMS implementation and documentation for CE/UKCA and FDA approval Impact AI-ECG could substantially reduce unnecessary echocardiograms, shorten diagnostic wait times, and streamline triage. Its integration into NHS workflows may accelerate heart failure diagnosis, ease imaging service pressures, and reduce costs. Conclusion This project marks a step-change in cardiac diagnostics by transforming the widely used ECG into a powerful triage tool through AI. By improving accuracy, reducing delays, and enabling earlier diagnosis, AI-ECG will improve outcomes for patients with suspected heart disease while shortening waiting lists.