A novel AI-enabled measure of early systolic function to guide management of heart failure

Research questions Does a novel measure of left ventricular function - first phase ejection fraction (EF1) reliably predict the response to cardiac resynchronisation therapy (CRT)?

Does optimisation of CRT settings to maximise EF1 6-month after CRT to improve the proportion of patients who respond to the procedure? Can measurement of EF1 be automated with artificial intelligence (AI)? Background Heart failure is a common cardiac endpoint and has poor prognosis.

Current measures of heart function fail to accurately identify patients who may benefit from potential life-saving interventions such as CRT.

When selected by current criteria, more than 30% patients undergoing CRT do not improve and are subject to risks imposed by the procedure.

EF1, a novel measure of heart function I developed during my NIHR lectureship, has the potential to identify patients who respond to CRT and may be used as a target to optimise outcomes.

Aims and objectives To examine whether EF1 measured before CRT implantation predicts response at 6-month after procedure.

To examine whether patients who do not respond at 6-month after CRT benefit from an EF1 guided CRT optimisation compared to standard care (no-optimisation). To use AI to automate EF1 measurement. Methods Baseline visit: 400 patients referred for CRT will be prospectively recruited from 5 hospitals in 24-month.

EF1 will be measured from clinical echocardiography before implantation. 6-month visit: Patients will be followed-up at 6-months after CRT implantation according to standard care for the evaluation of their response as defined by the primary endpoint. (Objective 1) Randomisation: Patients who do not respond at the 6-month as defined by the primary endpoint will be randomised 1:1 to an EF1 guided CRT optimisation group or to standard-care (no-optimisation). 12-month visit: Randomised patients will be closely monitored for signs of adverse effects and followed-up at 12-month post implantation to examine whether response (defined by the primary endpoint) in the optimisation group is greater than that in the standard-care group. (Objective 2) 36-month visit: All patients will be followed-up at 36-month for the evaluation of clinical events.

AI algorithms will be tested in the proposed study and in 2300 patients with aortic stenosis collected during my lectureship. (Objective 3) Primary endpoint: A reduction in left ventricular end-systolic volume >15%. Secondary endpoint: An improvement in clinical composite score. Timelines for delivery Ethical approval will be in place prior to study start date.

Month 3: external sites initiation. Month 24: complete recruitment. Month 30: complete 6-month follow-up and randomisation. Month 36: complete 12-month follow-up. Month 60: complete 36-month follow-up. Month 24-60: AI development and health economic evaluation.

Anticipated impact and dissemination This study will improve patient selection and outcomes for patients undergoing CRT.

Together with collaborators and experts, I will provide stronger evidence to include the EF1 in guidelines for management of heart failure and aortic stenosis.

This technology has the potential to be widely adapted into routine clinical diagnostics without additional training and financial burden to the NHS. Dissemination will be through PPI groups, scientific conferences, and publications.