Probabilistic Calibration of a Cardiac Electromechanical Digital Twin

Cardiac conduction disorders, such as left bundle branch block, severely impair cardiac mechanical synchrony and pump function. While Cardiac Resynchronization Therapy (CRT) is the current gold-standard treatment, optimizing patient selection and pacing strategies remains challenging. In this context, computational models of cardiac electromechanics provide a robust framework to understand the underlying pathophysiological mechanisms. Advancing these models into digital twins represents a frontier in precision cardiology and requires rigorous model personalization. In this seminar, we will explore the development of a patient-specific computational framework to simulate cardiac electromechanics and the acute hemodynamic response to CRT. The first part of the talk will introduce the clinical context and the physics-based modeling framework. Specifically, it will describe the coupling between three-dimensional electrical activation and a closed-loop cardiovascular mechanics model (CircAdapt) to capture global hemodynamics. The second part of the seminar will focus on a key aspect of translating these tools into clinical practice: model personalization under uncertainty. Relying exclusively on non-invasive clinical data, we will discuss how integrating physics-based models with Bayesian optimization and Approximate Bayesian Computation enables the probabilistic calibration of these complex systems. By quantifying uncertainty in both the electrical activation patterns and the resulting mechanical outputs, this framework enables the simulation of various pacing strategies and the non-invasive estimation of their hemodynamic effects, ultimately advancing the development of uncertainty-aware digital twins for personalized CRT planning.