Computational cardiac modeling has matured into a well-established domain within biomedical computing and is increasingly transitioning from a research-focused discipline to a valuable tool for clinical decision support. As this shift progresses, ensuring the reliability and interpretability of model predictions becomes essential, particularly when combining data-driven AI approaches with traditional mechanistic models. In such hybrid modeling frameworks, uncertainty quantification (UQ) plays a critical role in evaluating the robustness of predictions and in building clinical trust.
We focus on advancing UQ methodologies in computational heart mechanics, where both the intrinsic variability of biological parameters and the limitations of data-driven components must be carefully assessed. Specifically, we discuss how the polynomial chaos expansion (PCE) method can be used for UQ in cardiac mechanics models. PCE is a non-intrusive meta-modeling approach that approximates the original high-fidelity model using orthonormal polynomial surrogates over the space of random inputs, and allows us to efficiently propagate uncertainty from model input parameters to clinically relevant output quantities such as stroke volume, ejection fraction, stress, and strain. The method has been applied to and quantify the impact of uncertainty in inputs such as material behavior, tissue microstructure (fiber orientation) and geometry, and offers a promising pathway toward safe and trustworthy personalized simulations.