Global sensitivity analysis through causal discovery for an electromechanical cardiac model
Published in Biomechanics and Modeling in Mechanobiology, 2026
Increased parameters in advanced cardiac modeling have improved accuracy at the cost of high complexity, making it challenging to identify how specific parameters impact clinical outputs. We propose a new, interpretable global sensitivity analysis (GSA) workflow that leverages causal discovery to disentangle and quantify the joint effects of electromechanical model parameters on clinical biomarkers. We evaluate this approach across three distinct geometries: a healthy heart, hypertrophic cardiomyopathy (HCM), and dilated cardiomyopathy (DCM). Our causal-based framework successfully captures geometry-dependent sensitivities and isolates a reduced subset of influential parameters for each biomarker, providing actionable guidance for pathology-informed model calibration. Compared to traditional GSA techniques like Sobol and Pawn, our method delivers more stable and interpretable results, particularly in data-limited settings. This work extends our previous study by incorporating an additive noise model for validation and expanding the analysis to four clinical biomarkers: ejection fraction, maximum pressure change, isovolumic relaxation time, and early passive filling. The code is publicly available on GitLab.
Recommended citation: Al-Ali S., Rodríguez Padilla J., Sermesant M., Balelli I. (2026). “Global sensitivity analysis through causal discovery for an electromechanical cardiac model.” Biomechanics and Modeling in Mechanobiology.