Key result
Causal discovery improves cardiac sensitivity analysis stability and interpretability over classical methods.
Why the study?
Advanced cardiac modeling has increased model complexity, making the identification and understanding of relationships between parameters and clinical outputs a central challenge.
Causal discovery provides a robust alternative for sensitivity analysis in complex cardiac electromechanical models, aiding in pathology-specific personalization.
Answers are grounded in this paper and cite their sources.
May aid pathology-specific cardiac model personalization; leaves open human translation from this animal study.
The rapid development of advanced cardiac imaging and modeling technologies has significantly increased the number of parameters required to accurately characterize cardiac function, thereby improving model accuracy at the cost of increased complexity. Identifying and understanding the relationships between model parameters and clinically relevant outputs has therefore become a central challenge in both cardiac modeling and pathology-specific personalization. In this work, we propose a new approach to perform interpretable global sensitivity analysis by leveraging causal discovery. More precisely, causal discovery is employed to disentangle and quantify the joint effect of multiple parameters of an electromechanical cardiac model on some clinical biomarkers, enabling a robust multi-output global sensitivity analysis. We further explore how these parameter–biomarker relationships vary across cardiac geometries, comparing a healthy heart with two pathological conditions: hypertrophic cardiomyopathy (HCM) and dilated cardiomyopathy (DCM). The resulting causal impacts highlight geometry-dependent sensitivities and identify a reduced subset of influential parameters for each biomarker, providing actionable guidance for model calibration under pathological conditions, and supporting the definition of pathology-informed priors for personalization workflows. Finally, compared to classical global sensitivity analysis techniques such as Sobol and Pawn, our proposed approach yields more stable and interpretable results, even when only a limited number of simulations is available. Overall, this study demonstrates that causal discovery offers a powerful and reliable alternative to perform sensitivity analysis in complex cardiac models, particularly in data-limited settings where biological constraints alone are insufficient. The proposed workflow is made publicly available on GitLab . This work extends our previous study (Al-Ali et al. 2025 ) by broadening the sensitivity analysis beyond the classical biomarkers–ejection fraction (EF) and the maximum rate of pressure change in the left ventricular cavity (max(dP/dt))–to include two additional clinically relevant outputs: the isovolumic relaxation time (iso r ), as an indicator of diastolic function, and the early passive filling of the left ventricle (QRS E ). Moreover, we considered two pathological cases to study the variability of parameters-outputs impacts, and we employed an additive noise model (ANM) to further support and validate our causal based global sensitivity analysis.
No takes yet. Share an insight, caveat, or question.
Al-Ali et al. (2026) studied Hypertrophic cardiomyopathy (HCM) and dilated cardiomyopathy (DCM). Causal discovery approach vs. Classical global sensitivity analysis techniques (Sobol and Pawn) was evaluated on Parameter-biomarker relationships (ejection fraction, max(dP/dt), isovolumic relaxation time, early passive filling). A causal discovery approach yielded more stable and interpretable global sensitivity analysis results for electromechanical cardiac models compared to classical techniques like Sobol and Pawn.
Synapse has enriched 10 related papers. Consider them for comparative context: