Post-MI Remodeling Mechanics of Left Ventricle: Microstructure-Informed Models, Identifiability, and Uncertainty for Patient-Specific Prediction

Abstract

Background: Myocardial infarction (MI) causes spatially heterogeneous loss of contractility and progressive extracellular matrix remodeling, altering left ventricular mechanics from the acute phase through chronic remodeling. This review integrates current understanding of infarct, border-zone, and remote-myocardial microstructure with organ-scale mechanics and patient-specific computational modeling. Methods: A narrative review and perspective were conducted using the literature identified through PubMed/MEDLINE, Scopus, and Web of Science, supplemented by targeted searches of IEEE Xplore and Google Scholar. Experimental, imaging, computational, and translational studies were synthesised, with emphasis on post-MI constitutive behaviour, finite-element and growth-and-remodeling models, imaging-informed personalization, inverse parameter estimation, identifiability, model calibration, verification and validation, and uncertainty quantification. No quantitative synthesis was performed because of substantial heterogeneity in study populations, imaging modalities, constitutive formulations, boundary conditions, calibration procedures, and reported outcomes. Results: Contemporary post-MI models can reproduce ventricular volumes, regional strain patterns, and selected haemodynamic measures, while enabling counterfactual simulations of infarct stiffness, border-zone contractility, and loading interventions. However, clinically credible prediction remains constrained by limited in vivo observability of regional tissue properties, poor parameter identifiability, confounding between material properties and loading conditions, and incomplete treatment of measurement, parameter, and model-form uncertainty. Conclusions: The novelty of this review lies in framing post-MI patient-specific modeling as an identifiability- and uncertainty-limited inverse problem rather than solely as a model-fitting exercise. It proposes that translation toward decision-grade prediction requires parsimonious models aligned with a defined clinical context of use, constrained by microstructure-informed priors, multimodal pressure–volume–strain data, longitudinal validation, and routine reporting of parameter identifiability and predictive uncertainty.