Cardiac diseases involve a plethora of dysfunction in subcellular processes, involving ionic currents, calcium signalling, and cross-bridge cycling dynamics, as well as tissue-level characteristics, involving mechanical properties. However, the link between these subcellular abnormalities and their clinical signatures, such as the left ventricular ejection fraction and QT interval, remain speculative in animal or iPSC-CM models of disease due to a lack of clarity for human translation. Human-based biophysically detailed computational modelling and simulation has been used to bridge this gap. However, the lack of convincing credibility evaluations of these models can hinder the widespread use of such techniques in clinically relevant research.
In this work, we compile a list of evaluation criterion biomarkers, with clinical justification, covering ECG, pressure-volume, displacement and strain biomarkers compiled from the UK BioBank, supplemented with deep phenotyping studies with smaller sample sizes. We evaluate our multi-scale model of human biventricular electromechanics and highlight limitations of using literature-based measurements of model parameters.
Using one-at-a-time sensitivity analysis we saw that simulated T-wave amplitude was positively correlated to passive mechanical properties of the myocardium and the pericardium and the QT interval was negatively correlated to aortic resistance. Simulated left ventricular ejection fraction decreased by 8% by doubling the baseline incompressibility parameter of the tissue and increased by 17% by doubling the baseline myosin calcium sensitivity. Peak systolic fibre shortening increased by 8.5% by doubling the L-type calcium current amplitude and decreased by 4% by doubling the pericardial stiffness parameter. Using these results, we then calibrated a new baseline model which is better able to replicate all listed physiological biomarkers.
Our framework for calibrating and evaluating the computational models of cardiac electromechanics helps to improve clinical uptake of computational methods of scientific exploration through building trust in the model behaviour and providing insights regarding multi-scale mechanisms of physiology.