Sensitivity Analysis of Electrocardiogram Features to Computational Model Input Parameters

Cardiac models of electrophysiology capable of generating simulated electrocardiogram (ECG) signals are an increasingly valuable tool for both personalised medicine and understanding cardiac pathologies. Sensitivity analysis (SA) can provide crucial insight into how simulation parameters affect ECG morphology. We use two SA methods, direct numerical evaluation of integrals and polynomial chaos expansion, to calculate main and total effects for ECG features extracted from QRS complexes generated by a cardiac ventricular model. The importance of stimulation site parameters on output ECG features is evaluated. SA methods can highlight and quantify important input parameters for different ECG morphology features, which in some cases can be linked to physiological explanations. For example R peak amplitude in lead II depends on apicobasal location of stimulation sites in the left ventricle. Furthermore, different SA methods have different strengths and weaknesses. Insight into parameter importance supports model development and allows for more nuanced and patient-specific simulation changes.