Image-based models of human cardiac electrophysiology (EP) are increasingly relevant as a clinical research tool. However, current clinical EP models typically lack patient-specificity as they mostly rely on generic data, or are simplified to fit within clinical time scales due to a lack of computational efficiency.

This study therefore aimed to develop an efficient, clinically-compatible automated workflow for patient-specific parameterization of human cardiac EP models using non-invasive standard ECG recordings. Specifically, we focused on the parameterization of the depolarization phase during sinus activation to reproduce QRS morphology in a left-ventricular (LV) model.

Two MRI-based LV models, A and B, were utilized. A simplified activation model was defined based on the assumption that activation patterns are determined by the locations, $\mathbf{x}$, of septal, anterior and posterior fascicle of the His-Purkinje system (HPS) and their relative activation timings. HPS is represented by a fast-conducting endocardial layer with principal fiber conduction velocity, $\mathbf{v_f}$. A reaction-eikonal model was employed to compute activation sequences, source distributions and ECGs. Latin hypercube sampling was used to sweep the input parameter space [$\mathbf{x}$, $\mathbf{t}$, $\mathbf{v}$]. Quantitative comparison between QRS complexes of simulated and measured ECGs was performed using a normalized correlation coefficient and L2 norm.

Activation sequences with corresponding QRS complex were simulated in approximately 11 seconds on 6 cores. Inherent morphological characteristics of the QRS complex could be represented by our model parameter space [$\mathbf{x}$, $\mathbf{t}$, $\mathbf{v}$]. Correlation coefficients and L2 norms of 0.86 and 20.22 were attained for model A, and 0.93 and 3.06 for model B, respectively.

The feasibility of generating patient-specific LV activation sequences based on measured QRS complexes in non-invasive ECG recordings was demonstrated. The efficiency of the proposed model will facilitate its use in further data-driven clinical EP model parameterization with increased complexity.

This is joint work with A. Prassl, J. Bayer, E. Vigmond, A. Neic, and G. Plank. This research was supported by the Grant F3210-N18 from the Austrian Science Fund, and the European Commission Grant CardioProof 611232.