A fast model for solving the ECG forward problem based on an evolutionary algorithm

The estimation of solutions of the ElectroCardioGraphy (ECG) inverse problem is a key issue for non-invasive diagnosis and therapy of cardiac arrhythmia. A number of methods have been proposed to estimate such solutions, but their quantitative evaluation is not simple due to the lack of reference data. One way to proceed to their evaluation is to solve the forward problem, by generating simulations using a mathematical model representing the initiation and propagation of the cardiac electrical activity through the heart and torso. These models allow for the synthesis of torso ECG potentials corresponding to known, simulated cardiac potential mappings. However, most of the existing cardiac propagation models are too complex for this kind of application, with a significant number of parameters to be tuned, leading to high computational costs. In this paper, we propose a reliable and fast framework for building a cardiac and torso propagation model that generates sufficiently realistic healthy ECGs to perform reference-based evaluations of inverse problem methods. We used a set of tissue-level structures representing the cardiac electrical activity through FitzHugh-Nagumo model and a monodomain formalism for cardiac propagation. Low-resolution 3D Finite Element Method (FEM) representations are performed. An evolutionary algorithm is used to identify the main model parameters that provide the best fit to a real healthy ECG. The obtained preliminary results show that it is possible to generate realistic healthy ECGs using such simplified 3D heart torso models with very low computational costs.