Acceleration of FEM-based transfer matrix computation for forward and inverse problems of electrocardiography

Abstract

The distributions of transmembrane voltage (TMV) within the cardiac tissue are linearly connected with the patient's body surface potential maps (BSPMs) at every time instant. The matrix describing the relation between the respective distributions is referred to as the transfer matrix. This matrix can be employed to carry out forward calculations in order to find the BSPM for any given distribution of TMV inside the heart. Its inverse can be used to reconstruct the cardiac activity non-invasively, which can be an important diagnostic tool in the clinical practice. The computation of this matrix using the finite element method can be quite time-consuming. In this work, a method is proposed allowing to speed up this process by computing an approximate transfer matrix instead of the precise one. The method is tested on three realistic anatomical models of real-world patients. It is shown that the computation time can be reduced by 50% without loss of accuracy.

Patient Specific Simulation of Body Surface ECG using the Finite Element Method

“…An alternative approach to the realization of patient specific ECG simulation is, of course, solving the inverse problem, which researchers have pursued for several years . Hitherto, attempts have been successful only in the estimation of epicardial maps, owing to the well‐known ill‐posed nature of the problem.…”

“…An alternative approach to the realization of patient specific ECG simulation is, of course, solving the inverse problem, which researchers have pursued for several years. 3,5,[18][19][20] Hitherto, attempts have been successful only in the estimation of epicardial maps, 3 owing to the wellknown ill-posed nature of the problem. However, the forward problem approach adopted in this study can provide information on the activation and repolarization sequences within the wall, but this requires iterative simulations with manual adjustment of the parameters.…”

“…Furthermore, the important information retrievable from the body surface ECG has recently been revisited . In fact, many attempts simulating the ECG based on either a mono‐ or bi‐domain formulation of the cell model for cardiac electrophysiology implemented in image‐based models of animal and human hearts have been reported, though relatively few studies combine these heart models with an image based torso with various organs segmented to reproduce a realistic ECG.…”

“…Researchers have circumvented this problem by using the boundary element method (BEM) either with the torso (thorax) as the isotropic volume conductor or with a realistic torso model using the two step approach. In this approach, a priori obtained electrical activity of the heart is mapped onto a coarse torso model and the body surface potential is calculated using either the approximate transfer matrix or a finite element solution . Despite detailed discussion of the convergence and accuracy of these methods, we are still unsure whether such approaches ensure clinically approved accuracy for diseased individuals with complex anatomy of both the thorax and heart accompanied by highly variable modes of cardiac excitation and repolarization.…”

The Nightingale Prize 2010 for best MBEC paper in 2009 awarded

“…Potse et al [14] introduced extracellular in addition to intracellular anisotropy into the models that predict ECG signals. Also the work of Farina et al [5] was directed to the ECG but then to accelerate computational methods. Ferrario et al [6] developed a method for fetal heart rate variability in order to identify early intrauterine growth-restriction of the fetuses.…”