Session S34.2

Evaluation of Approaches to Solving Electrocardiographic Imaging Problem

M Milanic, V Jazbinsek, DF Wang, J Sinstra,
R Macleod, DH Brooks, R Hren*

University of Ljubljana
Ljubljana, Slovenia

Electrocardiographic imaging (ECGI) is a widely used method of computing potentials on the epicardium from measured or simulated potentials on the torso surface. The discretization of the governing equations is typically done using either the boundary element method (BEM) or finite element method (FEM). The main challenge of the electrocardiographic imaging problem remains its intrinsic ill-posedness, which requires use of regularization techniques to smooth out the solution; the amount of smoothing that is still clinically acceptable is a subject of ongoing research.
The objective of this paper is twofold: (i) to systematically compare the performance of the FEM and BEM; (ii) to systematically compare various mathematical techniques in regularizing the ECGI problem.
For the purposes of the forward computations, we employed potentials measured on the cylindrical cage placed around the physiological source (canine heart) and situated in the electrolytic torso tank. The cylindrical cage was discretized into 1200 elements (602 nodes), torso surface into 1538 elements (771 nodes) and both FEM and BEM were used.
The inverse potentials on the »epicardial« surface were recovered using the following regularization techniques: zero-order and first-order Tikhonov regularization imposing either L1- or L2-norm constraint, truncated singular value decomposition (TSVD), conjugate gradient method (CG), LSQR method based on Lanczos bidiagonalization, and í-method. To choose the regularization parameter, we used a gold standard L-curve method. Finally, we also comparatively evaluated the performance of time-constrained regularization methods and activation-based imaging.

(Abstract Control Number: 105)