Andrew Keely

Infrastructure and Comparison of Inverse Solutions to the Problem of Electrocardiography

Date: 12/20/07

Abstract:

The inverse problem of electrocardiography is to characterize the electrical activity on the heart using electrical potential measurements on the torso along with a forward model of the volume between the heart and the torso. This problem is ill-posed, and requires regularization to solve. Though we can find a solution using traditional regularization techniques, we can use constraints based on the spatio-temporal nature of the cardiac source to improve our solutions. In this thesis, we present a formulation for the forward model using the boundary element method. We also present techniques for manipulating geometric representations of the heart and torso, along with methods to synthesize additional data from existing measured data. Finally, we present solutions to the inverse problem from several experiments using standard regularization techniques, the Greensite method, and the wave-based potential reconstruction (WBPR) method. We also introduce a new iterative solution using WBPR to improve robustness to poor initialization. Using these results, we show that the iterative WBPR method significantly improves upon other existing methods for solving the inverse problem in the simulations we tested, especially for data on the endocardial surface of the heart.

Committee:
Dana Brooks (advisor)
Gilead Tadmor
Jeroen Stinstra
Alireza Ghodrati