Session P35.4

The Determination of the Bidomain Conductivity Values of Heart Tissue

LS Graham*, D Kilpatrick, FP Sainsbury, AC Yong

The University of Tasmania
Hobart, Australia

A method for determining the bidomain conductivity tensors was developed. The investigation was due to the variability in the different sets of conductivity values reported in the literature.
The method involved mapping the propagation of the electrical activation of cardiac tissue, initiated by point stimulation, via extracellular electrodes. A transient bidomain model was used to simulate the electrical phenomena. The optimum set of conductivity values was achieved by minimising the difference between the bidomain model output and the measured extracellular potential, by means of inverse techniques in parameter estimation, such as least-squares (LS) and Singular Value Decomposition (SVD). The method takes a different approach to the conventional four-electrode technique, as it does not require the small electrode separation needed to separate the extracellular current from the intracellular.
A validation study of the parameter estimation methods was performed. For the validation, synthetic data generated from a bidomain model with a set of known conductivity values were used with a percentage of random Gaussian noise. The conductivity values were able to be retrieved back, reasonably close to their original values. Data from experiments reported in the literature was also used to test the method.
Other parameters in the bidomain model could also be estimated, such as the local myocardial fibre direction, and membrane capacitance. Though, parameters such as membrane capacitance adversely effected the performance of the parameter estimation, due to parameter insensitivity or correlation. Stricter convergence criteria were then required, resulting in longer computational times.
Overall, the LS method with the use of the Marquart parameter, which determines how large a step the parameters are updated between optimisation iterations, seemed to work best, whereas the SVD method tended to overshoot the optimum parameter set, when involving experimental data.

(Abstract Control Number: 138)