The boundary conditions for the solution to equation (9) require that the electric potentials at all nodes of the epicardium be known a priori. The epicardial potentials for these simulations were recorded from a patient during the surgical ablation of an accessory pathway to treat Wolff-Parkinson-White syndrome. An electrode sock containing 64 silver wire electrodes was placed over the ventricles during open chest surgery [26, 27], and all 64 channels were sampled every millisecond across several heartbeats. Applying equations (8) and (9) to the geometrical model requires a surface that encloses all the cardiac sources. The measured data that make up the boundary conditions, however, are limited to the outer surface of the ventricles. For these simulations, we created a boundary surface that included the ventricular epicardium and a pseudo-surface ``cap'' corresponding approximately to the base of the heart between ventricles and atria. The atria then fell outside the source region and were treated as passive, blood-filled compartments. In order to generate potentials for all 1112 epicardial surface nodes from the 64 recording sites, we performed surface interpolation. First, each measurement site was mapped to a node in the geometrical model; then, Laplacian interpolation was performed [28, 29]. For the purpose of these simulations, we selected 30 time instants, spaced at 2 millisecond intervals and covering the entire QRS complex. The QRS represents the ventricular activation phase of the heart cycle, and, thus, is the most rapidly changing portion of the ECG; potential distributions also exhibit the largest spatial gradients during this period. By focusing on this part of the signal, we hoped to apply a variety of boundary conditions and thus reveal a wide range of effects of inhomogeneities.