Pranav Bhagirath

138 Chapter 7 In order to incorporate the electrical pathways and to account for the differences in conductivity, the conduction system of the heart was also modeled. The origin of activation for sinus rhythm was defined at the lateral border of the superior vena cava and right atrium junction, approximating the location of the sinus node. The right and left bundle branches were also defined. Image segmentation and mesh generation The three perpendicular stacks of MRI images were loaded in a custom developed software tool. Subsequently, the pre-defined anatomical and electrical components were segmented. The extra-cardiac thoracic volumes (lungs and thorax) were assigned conductivity values (Σ) described in the literature; lungs (0.04 S/m) and thorax (0.2 S/m) [8] . To investigate the feasibility of incorporating tissue properties, two segmentations were created containing pre-defined regions of edema (speed 0.2 m/s and Σ 0.0325 S/m) and scar tissue (speed 0 m/s and Σ 0 S/m) respectively. The software tool was used to generate a script containing the geometrical description of the topologically correct segmentation result. The script generated by the segmentation tool was used as input to the GMSH mesh generator to construct the computational mesh [9] . Inverse potential mapping and simulation of cardiac activation In the healthy volunteers, the IPM’s were reconstructed by multiplication of the BSP’s with (T T T + λ 2 I) -1 T T where T is the transfer matrix and λ = 0.01. This provided the cardiac surface potentials from which the IPM was derived. Simulations were performed using a fixed conduction velocity model. In concordance with the literature, an effective conduction velocity of 0.6 m/s was defined for both atrial and ventricular myocardium [10] . Based on the same literature the bundle branches were assigned a speed (2 m/s) and delay (0 m/s). The simulations were performed (1) computing direct connections between all mesh nodes using the Möller-Trumbore algorithm [11] followed by (2) solving the shortest path problem amongst the computed paths using the Floyd-Warshall algorithm [12] . The mesh