Ventricular Electrophysiology Modeling Protocol

The test case for benchmarking code scalability is derived from the heart of a cardiac resynchronization therapy candidate, the model development has been described previously (Niederer et al., 2010 (link)). Briefly, ventricular epicardium and left and right ventricular endocardium were segmented from MRI images acquired at end diastole using cmgui². The segmented surfaces were fitted with cubic Hermite surfaces and converted into a volume mesh using CMISS³. The cubic Hermite mesh was converted to a binary image stack which was fed into the image-based mesh generation software Tarantula⁴ to generate a high resolution unstructured tetrahedral mesh. The resulting mesh consisted of 26 million nodes and 153 million elements with a mean edge length of 0.25 mm. To initialize the electrophysiological state of the ventricles, the isolated TNNP cell model was paced for 500 beats at 1 Hz to reach a limit cycle which was used then as the initial state in the whole ventricles. Left ventricular endocardial activation maps were used to fit monodomain conduction parameters giving values of σ_ml = 0.035 and σ_mt = 0.023 Sm⁻¹ in the fiber and transverse fiber direction, respectively. Activation at the septum was defined from left ventricular endocardial activation maps and right ventricular activation was approximated from the right ventricular activation in human hearts as described by Durrer et al. (1970 (link)). The stimulation current was applied at 50 μAmm⁻³ for 5 ms, first to the right ventricle activation region and then 33 ms later in the septum.