Simulation of clinical electrophysiology in 3D human atria: a high‐performance computing and high‐performance visualization application

Abstract

SUMMARYAtrial fibrillation (AF) is a common cardiac disease of genuine clinical concern with high rates of morbidity, leading to major personal and National Health Service costs. Computer modelling of AF using biophysically detailed cellular models with realistic 3D anatomical geometry allows investigation of the underlying ionic mechanisms in far more detail than with experimental physiology. We have developed a 3D virtual human atrium that combines detailed cellular electrophysiology including ion channel kinetics and homeostasis of ionic concentrations with anatomical details. The segmented anatomical structure and the multivariable nature of the system make the 3D simulations of AF computationally large and intensive. * Computational demands are such that a full problem-solving environment requires access to resources of high-performance computing (HPC), high-performance visualization (HPV), remote data repositories and backend infrastructure. This is a classic example of eScience and Grid-enabled computing. This study was carried out using multiple processor shared memory systems and massively parallel distributed memory systems. With the envisaged increase in anatomical and molecular detail in our cardiac models the requirement for HPC resources is predicted to increase many fold (∼ 1-10 teraflops). Distributed computing is essential, both through massively parallel systems (a single supercomputer) and multiple parallel systems made accessible through the Grid. Analysis and interpretation of results are enhanced by HPV, which in itself is a large data computing aspect of cardiac modelling.

Effects of elevated Homocysteine hormone on electrical activity in the human atrium: A simulation study

“…Distributed memory (MPI) parallalisation was implemented in the spatial simulations, especially in the 2D and 3D cases with formalisms developed previously [13].…”

“…The 3D model has a spatial resolution of 0.33 mm × 0.33 mm × 0.33mm. The model was numerically solved by the explicit Euler method with a time step 0.005 ms, which guarantees sufficiently accurate solutions [13]. Using similar stimulation protocols as used in a previous study [13], reentrant excitation scroll waves were initiated for both Control and elevated Hcy (500 µmol/L) conditions.…”

Computational assessment of arrhythmia potential in the heterogeneously perfused ventricle

“…In this study, the rectangular geometries that were used lend themselves to a straightforward box partitioning. It is known that realistic geometries occupy less than 30% of the total volume [34], which often leads to uneven load on processors in distributed memory machines. In addition, a larger amount of memory is required for full 3D simulations.…”

“…In addition, a larger amount of memory is required for full 3D simulations. To optimise memory use, a method for mapping of the dynamical system variables onto a 1D array was developed previously [34]. To further improve both the load balancing as well as memory use, graph theory based partitioning will be implemented in the future [35].…”

Estudio de la vulnerabilidad a reentradas a través de modelos matemáticos y simulación de la aurícula humana

“…El modelo anatómico más recientemente desarrollado es el de Sachse et al (2000), este a su vez obtenido del proyecto Humano Visible y referido por (Kharche et al 2008, Seemann et al 2006) y se diferencia de nuestro modelo en que no se consideró el seno coronario ni las uniones a la aurícula izquierda. En cuanto a las fibras solo se indican tres direcciones diferentes, frente a las 42 regiones definidas en nuestro modelo.…”