Skip to Main content Skip to Navigation
Conference papers

Feasibility of Whole-Heart Electrophysiological Models With Near-Cellular Resolution

Mark Potse 1, 2, 3 Emmanuelle Saillard 4 Denis Barthou 4, 5 Yves Coudière 2, 3, 1
2 CARMEN - Modélisation et calculs pour l'électrophysiologie cardiaque
IHU-LIRYC, IMB - Institut de Mathématiques de Bordeaux, Inria Bordeaux - Sud-Ouest
4 STORM - STatic Optimizations, Runtime Methods
LaBRI - Laboratoire Bordelais de Recherche en Informatique, Inria Bordeaux - Sud-Ouest
Abstract : Given the opportunity to use a new cluster computer with over a quarter million compute cores we tested the strong and weak scaling of a monodomain reaction-diffusion model of the human ventricles with Ten Tusscher-Panfilov dynamics. Element sizes down to 25 µm and a model size up to 11 billion nodes were tested with both explicit and implicit Euler integration methods. Time steps were 0.01 ms for the implicit method and were resolution-dependent for the explicit method. We found that the weak scaling (increasing model size) was good for both methods. Depending on the model size, strong scaling (speedup at a larger number of cores) was satisfactory for the explicit method, and more limited for the implicit method. The implicit solver was generally slower; only at a resolution of 25 µm and on a relatively small number of cores it was as fast as the explicit solver. We conclude that whole-heart simulations at 25 µm resolution are technically feasible, although not practical yet on currently available systems.
Complete list of metadatas

Cited literature [10 references]  Display  Hide  Download

https://hal.inria.fr/hal-02943513
Contributor : Mark Potse <>
Submitted on : Saturday, September 19, 2020 - 6:53:42 PM
Last modification on : Thursday, October 8, 2020 - 8:10:37 PM

File

126_CinCFinalPDF.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02943513, version 1

Collections

Citation

Mark Potse, Emmanuelle Saillard, Denis Barthou, Yves Coudière. Feasibility of Whole-Heart Electrophysiological Models With Near-Cellular Resolution. CinC 2020 - Computing in Cardiology, Sep 2020, Rimini / Virtual, Italy. ⟨hal-02943513⟩

Share

Metrics

Record views

49

Files downloads

73