Cherry Elizabeth M, Greenside Henry S, Henriquez Craig S
Department of Physics and Astronomy, Hofstra University, Hempstead, NY 11549, USA.
Chaos. 2003 Sep;13(3):853-65. doi: 10.1063/1.1594685.
A recently developed space-time adaptive mesh refinement algorithm (AMRA) for simulating isotropic one- and two-dimensional excitable media is generalized to simulate three-dimensional anisotropic media. The accuracy and efficiency of the algorithm is investigated for anisotropic and inhomogeneous 2D and 3D domains using the Luo-Rudy 1 (LR1) and FitzHugh-Nagumo models. For a propagating wave in a 3D slab of tissue with LR1 membrane kinetics and rotational anisotropy comparable to that found in the human heart, factors of 50 and 30 are found, respectively, for the speedup and for the savings in memory compared to an algorithm using a uniform space-time mesh at the finest resolution of the AMRA method. For anisotropic 2D and 3D media, we find no reduction in accuracy compared to a uniform space-time mesh. These results suggest that the AMRA will be able to simulate the 3D electrical dynamics of canine ventricles quantitatively for 1 s using 32 1-GHz Alpha processors in approximately 9 h.
一种最近开发的用于模拟各向同性一维和二维可兴奋介质的时空自适应网格细化算法（AMRA）被推广到用于模拟三维各向异性介质。使用Luo-Rudy 1（LR1）模型和FitzHugh-Nagumo模型，针对各向异性和非均匀的二维及三维区域研究了该算法的准确性和效率。对于具有与人类心脏中发现的旋转各向异性相当的LR1膜动力学的三维组织平板中的传播波，与在AMRA方法的最精细分辨率下使用均匀时空网格的算法相比，分别在加速和内存节省方面发现了50倍和30倍的因子。对于各向异性的二维和三维介质，我们发现与均匀时空网格相比，准确性没有降低。这些结果表明，使用32个1-GHz Alpha处理器，AMRA将能够在大约9小时内定量模拟犬心室的三维电动力学1秒。
