Miniaturizing High-performance Computation of 3D Cardiac Dynamics onto Personal Computers, Tablets, and Even Cell-phone Browsers Using Abubu.js

Abouzar Kaboudian1, Elizabeth Cherry2, flavio fenton3
1Georgia Institute of Technology, 2Rochester Institute of Technology, 3georgia tech


Cardiac disease is the leading cause of death in the industrialized world and globally. Numerical simulations have proven to be of great value in the study of deadly arrhythmias. However, a major bottleneck in numerical simulations has always been the computational costs of the mathematical models of the electrophysiology of cardiac cells and tissue connected cells, requiring solving billions to trillions of ordinary differential equations (ODEs) for each second of dynamics in two- and three-dimensional space. Traditionally, high-performance computing on supercomputer clusters was used to address this challenge. In this work, we will present how we can use Abubu.js library to offload high-performance schemes to personal computers' graphical processing units (GPUs) to achieve interactive real-time simulations of cardiac dynamics in two and three-dimensional settings effectively solving billions of ODEs on a personal computer such as a laptop or even a cell-phone. We will use the classical Beeler-Reuter model on a three-dimensional rabbit ventricular structure to show how we can quickly change some tunable parameters in this model to change the dynamics of arrhythmia from ventricular tachycardia into ventricular fibrillation. We will interactively introduce scars on the tissue and demonstrate how anatomical reentry can form in this model. All simulations that would naturally require a supercomputer will be performed in real-time and interactively in the presentation. We demonstrate how this approach can change the way we pose questions, solve problems, design our numerical studies, and even save our simulation results into just kilobytes of setup files instead of gigabytes of time-series.