Aceleração de simulações de arritmias cardíacas através de métodos numéricos baseados no esquema Alternating Direction Implicit e computação paralela

Abstract

Cardiac arrhythmia is a condition related to changes in the rhythm of the heartbeat from heart. It happens when electrical stimuli do not propagate correctly by the myocardium and can lead to sudden death. Computational modeling can be a important ally in identifying serious arrhythmias, in addition to enabling virtual tests personalized for the patient. In this context, the monodomain, a mathematical model that describes electrical propagation in cardiac tissue, is widely used. Solve it, in However, it is still a complex and time-consuming task. This work presents strategies numerical and computational techniques to accelerate simulations of ventricular arrhythmias. They are compared three numerical methods based on the Alternating Direction Implicit scheme (ADI): Operator Splitting ADI, Second-Order Semi-Implicit ADI and Mixed-Order SemiImplicit ADI. The implementations are parallelized using the OpenMP library, and four metrics are analyzed: the window of vulnerability, used to assess the risk of arrhythmia in the heart; the speed of propagation of the stimulus; the numerical error; and the time of execution. Furthermore, an alternative approach is presented with an implementation parallel for GPUs using the CUDA library which is compared to the MonoAlg3D simulator.