Ricardo Silva Campos

A GPU-based heart simulator with mass-spring systems and cellular automaton

This work proposes an electro-mechanical simulator of the cardiac tissue, so that its main feature is the low computational cost. This is necessary to run real-time simulations and on the fly applications. In order to achieve this, we used cellular automata and mass-spring systems to model the cardiac behavior, and furthermore parallelize the code to run in graphics processing unit (GPU) with compute unified device architecture. Sequentially, our simulator was quite faster than traditional partial differential equations simulators. In addition, we performed different load tests to evaluate our code behavior in GPUs, and spotted its potentials and bottlenecks.

An Electro-Mechanical Cardiac Simulator Based on Cellular Automata and Mass-Spring Models

The mechanical behavior of the heart is guided by the propagation of an electrical wave, called action potential. Many diseases have multiple effects on both electrical and mechanical cardiac physiology. To support a better understanding of the multiscale and multiphysics processes involved in physiological and pathological cardiac conditions, a lot of work has been done in developing computational tools to simulate the electro-mechanical behavior of the heart. In this work, we propose a new user-friendly and efficient tool for the electro-mechanical simulation of the cardiac tissue that is based on cellular automata and mass-spring models. The proposed tool offers a user-friendly interface that allows one to interact with the simulation on-the-fly. In addition, the simulator is parallelized with CUDA and OpenMP to further speedup the execution time of the simulations.

Comparison between genetic algorithms and differential evolution for solving the history matching problem

This work presents a performance comparison between Differential Evolution (DE) and Genetic Algorithms (GA), for the automatic history matching problem of reservoir simulations. The history matching process is an inverse problem that searches a set of parameters that minimizes the difference between the model performance and the historical performance of the field. This model validation process is essential and gives credibility to the predictions of the reservoir model. Four case studies were analyzed each of them differing on the number of parameters to be estimated: 2, 4, 9 and 16. Several tests are performed and the preliminary results are presented and discussed.

Uniformization Method for Solving Cardiac Electrophysiology Models Based on the Markov-Chain Formulation

This paper compares different numerical methods for the solution of myocyte models of cardiac electrophysiology. In particular, it presents how the technique called uniformization method substantially increases the stability of simple first-order methods such as Euler explicit method and Rush-Larsen (RL) method, for the solution of modern electrophysiology models that are based on continuous-time Markov chains (MCs) for the description of subcellular structures, such as ion channels. The MCs are often associated with stiff ordinary differential equations that severely limit the time step used by these traditional methods. By using the uniformization method, we could significantly increase the time steps for the solution of different cardiac electrophysiology models and improve the computational performance up to 150 times compared to the performance of Eulers and RLs methods.

3D Heart Modeling with Cellular Automata, Mass-Spring System and CUDA

The mechanical behavior of the heart is guided by the propagation of an electrical wave, called action potential. Many diseases have multiple effects on both electrical and mechanical cardiac physiology. To support a better understanding of the multi-scale and multi-physics processes involved in physiological and pathological cardiac conditions, a lot of work has been done in developing computational tools to simulate the electro-mechanical behavior of the heart. In this work, we implemented an aplication to mimic the heart tissue behavior, based on cellular automaton, mass-spring system and parallel computing with CUDA. Our application performed 3D simulations in a very short time. In order to assess the simulation results, we compared them with another synthetic model based on well-known partial differential equationsPDE. Preliminary results suggest our application was able to reproduce the PDE results with much less computational effort.

Adaptive Time Step for Cardiac Myocyte Models

Abstract The modeling of the electrical activity of the heart is of great medical and scientific interest as it provides a way to better understand the underlying biophysical phenomena, supports the development of new techniques for diagnoses and serves as a platform for drug tests. At cellular level, the electrical activity of cardiac myocytes may be simulated by solving a system of ordinary di_erential equations (ODEs) describing the electrical behavior of the cell membrane. Because the biophysical processes underlying this phenomenon are non-linear and change very rapidly, the ODE system is challenging to solve numerically. Furthermore, the implementation of these models is a hard task. This paper presents a tool to describe models using Ordinary Differential Equations. It is based on CellML standard and automatically generates C++ source-code, with numerical methods to solve the models equations. The aim of this work is to present a numerical method with adaptive time step based on the Euler Method and Second Order Runge-Kutta method. The proposed method accelerated the execution and kept numerical errors under control. Preliminary results suggest this adaptive method is up to 25 times faster than the explicit Euler method with fixed time step.

Approaching cardiac modeling challenges to computer science with CellML-based web tools

Cardiac modeling is being used in a variety of ways to support the tests of new drugs, the development of new medical devices and of non-invasive diagnostic techniques. Computer models have become valuable tools for the study and comprehension of the complex phenomena of cardiac electrophysiology. However, the complexity, the multi-scale and multi-physics nature of cardiac modeling still restrict its use to a few specialized research centers in the world. In addition, the issue of sharing and re-using such models has proven to be particularly problematic, with published models often lacking information that is required to accurately reproduce published results. In this work, with the aim of tackling the aforementioned issues, we present a web portal that provides support for cardiac electrophysiology modeling. This framework integrates different computer tools and allows one to bypass many complex steps during the development and use of cardiac models. The process of model development is supported by a Web-based editor for CellML, a mark-up language dedicated to the description of biological structures, processes and the associated models. The implementation of the cardiac cell models is automatically provided by a code generator that translates models described in CellML language to executable code that allows one to manipulate and solve the models numerically. The set up and use of the simulator is supported by a user-friendly graphical interface that offers the tasks of simulation configuration and execution, storage of results and basic visualization. All the tools are integrated in a Web Portal. As a result, the complex techniques and the know-how behind cardiac modeling are all taken care of by the web distributed applications.

A Parallel Genetic Algorithm to Adjust a Cardiac Model Based on Cellular Automaton and Mass-Spring Systems

This work presents an electro-mechanical model of the cardiac tissue and an automatic way to tune its parameters. A cellular automaton was used to simulate the action potential propagation, and a mass-spring system to model the tissue contraction. A parallel genetic algorithm was implemented in order to automatically adjust a simple and fast discrete model, to reproduce simulations of another synthetic well known model based on differential equations DEs. Our results suggest that the discrete model was able to qualitatively reproduce the results obtained by DEs with much less computational effort.

Performance Evaluation of a Human Immune System Simulator on a GPU Cluster

The Human Immune System HIS is a complex system that protects the body against several diseases. Some aspects of such complex system can be better understand with the use of mathematical and computational tools. Huge computational resources are required to execute simulations of the HIS, so the use of parallel environments is mandatory. This work presents a parallel implementation of a 3D HIS simulator on a GPU cluster that uses CUDA, OpenMP and MPI to speedup the execution of the application. A performance evaluation is then carried out, and the impact of the use of InfiniBand, a low latency network, and GPUs Error-Correcting Code ECC are measured. Speedups upi¾?to 956 were obtained by the parallel version that uses Infiniband and turns off ECC.

Comparing high performance techniques for the automatic generation of efficient solvers of cardiac cell models

In silico experiments have been used for a better understanding of the electrical activity of cardiac myocytes, usually via models based on nonlinear systems of ordinary differential equations. Many different models for cardiac myocytes are available that vary on the level of complexity, depending on how detailed the phenomena is described. Long simulations of realistic and complex models are computationally expensive. To cope with this problem, this work compares different techniques to automatically speed up the numerical solution of cardiac models: (a) adaptive time step method, (b) Partial Evaluation (PE) and Lookup Tables (LUTs), and (c) an automatic way to find and exploit code concurrency via OpenMP directives. All the techniques were implemented as part of an automatic code generator for the numerical solution of models that are described in the CellML markup language. Experimental results demonstrated that the adaptive time step simulations were up to 32 times faster than the traditional Euler that use fixed time step. Combined with parallel computing on a multicore processor the execution time was further decreased and simulations were 41 times faster. Finally, the LUTs and PE techniques resulted in a 117-fold improvement in computation time over the Euler method and 72-fold improvement when compared to the traditional Rush–Larsen method.