Abstract
Ionic models of cardiac action potentials (APs) may not reproduce all relevant datasets using their default settings, and tuning parameter values to improve fits is often difficult. To facilitate this task, we present CardioFit, a tool to fit cardiac AP model parameters to time-series data using particle swarm optimization (PSO). CardioFit quickly finds conductance parameter values for detailed human ventricular models, including those of ten Tusscher et al. (2006) and O'Hara et al., that match experimental data, within the capabilities of the models. CardioFit is implemented as a web-based tool using JavaScript and the WebGL graphics API, allowing PSO to take advantage of any available graphics-processing unit hardware to run in parallel. As the PSO algorithm requires the simultaneous evaluation of many candidate parameterizations when searching for the best fit, this method is well-suited to large-scale parallelism. Due to its fast parallel implementation, CardioFit obtains conductance parameters of detailed ionic models to match a given dataset in a few minutes on consumer-grade hardware, even though tens of thousands of model runs typically are required.