CardioFit: A WebGL-Based Tool for Fast and Efficient Parameterization of Cardiac Action Potential Models to Fit User-Provided Data

TL;DR

This work tackles the challenge of rapidly identifying parameterizations for cardiac action potential models that reproduce data under diverse pacing conditions. It introduces CardioFit, a web-based tool that employs particle swarm optimization implemented in JavaScript and accelerated by WebGL to fit voltage time series and APD data across multiple cycle lengths. The approach supports several phenomenological AP models, an interactive interface for specifying data, models, and parameter bounds, and demonstrates accurate fittings to both model-derived and experimental datasets while providing insights into parameter roles and identifiability. The result is a fast, accessible platform that enables patient-specific or scenario-specific parameterizations without requiring local software installation, with practical impact for research and potential clinical translation.

Abstract

Cardiac action potential models allow examination of a variety of cardiac dynamics, including how behavior may change under specific interventions. To study a specific scenario, including patient-specific cases, model parameter sets must be found that accurately reproduce the dynamics of interest. To facilitate this complex and time-consuming process, we present an interactive browser-based tool that uses the particle swarm optimization (PSO) algorithm implemented in JavaScript and taking advantage of the WebGL API for hardware acceleration. Our tool allows rapid customization and can find low-error fittings to user-provided voltage time series or action potential duration data from multiple cycle lengths in a few iterations (10-32), corresponding to a runtime of a few seconds on most machines. Additionally, our tool focuses on ease of use and flexibility, providing a webpage interface that allows users to select a subset of parameters to fit, set the range of values each parameter is allowed to assume, and control the PSO algorithm hyperparameters. We demonstrate our tool's utility by fitting a variety of models to different datasets, showing how convergence is affected by model choice, dataset properties, and PSO algorithmic settings, and explaining new insights gained about the physiological and dynamical roles of the model parameters.

Figures (15)

• Figure 1: User interface of CardioFit. Top row: Buttons to initiate and save the results of CardioFit runs and a progress display for the iterations. Left: Fit (red) to the data (black) for the selected cycle length. Right: Interface for choosing the model, selecting the parameters to fit, setting the bounds, and viewing the fit parameter values. Bottom: Interface for adding datasets to fit.
• Figure 2: Detailed view of several elements of the CardioFit user interface. (a) APD data entry interface for CardioFit. (b) Square stimulus parameters for CardioFit. (c) Biphasic stimulus parameters for CardioFit. (d) The interface to configure the PSO algorithm hyperparameters in CardioFit.
• Figure 3: MS model fit to voltage data obtained from the same model. Cycle lengths of 500ms, 400ms, and 300ms were fit simultaneously to the two action potentials shown. Reference data is plotted in black; the results of 20.0 separate fits using CardioFit are plotted in various colors. Fits were generated using 4096.0 particles, 100.0 iterations, and four pre-recording stimuli in all cases.
• Figure 4: FK model fit to voltage data obtained from the same model. Cycle lengths of 500, 400, and 300ms were fit simultaneously to the two action potentials shown. Reference data is plotted in black; the results of 20.0 separate fits using CardioFit are plotted in various colors. Fits were generated using 4096.0 particles, 100.0 iterations, and four pre-recording stimuli in all cases.
• Figure 5: BOCF model fit to voltage data obtained from the same model. Cycle lengths of 500, 400, and 300ms were fit simultaneously to the two action potentials shown. Reference data is plotted in black; the results of 20.0 separate fits using CardioFit are plotted in various colors. Fits were generated using 4096.0 particles, 100.0 iterations, and four pre-recording stimuli in all cases.