EP-PINNs: Cardiac Electrophysiology Characterisation Using Physics-Informed Neural Networks

Area of Science:

• Computational Biology
• Biophysics
• Medical Technology

Background:

• Accurate inference of electrophysiological (EP) tissue properties from action potential recordings is crucial for diagnosing and treating arrhythmias, such as atrial fibrillation.
• Current methods for EP parameter estimation are often limited by data requirements and complexity.
• Identifying localized pathologies like fibrosis, which contribute to arrhythmias, remains a challenge.

Purpose of the Study:

• To introduce EP-PINNs (Physics-Informed Neural Networks), a novel computational tool for accurate action potential simulation and EP parameter estimation.
• To demonstrate the capability of EP-PINNs in reconstructing spatio-temporal action potential evolution and predicting key EP parameters.
• To validate the utility of EP-PINNs in identifying tissue heterogeneities and characterizing drug effects on action potential duration (APD).

Main Methods:

• Development and application of Physics-Informed Neural Networks (EP-PINNs) for simulating electrophysiological activity.
• Utilizing 1D and 2D *in silico* data to train and validate EP-PINNs for action potential reconstruction and parameter prediction.
• Employing optical mapping data from *in vitro* biological preparations to assess EP-PINNs' effectiveness in characterizing anti-arrhythmic drug effects on APD.

Main Results:

• EP-PINNs successfully reconstructed spatio-temporal action potential evolution from sparse electrophysiological data.
• The tool accurately predicted parameters including action potential duration (APD), excitability, and diffusion coefficients.
• EP-PINNs demonstrated the ability to identify heterogeneities in EP properties and characterize drug-induced changes in APD.

Conclusions:

• EP-PINNs offer a powerful and accurate method for action potential simulation and electrophysiological parameter estimation.
• The tool's capability to detect tissue heterogeneities suggests potential for identifying localized pathologies like fibrosis.
• EP-PINNs show significant promise as a clinical tool for characterizing arrhythmias and guiding treatment strategies.