Skip to content

EP-PINNs implementation for 1D and 2D forward and inverse solvers for the Aliev-Panfilov cardiac electrophysiology model. Also includes Matlab finite-differences solver for data generation.

Notifications You must be signed in to change notification settings

meenakshi-mittal/EP-PINNs

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

EP-PINNs

Implementation of Physics-Informed Neural Networks (PINNs) for the solution of electrophysiology (EP) problems in forward and inverse mode. EP-PINNs is currently implemented with the Aliev-Panfilov EP model (Aliev & Panfilov, Chaos, 1996) and the monodomain equation in isotropic 1D and 2D settings. EP-PINNs use the DeepXDE library (Lu et al, SIAM, 2021).

Both forward and inverse modes provide a high-resolution spatio-temporal representation of the action potential. The inverse mode can additionally estimate global model parameters: a; b; D; a and D; b and D and spatial heterogeneities in D.

We also include Matlab files to simulate the cardiac (ventricular) Aliev-Panfilov electrophysiology model in a cable (1D) and rectangular (2D) domain with Neumann boundary conditions and several initial conditions (single ectopic focus, planar wave, spiral wave). The code also supports the inclusion of heterogeneities in D (which are rectangular, for the time being). Centred finite differences and a 4-step explicit Runge-Kutta solvers are used throughout. The Matlab code can be used to provide ground truth data to the EP-PINNs solver.

Further details can be found in a soon to be published article. In the meantime, email marta.varela (at) imperial.ac.uk with any comments or questions.

ExampleAlievPanfilov

About

EP-PINNs implementation for 1D and 2D forward and inverse solvers for the Aliev-Panfilov cardiac electrophysiology model. Also includes Matlab finite-differences solver for data generation.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • MATLAB 63.5%
  • Python 36.5%