fsahli/FiberNet

Physics-informed neural networks to learn cardiac fiber orientation from multiple electroanatomical maps

Carlos Ruiz Herrera*, Thomas Grandits*, Gernot Plank, Paris Perdikaris, Francisco Sahli Costabal, Simone Pezzuto

Project homepage: https://fsahli.github.io/research/fibernet.html
arXiv: https://arxiv.org/abs/2201.12362

2D example 3D example

This repository contains a demo implementation of our paper Physics-informed neural networks to learn cardiac fiber orientation from multiple electroanatomical maps. It contains a few 2D and 3D examples showcasing how FiberNet works and can optimize fiber orientations from multiple electroanatomical maps. For more technical information on the approach, please have a look at the project page or the arxiv paper.

Installation

All examples are provided in the form of jupyter-notebooks. To run, makes sure you have python3 installed and then run

pip install -r requirements.txt

from your command line. The notebooks can then be easily accessed by starting the jupyter-server via the command

jupyter-notebook

Provided Examples

• 2D_example

  Reconstructs the fiber orientation and velocity of two piecewise constant regions on a square

• 2D_example_aniso

  Same as above, but considers different levels of anisotropy between the two regions

• 3D_example

  Reconstructs the fiber orientation of a rule-based in-silico left atrial model for randomly paced pacing and measurement locations

• 3D_example_CS

  Same as above, but considers only the approximate Bachmann bundle and coronary sinus locations as pacing loactions

Citation

@article{ruiz_herrera_physics_informed_2022,
	title = {Physics-informed neural networks to learn cardiac fiber orientation from multiple electroanatomical maps},
	issn = {1435-5663},
	url = {https://doi.org/10.1007/s00366-022-01709-3},
	doi = {10.1007/s00366-022-01709-3},
	language = {en},
	urldate = {2022-07-22},
	journal = {Engineering with Computers},
	author = {Ruiz Herrera, Carlos and Grandits, Thomas and Plank, Gernot and Perdikaris, Paris and Sahli Costabal, Francisco and Pezzuto, Simone},
	month = jul,
	year = {2022},
	keywords = {Anisotropic conduction velocity, Cardiac electrophysiology, Cardiac fibers, Deep learning, Eikonal equation, Physics-informed neural networks},
}