APHYN-EP : Deep Learning for Model Correction in CardiacElectrophysiological Imaging

Test code associated with article accepted for the conference MIDL 2022 by Victoriya Kashtanova, Ibrahim Ayed, Andony Arrieula, Mark Potse, Patrick Gallinari and Maxime Sermesant.

Getting Started

Prerequisites

• Linux or macOS
• Python 3.7+
• CPU or NVIDIA GPU + CUDA CuDNN

Installation

• Clone this repo:

git clone https://github.com/Inria-Asclepios/APHYN-EP
cd APHYN-EP

Requirements

To run the code within this repository requires Python 3.7+ with the following dependencies

• torch
• torchdiffeq
• and some standard python libraries matplotlib, numpy, scipy etc.

which can be installed via

$ pip install -r requirements.txt

APHYN-EP train

Try :

python train.py --name aphynep --dataroot ./data_ttp/ --batch_size 4 --estim_param_names d,t_in

Data

To evaluate APHYN-EP framework, we used a dataset of transmembrane potential activation simulated with a monodomain reaction-diffusion equation and the Ten Tusscher – Noble – Noble –Panfilov ionic model (Ten Tusscher et al., 2004), which represents 12 different transmem-brane ionic currents. The simulations were performed with a recent version of the propag-5 software (Krause et al., 2012; Potse, 2018) and added into folder data_ttp.

You can use an open source package Finitewave, if you want to simulate more data with the same properties or/and with more complex geometries of cardiac tissue.