This repository contains code for the paper Physics-informed neural networks for tissue elasticity reconstruction in magnetic resonance elastography which is to be presented at MICCAI 2023.
Run the following to setup the conda environment and register it as a Jupyter notebook kernel:
mamba env create --file=environment.yml
mamba activate MRE-PINN
python -m ipykernel install --user --name=MRE-PINNThis notebook downloads the BIOQIC simulation data set and trains PINNs to reconstruct a map of shear elasticity from the displacement field.
The notebook takes roughly 2.5 h to train for 100,000 iterations on an RTX 5000 and uses 2.5 GiB of GPU memory.