fastmri-challenge

fastmri-challenge

Image reconstruction from single-coil MRI data with subsampling mask according to the fastMRI challenge (2019).

How to run this experiment

1. Check (and modify if necessary) the configuration file config.py. It specifies the directory paths for the data and results, as well as all relevant parameters of the experimental setup. By default, the data is stored in the subdirectory raw_data and results and model weights are stored in the subdirectory results.
2. Download the fastMRI Knee-MRI dataset and place it in the data folder specified in the configuration.
3. Prepare the data by running data_management.py.
4. Train networks using the scripts named script_train_*.py.
5. Determine the regularization parameters of the total variation minimization reconstruction method by running script_grid_search_l1.py.
   *Remark: This script is designed to be run in parallel for multiple noise levels, making use of the batch-job capabilities of Sun Grid Engine cluster computing. You can adapt this to run sequentially, but be aware that this will be slow. *
   Collect the results of the grid search by calling the combine_results() function from the grid search script.
6. Check (and modify if necessary) the configuration files config_robustness.py. It specifies the relevant parameters for the robustness analysis. Adapt the list of networks to compare according to the ones that you have actually trained.
7. Analyze different aspects of the robustness comparisons by running the scripts named script_robustness_*.py.