Repository with code to reproduce the results for SMUG in our paper and our upcoming journal version
This repository provides code to reproduce the results from the SMUG method for robust MRI reconstruction, as described in our paper. SMUG systematically integrates Regularization by Smoothing (RS) with MoDL using a deep unrolled architecture. It addresses the instabilities of MoDL by optimizing where to apply RS in the unrolled architecture and introduces a novel unrolling loss to enhance training efficiency.
- Robust MRI Reconstruction: Enhances MoDL performance through systematic integration of RS.
- Deep Unrolled Architecture: Utilizes unrolling techniques for improved efficiency.
- Instability Mitigation: Significantly reduces three major types of instabilities in MoDL.
- Clone the repository:
git clone https://github.com/sjames40/SMUG_journal.git
2 Install the required dependencies: '''bash pip install -r requirements.txt
3 Usage Download the dataset from Dropbox: Data avaliable on https://www.dropbox.com/scl/fi/801dxovhbkp2bkl2krz5x/NEW_KSPACE.zip?rlkey=4u3b32f6c4pfujsv3kp7z5bdk&st=hwe9thrv&dl=0
4 Refer to the train_SMUG.py script for training the SMUG model. Additional scripts such as test.py and train_RSE2E.py provide testing and alternative training routines.
Directory Structure data/: Contains datasets for MRI reconstruction tasks. models/: Model architectures for SMUG and related experiments. options/: Configuration files for different experiments. util/: Utility scripts for data handling and model evaluation.