MRIDC is a toolbox for applying AI methods on MR imaging. A collection of tools for data consistency and data quality is provided for MRI data analysis. Primarily it focuses on the following tasks:

The following models are implemented for accelerated MRI reconstruction: 1.Cascades of Independently Recurrent Inference Machines (CIRIM) , 2.Compressed Sensing (CS), 3.Convolutional Recurrent Neural Networks (CRNN), 4.Deep Cascade of Convolutional Neural Networks (CCNN), 5.Down-Up Net (DUNET), 6.End-to-End Variational Network (E2EVN), 7.Joint Deep Model-Based MR Image and Coil Sensitivity Reconstruction Network (Joint-ICNet) , 8.Independently Recurrent Inference Machines (IRIM), 9.KIKI-Net, 10.Learned Primal-Dual Net (LPDNet), 11.MultiDomainNet, 12.Recurrent Inference Machines (RIM) , 13.Recurrent Variational Network (RVN), 14.UNet, 15.Variable Splitting Network (VSNet), 16.XPDNet, 17.and Zero-Filled reconstruction (ZF).

The following models are implemented for quantitative imaging: 1.quantitative Cascades of Independently Recurrent Inference Machines (qCIRIM) , 2.quantitative End-to-End Variational Network (qE2EVN) , 3.quantitative Independently Recurrent Inference Machines (qIRIM), 4.quantitative Recurrent Inference Machines (qRIM) .

Note: Currently only the above models are implemented. More models can be added by extending the reconstruction models for quantitative imaging. If you wish to extend the toolbox, please open an issue.

Coming soon...

Check the projects page for more information of how to use mridc.

MRIDC is best to be installed in a Conda environment.

conda create -n mridc python=3.9
conda activate mridc

Use pip installation if you want the latest stable version.

pip install mridc

Use source installation if you want the latest development version, as well as for contributing to MRIDC.

git clone https://github.com/wdika/mridc
cd mridc
./reinstall.sh

Access the API Documentation here

MRIDC is based on the NeMo framework, using PyTorch Lightning for feasible high-performance multi-GPU/multi-node mixed-precision training.

For the reconstruction methods:

• the implementations of 6 and 14 are thanks to and based on the fastMRI repo.
• The implementations of 7, 9, 10, 11, 13, and 16 are thanks to and based on the DIRECT repo.

Please cite MRIDC using the "Cite this repository" button or as

@misc{mridc,
    author = {Karkalousos Dimitrios and Caan Matthan},
    title = {MRIDC: Data Consistency for Magnetic Resonance Imaging},
    year = {2022},
    url = {https://github.com/wdika/mridc},
}

The following papers use the MRIDC repo:

[1] Karkalousos, D. et al. (2021) ‘Assessment of Data Consistency through Cascades of Independently Recurrent Inference Machines for fast and robust accelerated MRI reconstruction’

[2] Zhang C, Karkalousos D, Bazin PL, Coolen BF, Vrenken H, Sonke JJ, Forstmann BU, Poot DHJ, Caan MWA. ‘A unified model for reconstruction and R2* mapping of accelerated 7T data using the quantitative Recurrent Inference Machine’