Deep Cascade of Convolutional Neural Networks for MR Image Reconstruction
Reconstruct MR images from its undersampled measurements using Deep Cascade of Convolutional Neural Networks (DC-CNN). This repository contains the implementation of DC-CNN using Theano and Lasagne and the simple demo. Note that the library requires the dev version of Lasagne and Theano, as well as pygpu backend for using CUFFT Library. Some of the toy dataset borrowed from <http://mridata.org>.
1. 2D Reconstruction
Usage:
python main_2d.py --num_epoch 5 --batch_size 2
2. Dynamic Reconstruction
Reconstruct dynamic MR images from its undersampled measurements using DC-CNN with Data Sharing layer. Note that the library requires CUDNN in addition to the requirement specified above.
Usage:
python main_3d.py --acceleration_factor 4
Citation and Acknowledgement
If you use the code for your work, or if you found the code useful, please cite the following works.
2D Reconstruction:
Schlemper, J., Caballero, J., Hajnal, J. V., Price, A., & Rueckert, D. A Deep Cascade of Convolutional Neural Networks for MR Image Reconstruction. Information Processing in Medical Imaging (IPMI), 2017
The paper is also available on arXiv: <https://arxiv.org/pdf/1703.00555.pdf>
Dynamic Reconstruction:
Schlemper, J., Caballero, J., Hajnal, J. V., Price, A., & Rueckert, D. A Deep Cascade of Convolutional Neural Networks for Dynamic MR Image Reconstruction. ArXiv 1704.02422
The paper is also available on arXiv: <https://arxiv.org/pdf/1704.02422.pdf>