Example codes and data for motion-robust T2 mapping using MOST-DL
Author: Qinqin Yang (qqyang@stu.xmu.edu.cn)
This project relies on SPROM or MRiLab, the full version can be downloaded from zenodo platform:
MOdel-based SyntheTic Data-driven Learning (MOST-DL): Application in Single-shot T2 Mapping with Severe Head Motion Using Overlapping-echo Acquisition
arXiv: https://arxiv.org/abs/2107.14521
IEEE Xplore: https://doi.org/10.1109/TMI.2022.3179981
Table of Conents
- Dependencies
- CNN1_ParallelRecon
- CNN2_T2mapping
- ParametricTempGen
- SyntheticDataGen
- Pipeline
- Reference
MOST-DL:
The deep learning codes have been tested in Anaconda 3.6 with Pytorch 1.6.0-cuda. The training data is in the mat format.
The pre-processing codes have been tested in MATLAB R2019a.
The codes in this directory were used to deep learning parallel reconstruction.
bingo_train.py: This is the training code, you'd better run it in PyCharm IDE.
bingo_test.py: This is the testing code, make sure the trained model has been included in the ./models/ directory.
network/ResUNet.py: Network architecture of CNN1.
The codes in this directory were used to deep learning T2 mapping with motion correction.
bingo_train.py: This is the training code, you'd better run it in PyCharm IDE. Note that some techniques were applied for accelerating the training (refer to blog).
bingo_test.py: This is the testing code, make sure the trained model has been included in the ./models/ directory.
network/UNet.py: Network architecture of CNN2.
Example codes for generating MRI parametric templates (T2 and M0) from IXI database
datapro_IXI2map.m: Convert multi-contrast MRI images to quantitative parametric maps.
datapro_map2**.m: Convert parametric maps to templates (virtual object, VObj) for SPROM/MRiLab simulation.
show_**.m: Show the VObj files.
multicoil_data_MOLED.mat: A motion-corrupted multi-coil real data for evaluating the pipeline.
main.m: This is a Python-Matlab code to compare the conventional method and MOST-DL, make sure the trained models (see full version) have been included in the ./models/ directory.
Example codes for pre-processing the outputs from SPROM or MRiLab, the Bloch-based motion-corrupted data is generated from SPROM. A new version of MRiLab for batch processing is being polished and will be released soon.
CoilmapGen.m: Generating coil sensitivity maps from ACS data using ESPIRiT algorithm.
scan2dataset_**.m: Processing the outputs of SPROM/MRiLab software for deep learning.
OLED: C. Cai, C. Wang, Y. Zeng, S. Cai, D. Liang, Y. Wu, Z. Chen, X. Ding, and J. Zhong, “Single-shot T-2 mapping using overlapping-echo detachment planar imaging and a deep convolutional neural network,” Magn. Reson. Med., vol. 80, no. 5, pp. 2202-2214, Nov, 2018. doi: 10.1002/mrm.27205.
MOLED: J. Zhang, J. Wu, S. Chen, Z. Zhang, S. Cai, C. Cai, and Z. Chen, “Robust single-shot T2 mapping via multiple overlapping-echo acquisition and deep neural network,” IEEE Trans. Med. Imag., vol. 38, no. 8, pp. 1801-1811, Aug, 2019. doi: 10.1109/tmi.2019.2896085.
SPROM: C. Cai, M. Lin, Z. Chen, X. Chen, S. Cai, and J. Zhong, “SPROM - an efficient program for NMR/MRI simulations of inter- and intra-molecular multiple quantum coherences,” C. R. Phys., vol. 9, no. 1, pp. 119-126, Jan, 2008. doi: 10.1016/j.crhy.2007.11.007.
MRiLab: F. Liu, J. V. Velikina, W. F. Block, R. Kijowski, and A. A. Samsonov, “Fast realistic MRI simulations based on generalized multi-pool exchange tissue model,” IEEE Trans. Med. Imag., vol. 36, no. 2, pp. 527-537, Feb, 2017. doi: 10.1109/tmi.2016.2620961.
Please contact us if you have any questions about our work. We hope that synthetic data can help you solve challenging problems in MRI and other fields.