Skip to content

README.md

Latest commit

 

History

History
58 lines (41 loc) · 4.13 KB

README.md

File metadata and controls

58 lines (41 loc) · 4.13 KB
Raw

Low_Field_MRI_Recon

Accelerating Low-field MRI: From Compressed Sensing to Deep Learning Reconstruction with CNNs and Transformers

Authors: David Waddington, Efrat Shimron, Shanshan Shan, Neha Koonjoo.

Public repository with data and code supporting Shimron et al. 2024. This work investigates and compares leading compressed sensing and AI-based methods for image reconstruction at ultra-low magnetic fields. A preprint of the manuscript is available at https://arxiv.org/abs/2411.06704 .

Reconstruction accuracy across field strengths

Key Files

The following Python scripts generate the figures in the manuscript:

  • Fig2_fast_mri_sampling_experiment.py: Figure 2
  • Fig3and4_fast_mri_noise_experiment_complete-R2_R4.py: Figures 3 and 4
  • Fig5_phantom_experiment_plot.py: Figure 5
  • Fig6_retro_ulf_3T_compare.py: Figure 6
  • Fig7_prospective_recon-phantom-brain.py: Figure 7
  • FigS3_fast_mri_noise_curves_all.py: Supplementary Figure S3
  • FigS4_dataset_size.py: Supplementary Figure S4
  • FigS6_retro_3T_otherseqs.py: Supplementary Figure S6
  • FigS7_nonoise_sweep.py: Supplementary Figure S7

The following Python modules are shared dependencies:

  • automap_fns.py: Applies AUTOMAP to k-space data.
  • display_fns.py: Displays images as subplots comparing reconstruction methods.
  • metrics.py: Calculates image quality metrics (NRMSE, SSIM, BRISQUE, Tenengrad).
  • snr_calc.py: SNR estimation utilities.
  • ulf_recon_fns.py: Masks fully-sampled datasets and performs IFFT and CS reconstruction.
  • unrolling_fns.py: Applies Unrolled network reconstruction to k-space data.
  • unet_fns.py: Applies UNet reconstruction to k-space data.
  • swin_fns.py: Applies Swin Cascade reconstruction.

Training scripts for all models are provided in training_scripts/.

A requirements.txt file details the pip packages required to run the figure scripts.

Data access

For access to raw data, please contact the corresponding author to ensure compliance with IRB requirements.

Further Unrolled reconstruction code is available here: https://github.com/shanshanshan3/DCReconNet Further AUTOMAP code is available here: https://github.com/MattRosenLab/AUTOMAP

Acknowledgements

Reconstruction approaches were adapted from code associated with the following publications:

  1. F. Ong, M. Lustig, SigPy: a python package for high performance iterative reconstruction. 27th Annual Meeting of the International Society of Magnetic Resonance in Medicine (2019), p. 4819.
  2. S. Shan, Y. Gao, P. Z. Y. Liu, B. Whelan, H. Sun, B. Dong, F. Liu, D. E. J. Waddington, Distortion-Corrected Image Reconstruction with Deep Learning on an MRI-Linac. Magnetic Resonance in Medicine 90, 963-977 (2023).
  3. N. Koonjoo, B. Zhu, G. C. Bagnall, D. Bhutto, M. S. Rosen, Boosting the signal-to-noise of low-field MRI with deep learning image reconstruction. Scientific Reports 11, 8248-8248 (2021).
  4. T. Rahman, A. Bilgin, S. D. Cabrera, Multi-channel MRI reconstruction using cascaded Swinμ transformers with overlapped attention. Physics in Medicine and Biology 70, 075002 (2025).

Models were trained using data sourced from the following publications and their public repositories:

  1. Q. Fan, T. Witzel, A. Nummenmaa, K. R. A. Van Dijk, J. D. Van Horn, M. K. Drews, L. H. Somerville, M. A. Sheridan, R. M. Santillana, J. Snyder, T. Hedden, E. E. Shaw, M. O. Hollinshead, V. Renvall, R. Zanzonico, B. Keil, S. Cauley, J. R. Polimeni, D. Tisdall, R. L. Buckner, V. J. Wedeen, L. L. Wald, A. W. Toga, B. R. Rosen, MGH–USC Human Connectome Project datasets with ultra-high b-value diffusion MRI. NeuroImage 124, 1108-1114 (2016).
  2. F. Knoll, J. Zbontar, A. Sriram, M. J. Muckley, M. Bruno, A. Defazio, M. Parente, K. J. Geras, J. Katsnelson, H. Chandarana, Z. Zhang, M. Drozdzalv, A. Romero, M. Rabbat, P. Vincent, J. Pinkerton, D. Wang, N. Yakubova, E. Owens, C. L. Zitnick, M. P. Recht, D. K. Sodickson, Y. W. Lui, fastMRI: A Publicly Available Raw k-Space and DICOM Dataset of Knee Images for Accelerated MR Image Reconstruction Using Machine Learning. Radiology: Artificial Intelligence 2, e190007-e190007 (2020).

Refactored with the assistance of Claude Sonnet 4.6 on 6th August 2026.