RETSE_Metal

This repository contains the code and datasets for "Distortion correction in TSE near titanium implants at 0.55T using reversed frequency-encoding and model-based reconstruction", by Bochao Li, Nam G. Lee, Daehyun Yoon, Kubra Keskin, Alexander R. Toews, Jay Acharya, Jordan S. Gross, Brian A. Hargreaves, Krishna S. Nayak.

Bochao Li, University of Southern California, July 2025.

Distortion Correction Pipeline for Cartesian MRI Data

This repository contains MATLAB code to correct gradient non-linearity (GNL) and off-resonance distortion in Cartesian MRI acquisitions using dual frequency-encoding techniques. It includes tools for image reconstruction, off-resonance displacement estimation, and phase-constrained SENSE based joint reconstruction.

Setup

1. git clone https://github.com/usc-mrel/RETSE_Metal.git

2. Required Installation

• BART (Berkeley Advanced Reconstruction Toolbox)

• ISMRMRD (MRI raw data format)

3. Required third-party dependencies

Before running the pipeline, make sure the following third-party dependencies are downloaded and placed under the directory: /Helper_Functions/thirdparty/ (Please create it for first use)

1. BART (MATLAB interface)
   https://github.com/mrirecon/bart/tree/master/matlab

2. FINUFFT
   https://github.com/flatironinstitute/finufft

• Please install it correctly according to the online document on your system. This is very important!!

3. mapVBVD
   https://github.com/pehses/mapVBVD

4. export_fig
   https://www.mathworks.com/matlabcentral/fileexchange/23629-export_fig

Folder Structure

• main_recon.m: Main entry point for the reconstruction pipeline with automatic procedures.
• Helper_Functions/: Contains custom MATLAB functions and third-party dependencies.
• results/: Stores output images and figures (not included by default).
• raw/: Directory for raw data (available on Zenodo).

Data

To run this pipeline, you need reverse-encoding Turbo Spin Echo (TSE) raw MRI data acquired in two frequency-encoding directions data.

A sample dataset (THA, rBW = 150 Hz/Px) is publicly available on Zenodo:

Folder Setup

After downloading the dataset, please unzip it and copy to the code folder, then you can get the following directory structure manually:

\raw\
  |- PE_RL\
    |- <PE_RL_data>.dat
  |- PE_LR\
    |- <PE_LR_data>.dat

How to Run

Before running main_recon.m, please update the following variables at the top of the script:

ismrmrd_path = 'path_of_your_ismrmrd_installation'; (not BART MATLAB codes under \Helpfer_Functions)
bart_path    = 'path_of_your_bart_installation';

Outputs

• .json are the configuration files that are generated in the pipeline, saved in \raw folder.

• Estimated off-resonance displacement maps in meter are saved under offres .

• Corrected image per slice using proposed phase-constrained SENSE method are saved (img_typ1_xxx.cfl) in \result\dualrecon_pcs folder that is generated automatically; \result\meas_xxx folder contains non-corrected, corrected image for each direction's data.

• After show_results, img_recon has 5 multi-slice TSE images, including Dir.1 TSE, corrected Dir.1 TSE, Dir.2 TSE, corrected Dir.2 TSE, corrected image from PES-SENSE.

References and Acknowledgements

1. Mullen M, Garwood M. Dual polarity encoded MRI using high bandwidth radiofrequency pulses for robust imaging with large field inhomogeneity. Magn Reson Med. 2021;86(3):1271-1283. doi:10.1002/mrm.28771
2. Tao S, Trzasko JD, Shu Y, Huston J 3rd, Bernstein MA. Integrated image reconstruction and gradient nonlinearity correction. Magn Reson Med. 2015;74(4):1019-1031. doi:10.1002/mrm.25487