Structural Connectivity Analysis

Overview

This repository contains two scripts: tractography.sh for processing diffusion MRI data to perform tractography and generate the connectivity matrix and count_streamlines.sh for processing the tractogram and calculating the number of streamlines for each ROI of an atlas.

Table of Contents

1. Software Prerequisites
2. Input Requirements
3. Usage
   • Running Tractography
   • Running Count Streamlines
4. Notes
5. Troubleshooting
6. Contact

Software Prerequisites

• MRtrix3 (we used version 3.0.3)
• FSL (we used version 6.0.5)

Input Requirements

The scripts can require the following inputs:

1. DWI: Diffusion-weighted image (in NIFTI format, multi-shell)
2. BVEC: b-vector file
3. BVAL: b-value file
4. ANAT: Structural image (e.g., T1-weighted, in NIFTI format)
5. ATLAS: Atlas file in standard space (in NIFTI format)
6. WARP_STD2STRUCT: Warp of the transformation from standard space to structural space (in NIFTI format)
7. TRACT_FILE: tract file (.tck)
8. REMOVE_TEMP: Flag (0 or 1) for removing temporary files

Usage

First, make sure that the scripts have the right permisison to be run, by using chmod:

chmod 755 tractography.sh
chmod 755 count_streamlines.sh

Running Tractography

./tractography.sh <DWI> <BVEC> <BVAL> <ANAT> <ATLAS> <WARP_STD2STRUCT> <REMOVE_TEMP>

Example:

./tractography.sh subject1_dwi.nii.gz subject1_dwi.bvec subject1_dwi.bval subject1_t1.nii.gz atlas.nii.gz subject1_std2struct_warp.nii.gz 1

Running Count Streamlines

./count_streamlines.sh <DWI> <ANAT> <WARP_STD2STRUCT> <ATLAS> <TRACKS_FILE> <REMOVE_TEMP>

Example:

./count_streamlines.sh subject1_dwi.nii.gz subject1_t1.nii.gz subject1_std2struct_warp.nii.gz atlas.nii.gz tracts.tck 1

Notes

• For multi-shell acquisitions, the script uses the "dhollander" method for response function estimation. Other methods might be suited depending on the acquisition. For single-shell acquisitions, it is recommended to change to the "tournier" method.
• For single-shell acquisitions, the dwi2fod function should use the "csd" methods instead of the "msmt_csd".
• In the 5ttgen function, it is imperative to use the -nocrop flag so that the output image has the same size as the input image. Additionally, the -premasked flag was used since the T1 image was already skull-stripped.

Troubleshooting

• Ensure all input files exist and are in the correct format.
• Check if FSL and MRtrix3 are correctly installed and available in your path. MRtrix3 can run into some issues with Python>=3.10, with failed imports being the main symptom. If that's the case, downgrade to, at most, python3.9.
• For coregistration, do a manual inspection of intermediate images.
• If you find some other problem, feel free to create an issue.

Contact

For questions or other business contact @anamatoso

If you use this code in your research, please cite this repository or this paper.