Data analysis pipeline for the pediatric spinal cord

Project description

This repository contains a pipeline for pediatric spinal cord data analysis.

For this project, MRI data from the philadelphia-pediatric dataset was used, which includes typically-developing subjects aged 6 to 17.

Dependencies

To run the scripts in this repository, you will need the following installed or cloned on your local computer:

• The Spinal Cord Toolbox (SCT), version 7.0
• A local clone of the spinal cord rootlet segmentation model (release r20250318)

Usage

1. Modify the contents of the configuration files

To execute the scripts in this repo, you will need the following configuration files, located inside the config folder :

• config_morphometrics.yaml : configuration file to process morphometrics
• config_rootlets.yaml : configuration file to process spinal rootlets

Templates versions of these configuration files are provided in the config folder, with a _template suffix. Before running the scripts, rename each config file by removing the _template suffix. Inside each configuration file, you will need to set the following parameters:

• path_data : the path to your local copy of the dataset
• path_output : the path to where you want the output results saved
• jobs : the number of cores you want to use to run subjects in parallel
• script_args : the path to your local clone of the spinal cord rootlet segmentation model (r20250318)
• include_list: the list of subjects you want to process

2. Activate SCT's virtual environment and change directory

To run the scripts in this repository, you first need to activate the Spinal Cord Toolbox (SCT) virtual environment named venv_sct. You can activate it with the following command:

conda activate /path/to/your/venv_sct

*To find the exact path to venv_sct on your system, run conda env list to display all conda environments

To run the scripts described in the following sections, make sure to change your directory to the project’s repository in your terminal:

cd path/to/your/local/clone/pediatric_SC_morphometrics

3. T2w data preprocessing

The script T2w_data_preprocessing.sh (inside scripts/preprocessing) performs the following preprocessing steps on the dataset :

• Segmentation of spinal cord from T2w data (sct_deepseg_sc)
• Detection of PMJ from T2w data (sct_detect_pmj)
• Labeling of vertebral levels from T2w data (sct_label_vertebrae)
• Segmentation of rootlets from T2w data (model-spinal-rootlets_ventral_D106_r20250318)

In order to run the script in parallel, the sct_run_batch command can be used as follows :

sct_run_batch -config config/config_preprocessing.yaml -script scripts/preprocessing/pediatric_data_preprocessing.sh

4. Extract morphometrics

The script morphometrics.py (inside scripts/analysis) allows to :

• Get the labeled segmentation from the disc labels (using sct_label_vertebrae)
• Run sct_process_segmentation to compute spinal cord morphometrics (cross-sectional area, diameter, etc.)

To run this python script in pararallel, you can use the bash wrapper called wrapper_morphometrics.sh inside the sct_run_batch command :

sct_run_batch -config config/config.yaml -script wrappers/wrapper_morphometrics.sh

5. Process rootlets

The script rootlets.py (inside scripts/analysis) :

• Runs zeroing_false_positive_rootlets.py from the model-spinal-rootlets/pediatric_rootlets directory to remove false positive rootlets below the Th1 level.
• Runs 02a_rootlets_to_spinal_levels.py from the model-spinal-rootlets/inter-rater_variability directory to extract spinal levels from the rootlets segmentation
• Extracts the center of mass of each spinal level and saves it in a CSV file

The wrapper_rootlets.sh wrapper can be used to run the rootlets.py script in parallel with sct_run_batch :

sct_run_batch -config config/config.yaml -script wrappers/wrapper_rootlets.sh