Automatic Segmentation of Spinal Nerve Rootlets

This repository contains the code for deep learning-based segmentation of the spinal nerve rootlets. The code is based on the nnUNet framework.

Citation Info

If you find this work and/or code useful for your research, please cite the following paper:

@article{10.1162/imag_a_00218,
    author = {Valošek, Jan and Mathieu, Theo and Schlienger, Raphaëlle and Kowalczyk, Olivia S. and Cohen-Adad, Julien},
    title = "{Automatic Segmentation of the Spinal Cord Nerve Rootlets}",
    journal = {Imaging Neuroscience},
    year = {2024},
    month = {06},
    issn = {2837-6056},
    doi = {10.1162/imag_a_00218},
    url = {https://doi.org/10.1162/imag\_a\_00218},
}

Model Overview

The model was trained on T2-weighted and MP2RAGE (T1-weighted INV1 and INV2, and UNIT1) images and provides semantic (i.e., level-specific) segmentation of C2-T1 dorsal and ventral spinal nerve rootlets.

How to use the model

Install dependencies

• Spinal Cord Toolbox (SCT)—follow the installation instructions here

Usage SCT v7.0+

Once the dependencies are installed, download the latest rootlets model (r20250318—segmenting C2-T1 dorsal and ventral rootlets from T2w and MP2RAGE images):

sct_deepseg rootlets -install

Getting the rootlet segmentation

To segment a single image, run the following command:

sct_deepseg rootlets -i <INPUT> -o <OUTPUT>

For example:

sct_deepseg rootlets -i sub-001_T2w.nii.gz -o sub-001_T2w_label-rootlets_dseg.nii.gz

Usage SCT v6.2+

Instructions for 6.2 (with old syntax)

Once the dependencies are installed, download the rootlets model (r20240730—segmenting C2-C8 dorsal rootlets from T2w images):

sct_deepseg -install-task seg_spinal_rootlets_t2w

Getting the rootlet segmentation

To segment a single image, run the following command:

sct_deepseg -i <INPUT> -o <OUTPUT> -task seg_spinal_rootlets_t2w

For example:

sct_deepseg -i sub-001_T2w.nii.gz -o sub-001_T2w_label-rootlets_dseg.nii.gz -task seg_spinal_rootlets_t2w