Automatic segmentation of spinal canal

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

The spinal canal was defined using the anatomical boundary of the dural sac. The model was trained to segment all the structures within the dural sac, including the spinal cord, cerebrospinal fluid (CSF), and nerve rootlets.

Model Overview

The model is a 3D nnUNet, which was trained on T2-weighted images to segment the spinal canal.

How to use the model

Install dependencies

• Spinal Cord Toolbox (SCT) v6.5 or higher -- follow the installation instructions here
• conda
• Python

Once the dependencies are installed, download the latest canal model:

sct_deepseg -install-task canal_t2w

Getting the canal segmentation

To segment a single image, run the following command:

sct_deepseg -i <INPUT> -o <OUTPUT> -task canal_t2w 

For example:

sct_deepseg -i sub-001_T2w.nii.gz -o sub-001_T2w_canal_seg.nii.gz -task canal_t2w