Repository for generating high resolution ex-vivo MRI template.
Gros, C., Asiri, A., De Leener, B., Watson, C., Cowin, G., Ruitenberg, M., Kurniawan, N., Cohen-Adad, J., 2020. Ex vivo MRI template of the human cervical cord at 80μm isotropic resolution, in: Proceedingsof the 28th Annual Meeting of ISMRM. Presented at the ISMRM.
The ex-vivo template is available in this repository, which contains:
template.nii.gz: ex-vivo template.mask_spinalcord.nii.gz: binary spinalcord mask.map_greymatter.nii.gz: probabilistic grey matter map.mask_spinalsegments.nii.gz: mask of spinal segments, the value corresponds to the cervical segment, e.g. voxel value of 5 corresponds to C5 spinal level.mask_motortracts.nii.gz: mask of spinal motor tracts, the value corresponds to a specific motor tract, see Labels sub-section for details.
To install, run:
git clone https://github.com/sct-pipeline/exvivo-template.git
cd ivadomed
pip install -e .
To use the tools to generate the template, you will need to install additional dependecies, as described here.
Manual labelling of the rostral and caudal extent of each nerve root was performed using fsleyes. Each spinal level was then identified by orthogonal projection of these labels onto the spinal cord centerline, using sct_label_utils -create-seg.
Spinal cord and grey matter tissues were automatically segmented using a deep learning model trained and applied using IVADOMED. For each subject, the network was trained on 20 randomly picked and manually segmented slices, then inferred on the ~1,000 remaining slices. Results were reviewed and manually corrected when needed (~5%). The trained model is available here.
Preprocessing pipeline has been adapted from this project. The code is available under generate_template/. To run it:
source sct_launcher
python generate_template/pipeline.py
To generate the template with N (here N=13) subjects:
python -m scoop -n 13 -vvv generate_template/generate_template.py
Charley Gros, Nyoman Kurniawan, Benjamin De Leener, Charles Watson and Julien Cohen-Adad.