ivadomed / ivadomed/model-spinal-rootlets

Rootlets-informed lumbar registration

Open
#92 3 comments 1 reaction 0 assignees View on GitHub
lumbar rootlets
Dominant language
Python
Stars
8
Forks
2
PR merge metrics
No merged PRs in 30d

Description

## Description
Similarly to the cervical spinal cord ([repo](https://github.com/sct-pipeline/rootlets-informed-reg2template), [preprint](https://arxiv.org/abs/2505.00115)), we could leverage rootlets or spinal levels to inform the registration to the PAM50 template.

> [!NOTE]
> Since lumbar rootlet segmentation has proven to be challenging, we may need to manually label the spinal levels based on the rootlet entry zones. Or develop a method to do so automatically. (For the subject below, manually segmented rootlets are available.)

![Image](https://github.com/user-attachments/assets/3c9dbfa7-67b5-460a-8360-aee86e376d6a)

Pseudo-commands:

```bash
# 1. Cropping
# I cropped the image manually to reduce its size in FSLeyes using Tools/Crop. I saved it with the suffix _roi

# 2. Cord segmentation
sct_deepseg spinalcord -i sub-CTS04_ses-SPpre_acq-ax_T2w_roi.nii.gz

# 3. Cord segmentation labeling
# I manually labelled the cord segmentation in FSLeyes based on the rootlets entry zones. I saved it with the suffix roi_seg_labeled

# 4. Rootlet midpoints
sct_label_utils -i sub-CTS04_ses-SPpre_acq-ax_T2w_roi_seg_labeled.nii.gz -cubic-to-point -o labels.nii.gz

# 5. Registration
sct_register_to_template -i sub-CTS04_ses-SPpre_acq-ax_T2w_roi.nii.gz -s sub-CTS04_ses-SPpre_acq-ax_T2w_roi_seg.nii.gz -lspinal labels.nii.gz -c t2
```

## TODO

- [ ] tweak registration parameters
- [ ] try different number of labels (2 or 3+; [details](https://spinalcordtoolbox.com/stable/user_section/tutorials/vertebral-labeling/how-many-labels-for-registration.html))

Contributor guide

No contributing guide indexed for this repository

Research direction

Start by reproducing the listed cropping, spinal-cord segmentation, labeling, midpoint extraction, and sct_register_to_template commands on the provided subject. Compare registration results while varying the spinal-level label count and registration parameters; done means documenting a reliable configuration or identifying the remaining limitations.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
computer-vision, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
32/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.