Spinal Cord Lesion Frequency Map (LFM) with SCT

Pipline to make Lesion Frequency Map (LFM) along the spinal cord using SCT.

It's a bash script inspired by the Python code in neuropoly/lesion-mapping.

• Dependencies
• Dataset structure
• Prerequirments
• Steps
• Usage
• Usage
• Examples

Dependencies

• Spinal Cord Toolbox (SCT)

Dataset structure

• Similar to the result of running sct_run_batch.

Pre-Requirements

Before running the script you should have in each subject's folder the following:

1. Spinal cord segmentation (Can be generated with sct_deepseg_sc).
2. Lesions segmentation (Can be generated with sct_deepseg_lesion).
3. Warping fields to template space (Can be generated with sct_register_to_template).

Steps

The script perform the following steps:

• for each subject:
  • Look in subject folder for all cord segmentations.
  • Look in subject folder for matching warping fields.
  • Merge all cord segmentations into the template space using the warping fields (nn interpolation).
    • create <subject><cord suffix>_template.nii.gz.
  • Look in subject folder for all lesions segmentations.
  • Look in subject folder for matching warping fields.
  • Merge all lesions segmentations into the template space using the warping fields(linear interpolation).
    • create <subject><lesions suffix>_template.nii.gz.
• Then:
  • Sum all cord segmentation.
  • Sum all lesions segmentation.
  • Generate LFM with sum of all lesions divided by the sum of all cords.

Usage

make_lfm.sh \
    [-d <Data directory. default: "output/data_processed">] \
    [-s <Subjects file - text file with subjects names, subject per line. default: "subjects.txt">] \
    [-f <Data directory in subject folder. default: "anat">] \
    [-i <Image pattern (Will look also with "${SUBJECT}_" as prefix by default). default: "t2w">] \
    [-l <Lesions segmentation file suffix (Must be a binary map). default: "_lesionseg">] \
    [-c <Spinal cord segmentation file suffix (Must be a binary map). default: "_seg">] \
    [-w <Warping fields to template space file suffix pattern. default: ".*?warp_anat2template">] \
    [-o <Output LFM file. default: "LFM.nii.gz">] \
    [-r <Overwrite (0/1). If 1, the analisis will run again even if files exist. default: 1>] \
    [-m <Set all spinal cord area wich is not covered by all subject's cord segmentation to 0 (0/1). default: 1>] \
    [-t <Template prefix. default: "${SCT_DIR}/data/PAM50/template/PAM50_">] \
    [-a <Min level of sponal cord to be in result. default: 1 (C1)>] \
    [-b <Max level of sponal cord to be in result. default: 20 (T12)>]

How to run

For the following dataset:

- ~/
  - some_study/
    - output/
      - data_processed/
        - sub-1/
          - anat/
            - t2.nii.gz
            - t2_seg.nii.gz # cord mask (binary)
            - t2_lesionseg.nii.gz # lesion mask (binary)
            - t2_warp_anat2template.nii.gz # warping fields to template
        - sub-2/
          - anat/
            - t2.nii.gz
            - t2_seg.nii.gz # cord mask (binary)
            - t2_lesionseg.nii.gz # lesion mask (binary)
            - t2_warp_anat2template.nii.gz # warping fields to template
        - ...

• Download (or git clone) this repository:

git clone https://github.com/yw7/sc_lesion_frequency_map.git

• Add executable permissions:

chmod +x sc_lesion_frequency_map/make_lfm.sh

• Make subjects_all.txt (You can edit this file to make LFM for subgroup of subjects):

ls some_study/output/data_processed > some_study/subjects_all.txt

• Run the script (make LFM for Spinal Cord level C1 to C7):

sc_lesion_frequency_map/make_lfm.sh -d some_study/output/data_processed -s some_study/subjects_all.txt -o some_study/LFM_all.nii.gz -i t2 -a 1 -b 7

More Examples

Do not set all spinal cord area wich is not covered by all subject's cord segmentation to 0 (If some template voxel cover by some subjects it's value in the LFM will be the lesion frequency among those subjects):

sc_lesion_frequency_map/make_lfm.sh -d some_study/output/data_processed -s some_study/subjects_all.txt -o some_study/LFM_all.nii.gz -i t2 -m 0 -a 1 -b 7

Without overwriting previous analysis (For cases when running the script multiple time with overlaping groups, do not transform segmentations to template again):

sc_lesion_frequency_map/make_lfm.sh -d some_study/output/data_processed -s some_study/subjects_all.txt -o some_study/LFM_group1.nii.gz -i t2 -a 1 -b 7 -r 0

If the files are in subject's folder (not in 'anat' folder):

sc_lesion_frequency_map/make_lfm.sh -d some_study/output/data_processed -s some_study/subjects_all.txt -o some_study/LFM_group1.nii.gz -i t2 -d . -f "" -m 0 -r 0 -a 1 -b 7