spine-park

Pipeline for multicontrast analysis in PD patients.

How to use

Install dependencies

• Install Spinal Cord Toolbox (tested on v6.3)
• Install manual-correction

Clone this repository

git clone https://github.com/sct-pipeline/spine-park.git
cd spine-park

Install Python dependencies

pip install -r requirements.txt

Declare variables

PATH_DATA_RAW=<PATH TO ORIGINAL NIFTI DATASET>
PATH_DATA_BIDS=<PATH TO OUTPUT BIDS DATASET>

Convert data to BIDS

python convert_to_bids.py $PATH_DATA_RAW $PATH_DATA_BIDS

Perform manual vertebral labeling

Because automatic vertebral labeling is unreliable on this dataset (see #21 #24), manual labeling should be done instead. The procedure is as follows:

• Install manual-correction (see Install dependencies)
• Go in the folder:

  cd manual-correction

• Create a configuration file that lists all subjects in the dataset

  echo "FILES_LABEL:" > config.yml && find $PATH_DATA_BIDS -type f -name "*_T2.nii.gz" -exec basename {} \; | awk '{print "- " $0}' >> config.yml

• Perform manual labeling of discs by running this commmand:

  python manual_correction.py -path-img $PATH_DATA_BIDS -config config.yml

• If you want to quit and resume later, click on the Terminal window and press CTRL+C (KeyboardInterrupt). The manual-correction software will quit, and the config.yml will be modified such that the next time you re-run manual correction, you won't have to re-do the labels that you already did.

Here is a video tutorial:

Run processing across all subjects

sct_run_batch -script batch_processing.sh -path-data $PATH_DATA_BIDS -path-output <PATH_RESULTS> -jobs -1

To only run the processing in one subject (for debugging purpose), use this:

sct_run_batch -script batch_processing.sh -path-data $PATH_DATA_BIDS -path-output <PATH_RESULTS> -include-list sub-BB277

Run QC and manually correct the segmentations

Launch the QC report and flag with a ❌ the segmentations that need to be manually corrected. Then, download the YML file that list the problematic segmentations (bottom left button on the QC web browser) and run the script manual-correction to go through all the segmentations to be corrected.

The corrected segmentations need to be created under the derivatives/labels/ folder, at the root of the input dataset. This is done automatically by the manual-correction script, but the correct path need to be specified when running manual-correction.

The data with the derivatives folder should look like this:

├── derivatives
│   └── labels
│       └── sub-BB277
│           └── anat
│               └── sub-BB277_T2_seg.nii.gz
├── sub-BB277
│   ├── anat
│   │   ├── sub-BB277_T1map.json
│   │   ├── sub-BB277_T1map.nii.gz
│   │   ├── sub-BB277_T2.json
│   │   ├── sub-BB277_T2.nii.gz
│   │   ├── sub-BB277_UNIT1.json
│   │   ├── sub-BB277_UNIT1.nii.gz
│   │   ├── sub-BB277_mt-off_MTS.json
│   │   ├── sub-BB277_mt-off_MTS.nii.gz
│   │   ├── sub-BB277_mt-on_MTS.json
│   │   └── sub-BB277_mt-on_MTS.nii.gz
│   └── dwi
│       ├── sub-BB277_chunk-1_DWI.bval
│       ├── sub-BB277_chunk-1_DWI.bvec
│       ├── sub-BB277_chunk-1_DWI.json
│       ├── sub-BB277_chunk-1_DWI.nii.gz
│       ├── sub-BB277_chunk-2_DWI.bval
│       ├── sub-BB277_chunk-2_DWI.bvec
│       ├── sub-BB277_chunk-2_DWI.json
│       ├── sub-BB277_chunk-2_DWI.nii.gz
│       ├── sub-BB277_chunk-3_DWI.bval
│       ├── sub-BB277_chunk-3_DWI.bvec
│       ├── sub-BB277_chunk-3_DWI.json
│       └── sub-BB277_chunk-3_DWI.nii.gz
└── sub-DEV206Sujet10

Once segmentation masks are corrected, you can re-run the script, and the corrected segmentation will be used (if they exist).