Choroid Plexus Segmentation Pipeline

The proposed pipeline is developed on the Freesurfer built-in pipeline and further fixs the Type I issue (extra masks around lateral ventricle area), aiming to get a more accurate choroid plexus segmentation. To better avoid the potential inconvenience from software installation, we suggest to use a estabilished singularity image from PennSIVE group that contains all required packages.

To better use the pipeline, the following steps should be followed:

• Organize MRI data into a specific format
• Set up singularity on the cluster
• Run Freesurfer built-in Pipeline
• Run Type I Issue Fixing Script
• Segmentation Quality Check
• Run Stats Regeneration and Extraction Pipeline

Data Structure

On the current stage, the proposed pipeline highly relied on the data structure. Therefore, we suggest users to manage their MRI data using the following format:

Main Directory: UVM_data (as you name)

(Note: users can download code and license from Github and save it in the main directory)

• data
  • subject
    • anat
      • flair.nii.gz
      • mprage.nii.gz
      • T2star.nii.gz
• code
  • automatic_typeI_fix.R
  • extract_mask.R
  • recon_all.sh
  • seg_edit.sh
  • TypeI_fixing.R
• license
  • license.txt
  • ASegStatsLUT.txt
• choroid_plexus
  • automatic_pipeline
  • editted
  • quality_check
  • volume

Some commands to help create choroid_plexus folder under the main directory:

cd /path/to/main_dir(UVM_data)

mkdir choroid_plexus
mkdir choroid_plexus/automatic_pipeline
mkdir choroid_plexus/editted
mkdir choroid_plexus/quality_check
mkdir choroid_plexus/volume

Set up Singularity

module load singularity
# you may have a different version of singularity on your cluster, please check for the right version.

sin_path=/path/to/save/image/neuror.sif

singularity pull -F $sin_path docker://pennsive/neuror

Run FreeSurfer Built-in Pipeline

SINGULARITYENV_TMPDIR=/scratch
# set singularity tmpdir to the cluster tmpdir, might be different from cluster to cluster.

bash code/recon_all.sh $sin_path "mprage.nii.gz" 

Note: Need to replace "/scratch" here and inside recon_all.sh to your cluster tmp directory. Be causious about how to submit a job on cluster, it might be different from PennSIVE cluster

(If you met any error regarding to "ERROR: crypt() returned null with 4-line file", feel free to let me know!!)

Type I Issue Fixing

After the Freesufer Pipeline is done for all subjects, we run the Type I Issue Fixing Step.

patient=`ls ./choroid_plexus/automatic_pipeline | grep -v "fsaverage"`

for p in $patient;
do
    bsub -J seg_fix -o out.log -e err.log singularity run --cleanenv \
    -B $PWD \
    -B /scratch \
    --env FS_LICENSE=$PWD/license/license.txt \
    --env SURFER_FRONTDOOR=1 \
    $sin_path Rscript $PWD/code/automatic_typeI_fix.R p=$p thre=0.2
done

Note: Again, be causious about submiting a job on cluster and binding cluster tmp directory.

Quality Check

Upon finishing the Type I Issue fixing step, make sure to conduct a quality check for editted choroid segmentation (we use a default threshold of 0.2 to fix the Type I issue, but some patients might need a smaller threshold due to the remaining of Type I issue). Check quality_summary_.csv file in the data/quality_check directory, go through each mask and record which patients need smaller threshold.

Then run Type I fixing for patients that require a different threshold:

• 20140213-1445
• 20160615-1501

patient="20140213-1445"
bsub -J seg_fix -o out_seg.log -e err_seg.log singularity run --cleanenv \
    -B $PWD \
    -B /scratch \
    --env FS_LICENSE=$PWD/license/license.txt \
    --env SURFER_FRONTDOOR=1 \
    $sin_path Rscript $PWD/code/automatic_typeI_fix.R p=$patient thre=0.1

Use patient "20140213-1445" for illustration

Stats Regeneration

bash code/seg_edit.sh $sin_path

Extract Volume Data

bsub -J seg_fix -o out.log -e err.log singularity run --cleanenv \
    -B $PWD \
    -B /scratch \
    --env FS_LICENSE=$PWD/license/license.txt \
    --env SURFER_FRONTDOOR=1 \
    $sin_path Rscript $PWD/code/extract_mask.R "aseg_editted.stats"