This page details the different steps of our analysis. We assume that we have

• a T1 image, used for the segmentation of different anatomical structures
• an ASL/Proton Density image pair for each Post Labelling Delay. The PD image is used for computing the Cerebral Blood Flow (CBF) in absolute units (ml/100g/min).

The different steps are:

• Conversion of DICOM data to NIfTI
• Plexus choroid segmentation on T1 image
• FastSurfer segmentation on T1 image
• Rigid co-registration of all ASL/PD images
• Computation of CBF from ASL images
• Calibration of CBF into absolute units (ml/100g/min) using PD images
• Rigid registration of T1 onto ASL/CBF
• Computation of statistics for each label in the segmentation

FastSurfer must be cloned from GitHub.

git clone https://github.com/Deep-MI/FastSurfer

I recommend to create a conda environment for FastSurfer in order to install the necessary packages (https://github.com/Deep-MI/FastSurfer/blob/master/requirements.txt). FreeSurfer does not need to be installed. A GPU enables fast segmentation (under 1 minute).

FSL can be loaded on the cluster

module load FSL/6.0.3

Use dcm2niix

module load dcm2niix
mkdir nifti_data
dcm2niix -o nifti_data -f '%p_serie_%s' dicom_data_directory

where dicom_data_directory is the directory containing all the DICOM data of our patient. The NIfTI files should appear in the directory nifti_data. For example, we may obtain

nifti_data
├── 3D_Ax_ASL_1025.nii
├── 3D_Ax_ASL_1025.json
├── 3D_Ax_ASL_1025a.nii
├── 3D_Ax_ASL_1025a.json
├── 3D_Ax_ASL_1525.nii
├── 3D_Ax_ASL_1525.json
├── 3D_Ax_ASL_1525a.nii
├── 3D_Ax_ASL_1525a.json
...
├── 3D_Ax_T1_MPRAGE_F.nii
└── 3D_Ax_T1_MPRAGE_F.json

Each .nii file is accompanied by a .json file containing the meta-data (acquision parameters, etc).

python plexus_segmentation.py -i nifti_data/3D_Ax_T1_MPRAGE_F.nii

This will add the file plexus_mask_3D_Ax_T1_MPRAGE_F.nii into the directory nifti_data. This mask contains continuous values between 0 and 1 and is in the native space of the T1 image.

Create a directory for the outputs:

mkdir nifti_data/fastsurfer_outputs

In order to run FastSurfer, we need to go to the directory into which we have cloned FastSurfer

cd fastsurfer_directory

Now we can run FastSurfer

./run_fastsurfer.sh --t1 .../nifti_data/3D_Ax_T1_MPRAGE_F.nii --sd .../nifti_data --sid .../nifti_data/fastsurfer_outputs --seg_only

This should create a file

nifti_data
└── fastsurfer_outputs
    └── mri
        └── aparc.DKTatlas+aseg.deep.mgz

The file aparc.DKTatlas+aseg.deep.mgz is in a FastSurfer conformed space. We will now move it back to the native space of the T1 image

mri_vol2vol \
--mov nifti_data/fastsurfer_outputs/mri/aparc.DKTatlas+aseg.deep.mgz \
--targ nifti_data/3D_Ax_T1_MPRAGE_F.nii \
--o nifti_data/fastsurfer_outputs/mri/aparc.DKTatlas+aseg.deep_native_space.nii.gz \ 
--regheader --interp nearest