MRIPreprocessor

This repository provides a simple pipeline to co-register different imaging modalities and skull strip them.

This package uses HD-BET (https://github.com/MIC-DKFZ/HD-BET) and ANTsPy (https://github.com/ANTsX/ANTsPy).

For example, this pipeline is designed for the BraTS dataset.

To install the package:

pip install  git+https://github.com/ReubenDo/MRIPreprocessor#egg=MRIPreprocessor

Example case 1:

Let's assume we have access to 4 imaging modalities (e.g. T1, T1c, T2, FLAIR) and we want to:

• co-register the scans using T1 as reference
• in the MNI space (1x1x1 mm)
• skull strip the scans using T1 as reference
• crop the skull-stripped scans to remove the zero padding

from MRIPreprocessor.mri_preprocessor import Preprocessor

# 4 Modalities to co-register to MNI space using an affine transformation
# T1 is used as reference for the coregistration
# No labelmap is used 
ppr = Preprocessor({'T1':'./data/example_T1.nii.gz',
                    'T2':'./data/example_T2.nii.gz',
                    'T1c':'./data/example_T1c.nii.gz',
                    'FLAIR':'./data/example_FLAIR.nii.gz'},
                    output_folder = './data/output',
                    reference='T1',
                    label=None,
                    prefix='patient001_',
                    already_coregistered=False,
                    mni=True,
                    crop=True)

ppr.run_pipeline()

The output folder will contain three folders nammed coregistration, skullstripping and cropping containing respectively the co-registered modalities, the skull-stripped and co-registered imaging modalities and the cropped versions of these latter skull-stripped scans. (example output './data/output/cropping/patient001_T1.nii.gz')

Example case 2:

Let's assume we have access to 4 co-registered imaging modalities (e.g. T1, T1c, T2, FLAIR) and we want to:

• co-registered them in the MNI space (1x1x1 mm) using T1 as reference
• skull strip the scans using T1 as reference
• crop the skull-stripped scans to remove the zero padding

from MRIPreprocessor.mri_preprocessor import Preprocessor

# 4 Modalities to co-register to MNI space using an affine transformation
# T1 is used as reference for the coregistration
# No labelmap is used 
ppr = Preprocessor({'T1':'./data/example_T1.nii.gz',
                    'T2':'./data/example_T2.nii.gz',
                    'T1c':'./data/example_T1c.nii.gz',
                    'FLAIR':'./data/example_FLAIR.nii.gz'},
                    output_folder = './data/output',
                    reference='T1',
                    label=None,
                    prefix='patient001_',
                    already_coregistered=True,
                    mni=True,
                    crop=True)

ppr.run_pipeline()

The output folder will contain three folders nammed coregistration, skullstripping and cropping containing respectively the co-registered modalities in the MNI space, the skull-stripped and co-registered imaging modalities and the cropped versions of these latter skull-stripped scans. (example output './data/output/cropping/patient001_T1.nii.gz')

Example case 3:

Let's assume we have access to 4 imaging modalities (T1, T1c, T2, FLAIR) and one segmentation drawn on the T1c scan. We want to:

• co-register the scans using T1c as reference
• in the MNI space (1x1x1 mm), including the labelmap
• skull strip the scans using T1c as reference
• crop the skull-stripped scans to remove the zero padding and apply the same cropping to the registered labelmap

Note that the reference scan must be the scan employed for the segmentation, here the T1c scan.

from MRIPreprocessor.mri_preprocessor import Preprocessor

# 4 Modalities to co-register to MNI space using an affine transformation
# T1 is used as reference for the coregistration
# A labelmap is used
ppr = Preprocessor({'T1':'./data/example_T1.nii.gz',
                    'T2':'./data/example_T2.nii.gz',
                    'T1c':'./data/example_T1c.nii.gz',
                    'FLAIR':'./data/example_FLAIR.nii.gz'},
                    output_folder = './data/output',
                    reference='T1c',
                    label='./data/example_Label.nii.gz',
                    prefix='patient001_',
                    already_coregistered=False,
                    mni=True,
                    crop=True)

ppr.run_pipeline()

The output folder will contain three folders nammed coregistration, skullstripping and cropping containing respectively the co-registered modalities and labelmap, the skull-stripped and co-registered imaging modalities and labelmap and the cropped versions of these latter skull-stripped scans. (example output './data/output/cropping/patient001_T1.nii.gz')