Preprocessing¶

BraTS challenge algorithms require preprocessed brain scans. This typically includes:

Co-registration of multi-modal MRI sequences
Brain extraction (skull stripping)
Registration to a challenge-specific brain atlas (SRI24 or MNI152)

Atlases¶

Atlas	Used by	Available at
SRI24	Pre-2024 Glioma Pre-Treatment, Africa, Meningioma, Metastases, Pediatric	Zenodo
MNI152	2024+ Adult Glioma (Pre & Post-Treatment)	Zenodo

Meningioma RT data remains in native image space and uses defacing only; it does not require an atlas.

The required atlas is also noted in each algorithm's table in the algorithm reference.

Using the Built-in Preprocessing¶

Install the preprocessing extra:

pip install brats[preprocessing]

The brats.preprocessing module provides convenience wrappers around the preprocessing package. The wrappers accept individual input and output image paths, not input/output directories:

from brats.preprocessing import preprocess_coreg_sri24reg_bet

preprocess_coreg_sri24reg_bet(
    t1_input="path/to/raw/t1.nii.gz",
    t1c_input="path/to/raw/t1c.nii.gz",
    t2_input="path/to/raw/t2.nii.gz",
    flair_input="path/to/raw/flair.nii.gz",
    t1_output="path/to/preprocessed/t1.nii.gz",
    t1c_output="path/to/preprocessed/t1c.nii.gz",
    t2_output="path/to/preprocessed/t2.nii.gz",
    flair_output="path/to/preprocessed/flair.nii.gz",
)

Available wrappers:

Function	Description
`preprocess_coreg_sri24reg_bet`	Co-registration + SRI24 atlas registration + brain extraction
`preprocess_coreg_sri24reg_defacing`	Co-registration + SRI24 atlas registration + defacing
`preprocess_coreg_mni152reg_bet`	Co-registration + MNI152 atlas registration + brain extraction
`preprocess_coreg_sri24reg_bet_allow_missing`	Same as above but tolerates missing modalities
`preprocess_deface_only`	Defacing only
`preprocess_for_challenge`	Router that picks the right pipeline for a given challenge; source-dataset-specific GoAT, inpainting, and Missing MRI atlas choices may require custom preprocessing

Custom Preprocessing¶

The modular architecture of BraTS Orchestrator allows you to use your own preprocessing routines. The preprocessing package can also be used independently to design custom pipelines tailored to your specific needs.