Register the segmentation back to the T1 space

[CCaredda]

Hi,

thank you for this great tool! I understood that Brats use preprocessed files (skullstripped and co-registered in the SRI-24 space).
With the pre-processing functions, is there a way to register the segmented volume back to the T1 space?

Thank you,
Best,
Charly

[neuronflow]

Hi Charly @CCaredda,
I transferred your question to the preprocessing repository. Assuming you are actually using this repo to preprocess your files?

Your answer depends on how you preprocess your files.

[CCaredda]

Hi,
Thank you for your answer.
This is exact, I use this repo to pre-process my files (PreProcessor class), see a portion of the code below:

` # define modalities
center = Modality(
modality_name="t1",
input_path=t1,
raw_bet_output_path=output_dir + "t1_bet.nii.gz",
atlas_correction=True,
normalizer=percentile_normalizer,
)

moving_modalities = [
    Modality(
        modality_name="flair",
        input_path=flair,
        raw_bet_output_path=output_dir + "fla_bet.nii.gz",
        atlas_correction=True,
        normalizer=percentile_normalizer,
    ),
]


preprocessor = Preprocessor(
    center_modality=center,
    moving_modalities=moving_modalities,
    registrator=ANTsRegistrator(),
    brain_extractor=HDBetExtractor(),
    temp_folder="temporary_directory",
    limit_cuda_visible_devices="0",
)
preprocessor.run()`

Best,

[neuronflow]

Hi Charlie,

Currently, we don't have a built-in method to warp back to native space.

This method allows you to save the intermediate computation steps and registration matrixes https://brainles-preprocessing.readthedocs.io/en/latest/brainles-preprocessing.html#brainles_preprocessing.preprocessor.Preprocessor.run

Depending on your registration framework, in your case, ANTs, one can now use the inverse of these steps to warp back to native space.

It would be nice to have such a method in the future. Ideally, we solve this one first, as this would boil it down to running a single transform:
#92

[CCaredda]

Thank you, I will take a look at this and let you know.
Best,