multimodallearning / DG-TTA

DG-TTA
https://pypi.org/project/dg-tta/
MIT License
10 stars 0 forks source link

how to do data normalization for different modalities #1

Closed argman closed 1 month ago

argman commented 1 month ago

thanks for the code, do you use the default normalization in nnunet ? i cannot see any code about data normalization

chr-wei commented 1 month ago

Hey @argman, yes we use the data normalization of nnunet. It is a bit hidden in the code. Please refer to the channel name in the dataset.json definition: https://github.com/multimodallearning/DG-TTA/blob/09df229d66d87383e9ad8c8ab132874cc67f0df1/dg_tta/__resources__/dummy_results/nnUNetTrainer_GIN_MIND_MultiRes__nnUNetPlans__3d_fullres/dataset.json#L112

For CT images you could apply CT-normalization as defined in the nnunet dataset definition. For MR images you can use z-score-normalization.