Can we directly use unilateral .nii.gz data for training and submission?

Hello,

I downloaded the dataset using the "unilateral" config from Hugging Face. It is already in .nii.gz format, preprocessed (resampled, cropped, and split left/right), and matches the structure described in the documentation.

I also saw there's a Parquet version and a preprocessing script to transform "default" into "unilateral" format.

Do we have to start from the Parquet or default version and apply the preprocessing ourselves, or is it acceptable to directly use the provided unilateral .nii.gz files for training and submission?

Thanks.

Steve

Hi Steve

When your container is run on the Grand Challenge site for evaluation, it will be given the inputs of the T2-weighted acquisition and the DCE acquisition (one pre-contrast phase and 2-7 post-contrast phases) in .mha format. This will be in their original form, not as a subtraction or unilateral image.

We provided the unilateral data along with the code to produce it for competitors to use or amend as they wished - we viewed it as a useful starting point.

Best wishes and good luck

Nick