Clarification on Data Format and Slice Dimension in the Project

[naajeehxe]

Hi!,

I’m working on a project based on your work and had a quick question about the nii.gz file format. Could you clarify which dimension in your data represents the slice index when training the models?

Specifically, is the slice index in the first dimension (e.g., (250, 512, 512) where 250 is the number of slices), or in the third dimension (e.g., (512, 512, 250) where 250 is the number of slices)?

Thanks for your help! I just want to make sure I’m handling the data correctly.

Best,
Jeehye

[blueyo0]

Hi!,

I’m working on a project based on your work and had a quick question about the nii.gz file format. Could you clarify which dimension in your data represents the slice index when training the models?

Specifically, is the slice index in the first dimension (e.g., (250, 512, 512) where 250 is the number of slices), or in the third dimension (e.g., (512, 512, 250) where 250 is the number of slices)?

Thanks for your help! I just want to make sure I’m handling the data correctly.

Best, Jeehye

Hi, the answer is the first one, i.e. the slice index in the first dimension (e.g., (250, 512, 512) where 250 is the number of slices), we organized the data in NC DHW format.

Whether the data is organized with the dim order NCHWD or NCDHW depends on how tio.ScalarImage loads them.
you can refer to https://github.com/fepegar/torchio/blob/7c6a5c9d43fc97d8753226699baa66c5ca58d0aa/src/torchio/data/io.py#L33 for details.