search

JuliaNeuroscience / NIfTI.jl

Julia module for reading/writing NIfTI MRI files
Other
71 stars 35 forks source link

Consider making it difficult to make (u)int64 NIfTI images #61

Open effigies opened 2 years ago

effigies commented 2 years ago

Hi all, over in the NIPY world, we've just decided to deprecate (u)int64 support in NIfTI: https://github.com/nipy/nibabel/issues/1089

Basically nobody else (AFNI, FSL, SPM) supports them, and there's no compelling use case for them, but they're easy to make in Python if you aren't paying careful attention to types, so we're going to be making it hard to do unintentionally. Not sure if your technical situation is similar, but in case it is, I wanted to bring this to your attention before it becomes a painful decision.

Tokazama commented 2 years ago

FSL doesn't support Int64?

effigies commented 2 years ago

Not according to https://nifti.nimh.nih.gov/nifti-1/support/FSLandNIfTI1. It's possible that it's changed since 2007, but it is also not a very sensible data type (Generally you want integers for labels/masks or to compress files by using scale factors), so adding support would be pretty low priority. FSLeyes will support them, but that's because it's written in Python with nibabel as its I/O layer.