Hi everyone,
I recently stumbled upon a great source of confusion and was hoping someone could point me towards the part of the documentation that can clarify this. Essentially I realised that my images seem LAS encoded in MRtrix but RAS encoded in FSL (at least that’s how they show up in mrview/fsleyes), and also LAS when I use nibabel’s get_fdata(). Reading the discussion e.g. here Vector orientation mismatch between FSL & MRtrix I thought this was a bug (“With a recent installation (more recent than April 15 2016, when this fix was released), there should be no mismatch.”) but looking into this Images and other data — MRtrix3 3.0 documentation it seems like this is an intentional difference in conventions.
Whichever it is (maybe someone could shed some light on that too), I’m now wondering: do I need to flip the images with mrtransform -flip and (how) will that affect bvec files? Will it always be the case that the orientations I see in mrview will not match those in e.g. nibabel arrays, or how can I check whether they don’t? Essentially what do people do to check if orientations cause issues in their pipeline and how do they mitigate this, or am I just overthinking it?
I’m a bit concerned because I only discovered this coincidentally when specific voxels identified by a script using nibabel didn’t match up with the images in mrview. I originally thought this was a mif/nifti issue and that the nifti image should match the nibabel array, but this is clearly not the case. Now I’m worried there may be other issues that I haven’t spotted because I didn’t quite understand how this was handled, and because I don’t know how to elegantly deal with the issue (or what exactly the issue is, to be honest).
Thanks in advance!
Tanja