Tom Holroyd wrote:
> Hmm. Yes, I grabbed a bunch of top level stuff from there,
> thanks. I'm not so sure about the time axis swapping issue...
> Can't you just transpose the axes without moving the data? I
> agree that d[0] is better than d[:,:,:,0], of course.
No data is moved in this implementation. It just reflects how the data is stored inside the binary blob of a nifti file and numpys default behaviour to access it.
Sure, one could swap the axis in any arbitrary order (numpy does this easily), but what for? In all use-cases I have discovered so far the current order of dimension works best IMHO.
What access method would you suggest?
Cheers,
Michael