Orientation - flag or detection?

[eilidhmacnicol]

Currently, fprodents expects images which are RAS+ with respect to the subject.
However, the software tools available for conversion to nifti from the scanner proprietary file type are not necessarily in agreement with which convention to follow, as demonstrated below.

Anatomical convention follows standard anatomical convention:
y = anterior(rostral)-posterior(caudal); z = superior(dorsal)-inferior(ventral)

Scanner convention follows the same convention used in humans:
y = superior-inferior; z = anterior-posterior.

This leads to inconsistencies in assigning the y and z axes for different datasets, which are further exaggerated when you consider incorrectly labelled scan protocols in paravision (e.g. recording that the animals are head-supine when they are head-prone so that it looks the right way round in spm without rotating the images 180º).

While we can continue to push for the anatomical convention from the start (such as converting with https://github.com/BrkRaw/bruker), we have to expect that until this is resolved in the larger community that we will have to have some net in place to catch this.

My instinct is that I'd rather not support it and instead encourage images in a standard anatomical convention. However, not supplying a standardised way to do this is almost certainly going to lead to problems further down the line.

I guess, I'd like to know what other people think about either providing a command-line flag which informs fprodents about which axis is A-P and which is S-I (L-R doesn't seem to be a problem, plus people sometimes leave a little marker, like toothpaste, on one side of the head coil so they can deduce which side is which). A better cooler way would be to auto-detect, perhaps leveraging the elongated A-P axis in rodents, but it might not work very well in fat/large animals with tissue that fills the image in the L-R, or in whole head images...