Mouse reference atlas: NRRD file header problem, and question on axes

Hi, I am looking at the nrrd file downloadable from the ReferenceSpaceCache, querying for the 100um template stack. If I look in the header, I get the following space description:

('space', 'left-posterior-superior'),
('sizes', array([132,  80, 114]))

The space entry does not match, as the stack axes order should be PIR (but see later) as specified in the documentation.

Second problem. I am confused by the fact that reading here one would assume that origin (voxel [0, 0, 0]) of the stack template stack numpy array should be anterior, superior, left; but the stack that I get with ReferenceSpaceCache has voxel [0, 0, 0] in the anterior, superior, right corner. Has this to do with the confusion between image/numpy ij indexing vs cartesian coordinate system?

1 Like


Here is some information from a colleague:

I am not sure why the ‘space’ in the header is defined as ‘left-posterior-superior’. In fact, the ‘sizes’ in the same header lists the axes as having sizes 132, 80, and 114 which correspond to the AP, DV, and ML axes respectively. So the ‘sizes’ are PIR in the header, and I am not sure what the ‘space’ metadata is describing. If you load the nrrd file in ITK Snap, the origin is at the anterior, left, superior corner of the reference space (PIR orientation), as described in, but the views are flipped so that the first window shows the sagittal, the second shows the coronal, and the third shows the horizontal sections. Again, I am not sure why ITK flips these or why the header is described that way, but the origin maps correctly for me.

Best regards,