[data] allen CCF v3 transform may be errorenous
as of siibra-0.4a51
allen ccf v3 template is at https://neuroglancer.humanbrainproject.eu/precomputed/AMBA/templates/v3/stpt
The transform
curl -L https://neuroglancer.humanbrainproject.eu/precomputed/AMBA/templates/v3/stpt/transform.json | jq
prints out to be
[
[
1,
0,
0,
-5737500
],
[
0,
1,
0,
-6637500
],
[
0,
0,
1,
-4037500
],
[
0,
0,
0,
1
]
]
Accounting for half voxel shift, (full voxel is 25um) the translation in um ends up being:
[ 5725, 6625, 4025 ]
Total volume size (in um), according to the neuroglancer info, is
[ 11400, 13200, 8000 ]
This means, where we regard as [0, 0, 0 ] is exactly shifted 1 voxel to the RAS direction.
This can be verified in the viewer. at [0, 0, Z], the viewer is not at the center of the brain, as far as left right hemisphere is concerned:
Here is what I propose:
- we fix all transforms associated with allen ccf v3
What impact does this fix have:
- if calculations have been done previously in CCF v3 in siibra, they will not be reproducable. After fix, the value will shift by 1 pixel (or 25um) in the LPI direction.