Datasets alignment and registration: Was ARA ever re-aligned with the CCFv3 average template?

Dear Community users,

I am a collaborator of the Blue Brain Project (EPFL Lab, Swiss) and I am contacting you because I am working on datasets to which several AIBS scientists have contributed. These datasets are the average template of the Allen Mouse Atlas CCFv3 and the Allen Reference Reference Atlas’ Nissl files .

I would be very grateful to you if you could answer the following question.

Were the Nissl volumes and Nissl images ( of the ARA re-aligned with the CCFv3 average template?

The last modification date of 2015 suggest that it could be the case, but direct inspection indicates that the ARA volumes are in better agreement with CCFv2 annotations than the CCFv3 annotations.

If the answer to the above question is negative, then I would dare to ask:

Do you have a registered version of the Nissl volumes or Nissl images in your CCFv3 registration framework? If you have such a dataset, it would be extremely useful for us. Indeed, this would help us compute more accurate cell densities in several regions of the mouse brain.

The following lines of the Allen Mouse CCF technical white paper on page 3 gave me some hopes:

" Reference data were registered to the average template using customized methods. Registration accuracy was limited due to the modest amount of deformation and tissue damage. Regardless, these datasets have great utility in providing anatomical details for delineating certain structures."

Many thanks in advance for your response,

Luc Guyot.

Hi Luc,

  • mouse_ccf/ara_nissl is the Nissl volume after it has been registered and rewarped to mouse_ccf/average_template

  • mouse_ccf/annotation/mouse_2011 is the “CCFv2 annotation based on the Nissl” rewarped in the same way as mouse_ccf/ara_nissl

  • mouse_ccf/annotation/ccf_2017 is the “CCFv3 annotation based on the average template” and as described in the whitepaper

There will be discrepancy between mouse_2011 and ccf_2017 due to multiple factors including:

  • Difference parcellation strategy of the original Allen Reference Atlas versus the 3D multi-modal approach of the CCFv3 including the use of registered data to ground the paracellation

  • Registration error between the Nissl volume and the average template

  • Limitation in the 3D reconstructed Nissl cause by deformation and tissue damage of the Nissl section

Thanks a lot, this definitely helps!

Lydia – is there a way to access the transform that was used to warp the ara_nissl and mouse_2011 volumes, i.e., so a point in CCFv2 can be transformed into its CCFv3 location?

Luc – I am specifically interested in getting the position data from the Blue Brain Cell Atlas in CCFv3 coordinates, I assume that’s related to what you’re working on. Please let me know if you have this information already!


@joshs Dear Josh, unfortunately we don’t have this information. There is some on going effort to combine CCFv2 and CFFv3 in order to get the level of details of the former (some hierarchical details have been lost in the CCFv3 annotation) and the smoothness of the latter. But nothing conclusive yet.

We would also like to have a way to map/warp the CCFv2 2011 geometry to better match the CCFv3 2017 geometry in order to reinstate lost regions within CCFv3 in a convincing way. The same transformation applied to the Nissl stained volume would potentially offer a better alignment of CCFv3 annotation and Nissl stains. But Lydia replied above that this transformation we are looking for has been already applied to CCFv2 2011.