DMBA post added - #514
Conversation
FWIW I would definitely test changes to the docs by building them locally. There are often issues you don't notice until you see the rendered HTML. |
| ### [Duke Mouse Brain Atlas](https://doi.org/10.1126/sciadv.adq8089) | ||
| This atlas from [Mansour et al. (2025)](https://doi.org/10.1126/sciadv.adq8089) is a 15 μm isotropic MRH-based stereotaxic atlas of the adult C57BL/6J mouse brain. The source atlas was built from five 90 ± 2 day male mice scanned in skull, limiting tissue deformation from cranial dissection and conventional histology. Micro-CT images of the skull provide bregma and lambda landmarks in the same stereotaxic space. | ||
|
|
||
| BrainGlobe includes the mean diffusivity template as the main reference image, the source RCCFv3 annotation image, generated meshes, 11 additional registered reference images, and two combined stereotaxic grid templates at 0.2 mm and 1 mm spacing that show the location of bregma. RCCFv3 is the Duke team's reduced version of the Allen CCFv3 label set, in which related structures smaller than 0.1 mm3 are grouped into larger regions to reduce registration noise from small structures. |
There was a problem hiding this comment.
could you please include how many unique regions are in this atlas vs the allen? Im curious how many regions were removed
There was a problem hiding this comment.
Thanks, good point. I checked both the paper and the packaged annotation volumes. In BrainGlobe, duke_mouse_15um contains 179 unique annotated region IDs, compared with 671 anatomical region IDs in allen_mouse_10um, excluding background and Allen’s root label. This is a net reduction of 492 annotation classes, or approximately 73%.
Aggregated is more accurate than removed since RCCFv3 combines related CCFv3 structures smaller than 0.1 mm³ into larger regions and also contains some RCCF-specific or “uncharted” labels. I’ll add these numbers and this clarification to the atlas description.
There was a problem hiding this comment.
this is such a cool gif. Im thinking about which should be the default image. I think the one which has the eyes perhaps? is this one the equivalent of t2?
There was a problem hiding this comment.
Thanks! The default reference image in the BrainGlobe package is the mean diffusivity (MD) volume.
The image showing the eyes is the unmasked multigradient-echo (mgre-unmasked) reference. It is more closely related to T2*-weighted contrast than conventional T2, so I would describe it as unmasked mGRE rather than T2.
The paper does not explicitly define one image as the default: DWI was used to construct the atlas template and as a registration target, while MD is used prominently for anatomical visualisation in the paper that's why I seleted MD as defult.
There was a problem hiding this comment.
ah ok so which is better for the default image in your opinion?
|
|
||
| For BrainGlobe, the main reference image is the DMBA mean diffusivity template, and the annotation image uses the RCCFv3 labels supplied with the atlas data. RCCFv3 is the Duke team's reduced version of the Allen CCFv3 label set: related structures smaller than 0.1 mm<sup>3</sup> were grouped into larger anatomical regions, reducing registration noise from small regions that are difficult to align reliably. | ||
|
|
||
| The BrainGlobe package includes 12 registered reference images. In addition to the main mean diffusivity template, the package includes AD, DWI, NQA, RD, FA, M0, M1, M2, M3, ISO, and unmasked mGRE. These complementary images can be useful for different anatomical questions because each contrast highlights different tissueproperties. |
There was a problem hiding this comment.
Could you explain what all these acronyms are? They won't mean much to anyone outside of the MRI world.
|
|
||
| **Figure 1. The 12 registered reference images included in the `duke_mouse_15um` atlas package.** | ||
|
|
||
| Another practical feature is the stereotaxic grid information. In the source atlas, micro-CT scans of the skull were registered to the MRH reference, giving bregma and lambda landmarks in the same coordinate space as the brain. BrainGlobe includes combined grid templates at 0.2 mm and 1 mm spacing that show the location of bregma, which can help with stereotaxic targeting and electrophysiology trajectory planning. |
There was a problem hiding this comment.
More of a question for the packaging side of this, but I find the inclusion of these grids a bit weird, as they are very different to other atlas data in the API.
I think we should expand more here, because users may read "can help with ..." and rightfully assume that BrainGlobe will provide tooling for this, rather than just showing them an image.
There was a problem hiding this comment.
we decided to remove these grids so this part of the blog can be removed
|
|
||
| ## Source and citation | ||
|
|
||
| The source data are available from [Duke CIVMImageSpace](https://civmimagespace.civm.duhs.duke.edu/tp_item_detail.php/view/item_number=DMBA/set_id=315). |
There was a problem hiding this comment.
I think this can be removed, it's not relevant to users, as long as there is a citation.
Co-authored-by: Adam Tyson <code@adamltyson.com>
3 participants
Description
What is this PR
Why is this PR needed?
This PR announces the addition of the Duke Mouse Brain Atlas (DMBA) to BrainGlobe and adds it to the website atlas documentation.
What does this PR do?
duke_mouse_15umentry to the atlas details page.References
Related atlas resources:
How has this PR been tested?
Local checks:
git diff --checkA full Sphinx build was not run locally because Sphinx is not installed in the current local Python environment.
Is this a breaking change?
No.
Does this PR require an update to the documentation?
Yes. This PR is itself a documentation/website update. It adds the Duke Mouse Brain Atlas to the atlas table and atlas details page, and adds a new blog post.
Checklist: