Human Brain Atlas Mesh Files

Technical
, , , , , ,
1

Hello,

I was wondering if the Allen Institute provides structural meshes for a segmented human reference brain similar to the mouse ccf meshes provided here: http://download.alleninstitute.org/informatics-archive/current-release/mouse_ccf/annotation/ccf_2017/structure_meshes/ply/

I know that these files exist: http://download.alleninstitute.org/informatics-archive/allen_human_reference_atlas_3d_2020/version_1/examples/convert_to_itksnap/

But I can’t find any documentation on what these files are or how to use them to get meshes in a useable file type (.ply, .stl, .obj, etc.).

Thank you,

Jonah

2

Hi Jopeter,

Welcome to the forums!

I think you can use the linked files to generate stls. The trick is a 3rd-party application called itk-SNAP. Here are the steps that worked for me just now:

  1. install and open itk-SNAP
  2. Download the annotation file and load it as a main image into itk-SNAP by using File->Open Main Image. Now I see:
    annot
    annot1704×1431 113 KB
  3. Download the segmentation file. Load it into itk-SNAP via Segmentation->Open Segmentation. I now see:
    segmentation
    segmentation1704×1431 119 KB
  4. Click update in the bottom left to show meshes derived from this segmentation. This results in:
    label_desc
    label_desc1704×1431 251 KB
  5. To save meshes out, go to Segmentation->Export as Surface Mesh. This opens up a dialog, with options to export one or all meshes. I chose “separate files”, and on the next page of the dialog changed the file format to stl.

As a sanity check, I then imported these stls into blender. Looks alright!

looks_familiar
looks_familiar1265×893 818 KB

Cheers,
Nile

3

Could not have been easier! Thank you immensely!

4

By the way, do you know if similar data files exist for the right hemisphere as well?

5

not able to download the annotation and segmentation link that you have provided above

6

Yikes! ^^^ Hopefully things have gotten better. This thread proves my point about the software ecosystem. In today’s world we should have something better than STL files, and on the other hand we shouldn’t have to download multi-gB h5 files just to get an astrocyte. We need to look ahead to what’s coming, we have an essential set of needs out of these files that aren’t being met. I should be able to take “any” neuron or astrocyte from “any” Allen or MICrONS database, and turn it into a watertight computational mesh on my PC by.pushing a button. This is absolutely possible with today’s technology and if it doesn’t exist let me know and I’ll build it. Quickly! Meshparty is lacking in basic functionality, it won’t heal a dirty mesh and it won’t even read OBJ files correctly. Finite elements on astrocytes are already here, there is a need to maintain compartments for ER and these formats aren’t giving us what we need. We need named separable objects and arbitrary data attached to each element (VTK can do that, kinda-sorta). These are basic needs for the workflow, the job can’t be done without them. (At least not in a reasonable time frame, without jumping through hoops and pulling teeth). It should be doable in Python without having to invoke a C compiler or download binaries.

7

Apparently this has been a well recognized problem for a long time.

These authors experienced the same silly set of problems I’m experiencing today (ten years later).

Pretty pictures don’t tell us very much, some of the power and popularity of what Allen’s doing is being mitigated by inconvenient software. Neuroscience and cell biology have peculiar issues, they don’t get addressed in other fields.

What is the purpose of having PyVista draw a mesh for you? It is computation, is it not? What else is a mesh good for? Surface extraction is computation, if you can do that you can get normals and boundaries and anything else you might need.