Cortex mesh and accompanying connectivity

1

Dear All,

I am a mathematician working on numerical methods for brain dynamics and, while I am numerically trained and can program in several languages, I have never worked first hand with a concrete dataset.

I must now test a numerical scheme on a model, using a realistic cortex, and with a somewhat realistic structural connectivity matrix. What I need is:

  • A triangulated surface mesh for a cortex with n vertices (so x,y,z coordinates for the vertices, and corresponding triangulations)

  • An accompanying structural connectivity matrix. Ideally, this would be an n-by-n matrix, potentially sparse, expressing some form of functional connectivity between each pair of vertices.

Do you know of any dataset giving access to the above in some standard format? I am aware that connectivity matrices can be hugely expensive to store, which is why I could compromise on resolution (n need not be huge).

Even synthetic data would be helpful for what I must do.

Also, I am aware that connectivities are sometimes expressed from region to region, rather than vertex to vertex. I would be happy with that, provided it would also be possible, with little (or possibly no) effort, to associate nodes to regions: in this way, I could take a “constant” connectivity matrix between group of nodes that correspond to the same region.

This seems quite a nontrivial but possibly standard thing to do, and I wonder if there is anything that I could use out of the box, or some dataset that serves as a benchmark. There is a plethora of available codes for displaying and manipulating this type of dataset, and I would not want to invest time into learning one specific tool if I can’t extract the dataset I need above.

Best

Daniele Avitabile
Associate Professor in Applied Mathematics
danieleavitabile.com

2

Hi All,

I was about to resubmit this post again, because after 2 years I’m still searching for a dataset that would allow me to test numerical schemes on realistic cortices. What I’m after is a triangulation of the grey matter (a surface mid-thickness manifold would suffice) together with an accompanying structural connectivity matrix that runs from vertex to vertex.

I understand that this may be nontrivial, but even a tutorial on how to do that would be useful at this point.

3

Hi @Daniele_Avitabile / @danieleavitabile ,

While I cannot think of any datset that has vertex-wise connectivity, the Human Connectome Project imaging data can be downloaded for free (HCP-Young Adult 2025) and has already processed FreeSurfer data (for surface meshes) and preprocessed DWI. The following paper postprocessed the HCP DWI to make vertex-wise structural connectivity matrices (https://www.sciencedirect.com/science/article/pii/S1053811920311800) , and you can try to use their code (neural-identity/codes at master · sina-mansour/neural-identity · GitHub).

EDIT: They also provide the vertex-wise connectivity matrices in their S3 repository: S3 Bucket Listing Generator

Best,

Steven

4

Thank you @Steven, this is very helpful