As detailed in issue #3415 opened in nilearn, vertex order does not match across all fsaverage resolutions.

I downloaded fsaverage 3 to 7 tarballs for the git annex of this repo and simply ran the following script to show that vertex order is not preserved across resolutions:

# %%

from nilearn import surface

# %%

fs3_path = "/home/alexis/singbrain/data/freesurfer_fsaverage/fsaverage3/surf/lh.pial"

fs5_path = "/home/alexis/singbrain/data/freesurfer_fsaverage/fsaverage5/surf/lh.pial"

fs7_path = "/home/alexis/singbrain/data/freesurfer_fsaverage/fsaverage7/surf/lh.pial"

fs3_coordinates, fs3_faces = surface.load_surf_mesh(fs3_path)

fs5_coordinates, fs5_faces = surface.load_surf_mesh(fs5_path)

fs7_coordinates, fs7_faces = surface.load_surf_mesh(fs7_path)

# %%

print(fs3_coordinates[:5])

print(fs5_coordinates[:5])

print(fs7_coordinates[:5])

Output:

[[-3.87359581e+01 -1.93433647e+01  6.72201385e+01]

 [-9.71759605e+00 -9.23328686e+00  4.65803757e+01]

 [-2.40194092e+01  4.31147995e+01  2.39262428e+01]

 [-5.98615112e+01  4.93119545e-02  8.97464848e+00]

 [-5.06454735e+01 -4.94050064e+01  4.78137321e+01]]

[[-3.87359581e+01 -1.93433647e+01  6.72201385e+01]

 [-1.66624870e+01 -6.90612259e+01  6.12812729e+01]

 [-9.71761799e+00 -9.23326397e+00  4.65803413e+01]

 [-2.40194283e+01  4.31148720e+01  2.39262142e+01]

 [-5.98614998e+01  4.92927618e-02  8.97456360e+00]]

[[-3.87359581e+01 -1.93433647e+01  6.72201385e+01]

 [-1.66624680e+01 -6.90612564e+01  6.12812119e+01]

 [-9.71762943e+00 -9.23325062e+00  4.65803223e+01]

 [-2.40194435e+01  4.31149139e+01  2.39261990e+01]

 [-5.98614883e+01  4.92815673e-02  8.97451496e+00]]

Here, we see that the second vertex of fsaverage3 doesn't match that of fsaverage 5 and 7.

I did some more extensive testing and it seems to me that:

fsaverage 5, 6, 7 have consistent vertex order

fsaverage 3 and 4's vertex order is different from 5, 6, 7 but can be reworked to be consistent, using a procedure I detailed in nilearn issue #3415 (MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be https://github.com/nilearn/nilearn/issues/3415)

Maybe we should reorder fsaverage 3 and 4's vertices for all maps, what so you think?

Thank you for the great software anyways! ☺️

Warm regards,

Alexis