Which MNI brain is used for the fsaverage aseg?

I apologize in advance for any terminology errors - I’m still new to neuroscience research. Please let me know if I can provide any information to clarify my question!

Hi all,

Does anyone know which MNI brain is used for the fsaverage aseg?

So far, I’ve come to the conclusion that the fsaverage aseg brain is not in MNI152NLin2009cAsym space. I overlaid the fsaverage aseg on a subject brain in MNI152NLin2009cAsym space using afni, but I found the subcortical structures do not line up. For context, the aseg brain was resampled to fit the resolution of the MNI152NLin2009cAsym subject brain, and the MNI152NLin2009cAsym subject brain is an output from fmriprep.

Likewise, does anyone know why the fsaverage aseg doesn’t overlap the fmriprep output for the MNI152NLin2009cAsym subject?


No MNI brain is used for fsaverage. It was developed in 1999 prior to the MNI 152 template in 2001, so it’s its own space. Many people who use FreeSurfer’s pipeline are interested in transforming their parcellation/annotations to a surface. In AFNI/SUMA, we take advantage of FreeSurfer’s registration to the FSaverage on a spherical surface by transforming all the surfaces and corresponding data annotations and parcellations to a common standard mesh. Using @SUMA_Make_Spec_FS with a selection of a standard mesh surface definition provides nodewise correspondence across all subjects.

For surface analysis, that is all that is typically required. To go between MNI in the volume and the standard mesh, you can map from surface to volume spaces. You can map for a particular subject’s surface to their own volume or from a surface generated for the MNI 2009c asymmetric template (generated by @ptaylor) to that template as a volume. This method allows any FS registered dataset to be remapped to any subject or template for which FS has generated a surface and particularly the sphere.reg dataset.

The MNI surface for MNI 2009c asymmetric template is stored here:

We have used this method to transform datasets from HCP’s Contee through FSaverage space into MNI. A method to transform the HCP Glasser atlas is described here using this method.


Briefly, we use @surf_to_vol_spackle to map data from the 2009c surface into the cortical ribbon mask or a dilated version of that maks. The program fills the mask first along the projection lines between the pial and the smoothwm surfaces and then fills remaining voxels with a non-zero mode value in a neighborhood around each voxel.

1 Like