Conversion of Glasser et al.(2016) atlas to each individual's surface space?


There is a need for the volume version atlas of Glasser et al.(Nature, 2016). However, due to some caveats raised by the contributors, I now target to convert the atlas to each individual’s surface.


I found a code for converting the volumetric version of Glasser et al.(2016) on the MNI2009a normalization space on the website Andreas Horn posted.

While the code makes use of the annotation files a user projected on the fsaverage space, I am now lost in how to proceed with each individual’s surface space. I’m currently using the FreeSurfer 6.x version implemented in Fmriprep 20.2.0.

Any suggestions how to perform the conversion? Any advice will be appreciated.

From what I understood, you need to move a .annot (Glasser .annot) from fsaverage space to your native surface space, for each participant.
In that case, Freesurfer has a tool to do that!
Check the mris_apply_reg (–help) function. It allow to input an annotation file, and a registration file (-lta).


1 Like

I really appreciate your help! I tried the function and it seems to receive the input annotation file(–src) and source-target registration pair(–streg srcreg1 trgreg1).

So by LTA you may mean the linear transformation from the fsaverage to individual surface and vice versa, provided by the Freesurfer. But where can I find the files? The things I could find in the Fmriprep output folder are the following two files:

  • sub-xxxx_ses-1_from-fsnative_to-T1w_mode-image_xfm.txt
  • sub-xxxx_ses-1_from-T1w_to-fsnative_mode-image_xfm.txt

which I guess have little to do with the transformation between two surfaces. Thank you.

Hummm Im not that familiar with fmriprep, nor I do have a freesurfer with me right now.
The registration matrix you are looking for, should be WITHIN the subject-freesurfer folder.

If not, you can compute it by yourself. Take a loot to mri_cvs_register, it might help you.


1 Like

While the HCP Glasser atlas was originally developed for surface analysis, it is now available for the volume in a couple ways now. That old script you referenced doesn’t really work for moving to MNI space. The HCP Glasser atlas is now distributed with AFNI by default and as NIFTI datasets aligned with the MNI 2009c asymmetric template. The atlas is described here:,160996,160996#msg-160996

Please let us know if you have any troubles using it by posting either here or on the AFNI message board.

1 Like

@moreblue, can I ask you to elaborate more on why you want to do this, and what exactly is your expectation?

1 Like

@dglen I really thank you for updating me. I read the descriptions, which were really helpful, including how to convert the atlas to the MNI surface.

Please bear with me asking one more question here: will it be a tolerable practice to normalize the volumetric atlas on MNI2009c to another common volumetric space or individual’s volumetric T1w space?

@oesteban Sure, I’m currently analyzing functional MRI signals preprocessed by fMRIprep with the output volume spaces of T1w and MNI152NLin6Asym, along with its recon-all option by default. Our team collected the data by ourselves.

One of my colleagues recommended using Glasser et al.(2016) atlas on the volume and I came across this forum post where it is recommended to use the “single-subject-specific” alignment to individual’s own surfaces.

So what I want to do is i) to convert the fsaverage-projected atlas to individual’s surface space and then ii) to convert the atlas on each individual’s surface space to each individual’s T1w(or another) volumentric space with the corresponding resolution of fMRI.

What I expect is a better alignment of the atlas to an individual’s volume space of functional mri. Actually there are little grounds for such expectation, so my original intention was to directly compare the results from this surface(average)->surface(individual)->volume conversion, to the publicly available surface(average)->volume conversion.

One way to use this atlas is to use the standard meshes generated by @SUMA_Make_Spec_FS applied to surfaces generated by FreeSurfer. The resulting datasets allow the standard mesh annotations in a niml.dset to be used on individual subjects. The standard mesh provides a node-wise correspondence across subjects. Standard mesh datasets are available for the MNI152_2009c asymmetric template and for the MNI N27 template. The HCP atlas is available for the surface for a low and high resolution standard mesh, in the same way that individual subject surfaces and parcellations are available. Using these standard meshes, one can either analyze the data on the surface or project from the surface into the volume (@surf_to_vol_spackle, 3dSurf2Vol).

1 Like