Problem loading the HCP anatomy data and the Deskian-Killiany Atlas

andylikeswater · February 24, 2022, 4:33am

Hi Brainstorm experts,

I am trying to run some source localization on the HCP MEG data and calculate the source connectivity using the Deskian-Killiany atlas.

In HCP, there are two different sets of anatomy that I can load into Brainstorm directly. One is the HCP preprocessed anatomy, and the other is the Freesurfer files.

I will need to load the HCP preprocessed anatomy because the MEG signals that I will be using also went through the same preprocessing pipeline and this allows me to have auto co-registration.

However, it seems like the Deskian-Killiany atlas (the *h.aparc.annot) is only defined in the Freesurfer surface that cannot be directly loaded onto the HCP preprocessed anatomy surface. These atlas can be automatically loaded and map to the cortical surface if I load the Freesurfer anatomy, but doing so does not allow me to do the co-registration with the sensor data. I am assuming the Freesurfer folder alone is missing some important information for auto co-registration because it didn't go through the HCP preprocessing pipeline.

Here is the co-registration with the HCP preprocessed anatomy on subject 100307:

Here is the co-registration with the Freesurfer anatomy on subject 100307:

You can see that these two are different. I am assuming that the HCP preprocessed one is more accurate.

Is there a way to work around this? Maybe somehow map the parcellation in Freesurfer surface to HCP preprocessed surface?

Thank you,

Francois · February 28, 2022, 8:27am

Unfortunately, I don't think there is any solution for that, as the information necessary for computing the coregistration between the original anatomy and the MEG was deleted for deidentification purposes.
See my notes in the HCP tutorial, in the dataset description and the section "Import the anatomy".
https://neuroimage.usc.edu/brainstorm/Tutorials/HCP-MEG

It could be possible to compute a transformation by fitting the head surface computed by Brainstorm from the FreeSurfer folder, and the one from the HCP FieldTrip-processed anatomy, and then apply it on the FreeSurfer cortex surface. But Brainstorm does not provide tools for doing that.

Or maybe there is a solution for computing directly a transformation to convert the FreeSurfer cortex into the same coordinate system as the FieldTrip surfaces, but I'm sorry I won't be able to guide you through this process. You can look for inspiration in the HCP import code:

github.com

brainstorm-tools/brainstorm3/blob/master/toolbox/io/import_anatomy_hcp_v3.m#L160

    
      
          % Convert transformations from "Brainstorm MRI" to "FieldTrip voxel"
          Tbst2ft = [diag([-1, 1, 1] ./ sMri.Voxsize), [size(sMri.Cube,1); 0; 0]; 0 0 0 1];
          % Set the MNI=>SCS transformation in the MRI
          Tmni = transform.vox07mm2spm * Tbst2ft;
          sMri.NCS.R = Tmni(1:3,1:3);
          sMri.NCS.T = Tmni(1:3,4);
          % Compute default fiducials positions based on MNI coordinates
          sMri = mri_set_default_fid(sMri);
          
          

          
%% ===== MRI=>SCS TRANSFORMATION =====
          % Set the MRI=>SCS transformation in the MRI
          Tscs = transform.vox07mm2bti * Tbst2ft;
          sMri.SCS.R = Tscs(1:3,1:3);
          sMri.SCS.T = Tscs(1:3,4);
          % Standard positions for the SCS fiducials
          NAS = [90,   0, 0] ./ 1000;
          LPA = [ 0,  75, 0] ./ 1000;
          RPA = [ 0, -75, 0] ./ 1000;
          Origin = [0, 0, 0];
          % Convert: SCS (meters) => MRI (millimeters)

You can try asking the FieldTrip developers, who prepared the HCP-MEG release, maybe they have an easy solution for this coregistration question.

Other option, longer but simpler: run CAT on the MRI from the Brainstorm database:
https://neuroimage.usc.edu/brainstorm/Tutorials/SegCAT12

Good luck