How to load merged hemisphere dset in SUMA?

AFNI Message Board
1

Dear afni experts,

I’m using AFNI/FreeSurfer/SUMA tool chain, and CoSMoMVPA matlab package to do surface-based MVPA, and encountered a few problems. Any help would be highly appreciated.

  1. According to the CoSMoMVPA demo, its surface-based searchlight routine seems to require merged hemisphere (mh, as opposed to lh+rh) format of the FS surface (in *.asc file). However, @SUMA_Make_Spec_FS generates separated (lh+rh) files. I used the following pseudo matlab code to merge the two:
    [left_verts, left_faces] = read_asc(‘lh.asc’);
    [right_verts, right_faces] = read_asc(‘rh.asc’);
    right_faces(:,1:3) = right_faces(:,1:3) + size(left_verts,1);
    write_asc(‘mh.asc’, [left_verts; right_verts], [left_faces; right_faces]);
    It seemed to work (but is it the correct way to do this?). But when I visualize the merged surfaces, I realized that the spatial location of the vertices from the two hemispheres in volumetric space are overlaped. Is that OK? Specifically, will that affect neighborhood definition or multiple comparison correction?

  2. After I got the surface-based searchlight result with CoSMoMVPA in a *.niml.dset file, which contains the results for both hemispheres, I tried to load and visualize it with SUMA. However, with SUMA object controller, I could load the dset onto only one of the two hemispheres (either left or right, but not both). When I loaded the same dset to the second hemisphere, SUMA complained and failed.
    Is there a way to visualize merged hemisphere dset with SUMA?

Thanks~
Chencan

2

I have no experience with that CoSMoMVPA software, but you could try using ConvertSurface -merge_surfs to create a merged version of the surfaces.

3

Thanks, Daniel~

But if I have a single *.niml.dset that contains vertices and associated surface data for both hemispheres, how can I load and visualize it in SUMA?

4

I see. You can convert the dset into a simple 1D file and then split that file. Convert niml.dsets with ConvertDset.

5

Greetings,

I’m the main author of CoSMoMVPA. Hopefully my responses below are helpful.

herrlich10 Wrote:

I’m using AFNI/FreeSurfer/SUMA tool chain, and
CoSMoMVPA matlab package to do surface-based MVPA,
and encountered a few problems. Any help would be
highly appreciated.

  1. According to the CoSMoMVPA demo, its
    surface-based searchlight routine seems to require
    merged hemisphere (mh, as opposed to lh+rh) format
    of the FS surface (in *.asc file). However,
    @SUMA_Make_Spec_FS generates separated (lh+rh)
    files. I used the following pseudo matlab code to
    merge the two:
    [left_verts, left_faces] =
    s] = read_asc(‘lh.asc’);
    [right_verts, right_faces] =
    s] = read_asc(‘rh.asc’);
    right_faces(:,1:3) = right_faces(:,1:3) +
    • size(left_verts,1);
      write_asc(‘mh.asc’, [left_verts; right_verts],
      ts], [left_faces; right_faces]);
      It seemed to work (but is it the correct way to do
      this?).

Yes, that should be fine.

But when I visualize the merged surfaces,
I realized that the spatial location of the
vertices from the two hemispheres in volumetric
space are overlaped. Is that OK?

You can add an offset to the coordinates, for example to the first column in left_verts and right_verts to add or subtract some value from the x coordinate.

Specifically,
will that affect neighborhood definition or
multiple comparison correction?

It should not really, as neighborhoods are defined through the faces. So even if surfaces from two hemispheres overlap spatially, there are no two nodes with one in each hemisphere that are neighbors.

  1. After I got the surface-based searchlight
    result with CoSMoMVPA in a *.niml.dset file, which
    contains the results for both hemispheres, I tried
    to load and visualize it with SUMA. However, with
    SUMA object controller, I could load the dset onto
    only one of the two hemispheres (either left or
    right, but not both). When I loaded the same dset
    to the second hemisphere, SUMA complained and
    failed.
    Is there a way to visualize merged hemisphere dset
    with SUMA?

It should work fine to load the merged hemisphere directly (i.e. using an mh.asc file as you defined earlier). It may require a custom .spec file that points to the merged hemisphere files, not left or right hemispheres. How does SUMA complain?

As a side note: PyMVPA’s pymvpa2-prep-afni-surf is a script that helps with merging hemispheres, writing spec files, and more. It takes as input the result from FreeSurfer’s recon-all. See http://www.pymvpa.org/generated/cmd_prep-afni-surf.html for details.

6

Hello Nick

I have the same problem. I did a surface-based searchlight in CoSMoMVPA [1]. The result is a niml.dset file for both hemispheres (ld 141). I tried to map the result on suma_TT_N27 [2] using the following command:
suma -spec std.141.MNI_N27_both.spec -sv mni305.nii -input searchlight.niml.dset

But I got the following error message:
– Error SUMA_OKassign (SUMA_Color.c:10134):
Node index range outside
0 and SO->N_Node
++ Notice SUMA_LoadDsetOntoSO_eng (SUMA_Color.c:10312 @21:59:05):
Could not assign dset to SO.
Trying to assign to domain parent.
– Error SUMA_OKassign (SUMA_Color.c:10134):
Node index range outside
0 and SO->N_Node
– Error SUMA_LoadDsetOntoSO_eng (SUMA_Color.c:10315):
Cannot assign dset to SO
or its local domain parent
– Error suma (SUMA_suma.c:1161):
Failed to load searchligh.niml.dset onto std.141.lh.smoothwm.gii

I think the problem is that there is no std.141.both.smoothwm.gii in suma_TT_N27.

I think I should split searchlight.niml.dset to the left and right hemispheres but I don’t know how to do it!

[1] https://www.cosmomvpa.org/_static/publish/run_surface_searchlight.html
[2] https://afni.nimh.nih.gov/pub/dist/tgz/suma_TT_N27.tgz