weird overlap in bilateral '.gii' import.

[iPsych]

For same data, below code directly loading *.pial of freesurfer works fine.

lh_pial=lh.pial
rh_pial=rh.pial
files = [lh_pial, rh_pial]
b1 = BrainObj(files)

However, when try to load gii file as faces and vertexes, then merge like below

lh_pial=test_L.gii
rh_pial=test_R.gii
(vert_l, faces_l) = read_gii(lh_surf_mesh)
(vert_r, faces_r) = read_gii(rh_surf_mesh)

vert=np.vstack([vert_l, vert_r])
faces=np.vstack([faces_l, faces_r])
b2 = BrainObj('Custom', vertices=vert, faces=faces)

The result is broken 3D shape as attached. Is there any specific step required before make brain Obj?

[EtienneCmb]

Hi @iPsych ,

I think here you need to increment the faces_r with the number of faces in the faces_l because the number of vertices changed since it's concatenated over left and right hemispheres. Can you try something like this please :

(vert_l, faces_l) = read_gii(lh_surf_mesh)
(vert_r, faces_r) = read_gii(rh_surf_mesh)
faces_r += faces_l.shape[0]
faces = np.vstack([faces_l, faces_r])

[iPsych]

Hi, @EtienneCmb, It seems, the error happens as below. I think some number mismatch between lh and rh cause it.

IndexError: index 258612 is out of bounds for axis 0 with size 258108

[EtienneCmb]

And with this ?

faces_r += faces_l.max() + 1

[iPsych]

Perfect! it works : )

[iPsych]

@EtienneCmb, After successful mesh loading, annotation with parcellize comment also show strange result. I guess it is somehow related faces # just changed above? when annotation file loaded single hemisphere, just yellow brain or yellow brain with one region shows in purple.

b_obj.parcellize(lh_annot, hemisphere='left')

b_obj.parcellize(rh_annot, hemisphere='right')

[EtienneCmb]

Hi @iPsych ,

Indeed, it might comes from the faces concatenation. A possible workaround could be to specify the indices that belongs to the left and to the right hemispheres (checkout the doc of the input lr_index) :

vert = np.vstack([vert_l, vert_r])
# left = True / right = False
lr_index = np.array([True] * vert_l.shape[0] + [False] * vert_r.shape[0])
# define your BrainObj
b2 = BrainObj('Custom', vertices=vert, faces=faces, lr_index=lr_index)
# try again the parcellize
[...]

[iPsych]

I already put the code as below, and seems to do the same function.

r _index = np.ones((vert_l.shape[0],), dtype=bool)
r_index = np.zeros((vert_r.shape[0],), dtype=bool)
lr_index = np.r_[l_index, r_index]

Tested both

b_obj.parcellize(lh_annot)
b_obj.parcellize(rh_annot)

and

b_obj.parcellize(lh_annot, hemisphere='left')
b_obj.parcellize(rh_annot, hemisphere='right')

but the only difference was the warning.

WARNING | left hemisphere(s) inferred from filename WARNING | right hemisphere(s) inferred from filename

It seems the code load annotation properly, but cannot map it properly. If I re-call the parcelize code as below, both hemisphere shows the proper segmented regions, but with changed color.

b_obj.parcellize(lh_annot, hemisphere='left')
b_obj.parcellize(rh_annot, hemisphere='right')
b_obj.parcellize(rh_annot, hemisphere='right')

[EtienneCmb]

Ok, indeed, there's probably an issue for selecting the correct left and right indices used for the parcellization.

The original purpose of your issue changed, could you please close this issue and reopen a new one, specifically for the parcelization issue?

[iPsych]

Issue closed and new issue opened.