ValueError when importing subject from Freesurfer

[dangom]

Recently came across pycortex and decided to give it a try. I'm having troubles getting it to import data from freesurfer, though, as pycortex raises a ValueError, as shown in the backtrace below. Any suggestions on how to proceed would be welcome.

In [10]: import cortex.freesurfer as fs
In [11]: fs.import_subj("spat01", whitematter_surf="white")
Are you sure you want to overwrite this existing subject?
This will delete all files for this subject in the filestore, including all blender cuts. Type YES
YES
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-11-a8a458747a00> in <module>
----> 1 fs.import_subj("spat01", whitematter_surf="white")

/home/user/anaconda3/envs/cortex/lib/python3.7/site-packages/cortex/freesurfer.py in import_subj(subject, sname, freesurfer_subject_dir, whitematter_surf)
    174
    175     # Make the fiducial files. Just make them.
--> 176     make_fiducial(subject, freesurfer_subject_dir=freesurfer_subject_dir)
    177
    178     # Freesurfer uses FOV/2 for center, let's set the surfaces to use the

/home/user/anaconda3/envs/cortex/lib/python3.7/site-packages/cortex/freesurfer.py in make_fiducial(subject, freesurfer_subject_dir)
    312     """
    313     for hemi in ['lh', 'rh']:
--> 314         spts, polys, _ = get_surf(subject, hemi, "smoothwm", freesurfer_subject_dir=freesurfer_subject_dir)
    315         ppts, _, _ = get_surf(subject, hemi, "pial", freesurfer_subject_dir=freesurfer_subject_dir)
    316         fname = get_paths(subject, hemi, "surf", freesurfer_subject_dir=freesurfer_subject_dir).format(name="fiducial")

/home/user/anaconda3/envs/cortex/lib/python3.7/site-packages/cortex/freesurfer.py in get_surf(subject, hemi, type, patch, flatten_step, freesurfer_subject_dir)
    406         surf_file = get_paths(subject, hemi, 'surf', freesurfer_subject_dir=freesurfer_subject_dir).format(name=type)
    407
--> 408     pts, polys = parse_surf(surf_file)
    409
    410     if patch is not None:

/home/user/anaconda3/envs/cortex/lib/python3.7/site-packages/cortex/freesurfer.py in parse_surf(filename)
    328         print(comment)
    329         verts, faces = struct.unpack('>2I', fp.read(8))
--> 330         pts = np.fromstring(fp.read(4*3*verts), dtype='f4').byteswap()
    331         polys = np.fromstring(fp.read(4*3*faces), dtype='i4').byteswap()
    332

ValueError: string size must be a multiple of element size

[dangom]

It seems there is a readline too much in parse_surf. Removing the line fixes the issue for me. I'm on Python 3.7.3

[alexhuth]

Hi Daniel! Thanks for both posting this issue and finding a fix :) I was actually surprised upon seeing this that (1) we are still using our own code to parse freesurfer surfaces instead of nibabel (I think this module predates the introduction of nibabel as a dependency for pycortex), and (2) that we haven't run into this issue before.

Can you tell us what version of freesurfer you're using here? Also, is there something particular or strange about this surface?

Finally, can you confirm that the nibabel function freesurfer.read_geometry can successfully load your surfaces? If so, I think it may be best for us to just switch to using nibabel to read freesurfer surfaces.

[dangom]

Uhm, interesting.
The surfaces were reconstructed with Freesurfer version freesurfer-local-build-dev-20190709, and recon-all.log tells me that recon-all finished without error.

But I also cannot nib.freesurfer.read_geometry for the smoothwm surface, so I suppose there must be something wrong with it. The surfaces pial and smoothwm are "broken", but all others work. Freeview can read and display them without issues though.

I'll investigate further and report back when I find out what the issue is.

[dangom]

Related: freesurfer/freesurfer#699

[mvdoc]

Oh, thanks for the pointer @dangom!