[pymvpa] Surface-based Searchlight?

[Nick Oosterhof]

Hi Matthias,

On Nov 28, 2013, at 6:00 AM, Matthias Ekman wrote:

> I'm currently trying to convert my Freesurfer (FS) surfaces to afni; the provided scrip ``prep_afni_surf`` is very helpful.

I assume you refer to pymvpa2-prep-afni-surf wrapper script in the PyMVPA bin directory? 

(There is a module lib_prep_afni_surf in mvpa2/support/afni that provides the actual implementation, but it is preferred to use this through the wrapper script).

Also please note that the most recent code is available here: https://github.com/nno/PyMVPA

> 
> However, in my case it fails as recent FS (v5.3) doesn't seem to produce the
> COR files anymore? ("grep: COR-.info: No such file or directory"). Running ``@SUMA_Make_Spec_FS`` with the "-use_mgz" flag however works. Maybe this is a more sensible default, as the FS wiki marks COR as deprecated [2]. 

Thanks for bringing this up. I may change the script to use use_mgz then, as long as it keeps working with older versions of FS.

> 
> Q1: Do I understand correctly that the EPI image must be coregistered to the anatomical before running ``prep_afni_surf`` (i.e. afni is not doing that 'magically' otherwise)?

You can coregister surfaces to either an anatomical or EPI image ('-a' or '-e' options). If you have both, you should coregister them yourself, for example using AFNI's align_epi_anat.py. 

If you analyze data in original (not template) space, one option is to coregister anatomical files and surfaces to the EPI image. This reduces the amount of interpolation applied to the EPI image.  

> Q2: How would one save/transform the results back to a volume format? E.g. the ``fn = 'ico%d-%d_%sh_%dvx.niml.dset' % (lowres_ld, highres_ld, hemi, radius)`` in the surface-based Searchlight example [3]. Afni seems to have "3dAFNItoNIFTI" but that doesn't seem to work with SUMA files; I'm a bit lost here.

AFNI provides 3dVol2Surf and 3dSurf2Vol to transform results between volume and surfaces. 

best,
Nick