Hi Nir,

I'm assuming you want to get a binary mask for one or more ROIs, right?

The first step is to look up the regions that you want and get their numbers; you can do that interactively using tkmedit or freeview, or you can find the region names here:
  https://surfer.nmr.mgh.harvard.edu/fswik...
E.g. if you want the left and right caudate the numbers are 11 and 50.

Consider what space you need the regions in: native (as acquired) or FreeSurfer's "conformed" (256 cubed, 1mm cubic voxels) space. For example, if you plan to do any further work (e.g. coregistration of a functional image, etc.) using FreeSurfer's T1.mgz, then you can work with the FreeSurfer space aparc+aseg, but if your other work is based on the original pre-FreeSurfer image, then you'll want to convert the aparc+aseg to native (rawavg) space. If you need native, see the mri_label2vol command here:
  https://surfer.nmr.mgh.harvard.edu/fswik...

You can use FreeSurfer's matlab function MRIread on an mgh/mgz file directly (without converting to NIfTI) or you can use it on a converted nii.gz. Then you can get a binary mask using the matlab function ismember together with a vector of your region numbers (e.g. 11 and 50 for the caudate, as noted above), i.e.
  img = MRIread('aparc+aseg.mgh');
  msk = ismember(img.vol, [11 50]);

Finally, you can write out the binary mask using MRIwrite, e.g.
 img.vol = msk;
 MRIwrite(img, 'caudates.nii');

I hope that's useful, but if you want further help, you will probably have more success on the FreeSurfer mailing list:
  https://surfer.nmr.mgh.harvard.edu/fswik...

All the best,
Ged