Dear Robin,
See spm_templates.man
% The MRI scans were aligned with tpm/TPM.nii via the Segment algorithm
% of SPM12b (version 5979, with settings all at default values), which
% resulted in deformation fields that were used to warp the label maps
% to MNI-space.
So yes, they should be in 6th gen space, within the limitations of the Segment algorithm (with Dartel or Shoot one would probably be able to increase overlap between subjects to some extent still, whether this would be a relevant one, and/or how it is reflected in a maximum probability map, is another issue).
I guess the atlas is perfectly fine (I haven't used it myself), but be aware of the fact that its options are limited, as only an MPM is provided (which might be sufficient though). In case individual labels are available one can conduct "probabilistic" ROI analyses, that is weight voxels accordingly (which should be preferable). In case structural volumes are included one can re-preprocess, trying to increase the overlap 1) between atlas subjects (e.g. because better registration algorithms have become available) 2) between atlas subjects and study subjects. One can also apply smoothing accordingly (which should be preferable = smoothed labels for smoothed data, unsmoothed labels for unsmoothed data).
Best regards
Helmut
|