Dear Ben,
The expression is correct. Please make sure which atlas/toolbox version you have installed, in v2 onward the range for the probabilistic maps is 0 to 1. There's also a new, 4th subregion AStr (haven't looked into the papers, stria terminalis?). The origin is still shifted though, thus if you want to create masks based on nii files in Anatomy toolbox shift them by [0 4 -5] to align to "SPM" MNI space.
If you go with threshold 0.5 and maps from v2 then AStr and CM masks will indeed contain (almost) no voxels due to low overlap. The section "Measures of consistency between subjects" in Saygin et al. (2011, Neuroimage) might be interesting in that context. Their simple T1 normlisation results in poor overlap between subjects, segmentation-based normalisation should be better but possibly still suboptimal. Not sure which algorithms were used for maps creation in Anatomy / how v2 differs from v1, I remember one paper stating something about "linear and non-linear elastic" transformations.
In principle no threshold shoud be preferable for ROI analyses, with voxels weighted by their corresponding probabilistic values to obtain average responses for certain regions.
Best
Helmut
|