Dear Elisabeth,
> the SPM segmentation without the cat12 toolbox always allowed me to simply put in the raw images without any prior "noise" reduction steps
This could well be the case. In fact, it would correspond to observations described in the following two papers:
Streitbürger et al. (2014, Neuroimage, http://dx.doi.org/10.1016/j.neuroimage.2013.10.051 ) segmented MP2RAGE images with SPM8 Segment, SPM8 New Segment, SPM12b Segment and state that the initial masking is not required for SPM12b, although I'm not exactly sure how they come to that conclusion. Maybe based on the comparison of SPM12b segmentation of the masked and unmasked images (reported in "Signal-intensity distributions and segmentations"), resulting in very similar intracranial volumes.
Seiger et al. (2015, Neuroimage, http://dx.doi.org/10.1016/j.neuroimage.2015.03.019 ) segmented MP2RAGE images with VBM8, stating that a masking was not necessary ("Although the ‘uniform’ 7 T MP2RAGE image exhibits a high amount of artificial background noise, the segmented gray matter volumes delivered good results and segmentation quality was not affected. Hence, the algorithm performed robustly despite the noisy background and therefore no binary masking had to be applied").
However, at least in theory it could be the case that unmasked images lead to a more noisy segmentation (due to the background noise), with e.g. some GM voxels misclassified as CSF and CSF voxels misclassified as GM. Thus one would probably have to provide some additional information on how the "goodness" was evaluated, especially looking at local misclassifications (which could cancel each other within certain anatomical regions and/or across subjects).
Best
Helmut
