Hi Carsten,
I think that your bounding box (volume to be written relative to the anterior commissure) is too small and that's why it ends up cutting off small parts of the brain. You can change the bounding box size at the normalization step and that should solve the problem.
About CFM although the paper by Crinion et al (2007) states that Unified Segmentation using medium regularization does not need additional masking (I think that Seghier references this paper when saying that no additional mask is needed), there is a recent paper by Andersen et al(2010) (http://www.ncbi.nlm.nih.gov/pubmed/20542122) which shows that, at least with a group of chronic stroke patients, not adding a mask reduces normalization accuracy and lesion volume.
I hope this helps!
Pablo
2011/7/12 Carsten Finke
<[log in to unmask]>
Hi Pablo,
many thanks for you answer! I have attached a screenshot - there you can see that (small) parts of the temporal lobe and the cerebellum are cut off.
And I actually chose the Segmentation and Normalization approach as a paper by Seghier et al. (2008) states that with this method additional lesion masking will bring no further benefit (hence, is not necessary)... The normalisation does indeed work great, just the little bit of the temporal lobe is missing...
carsten
