Dear SPMers,

I am trying to do some voxel based morphometry analyis using SPM12 and the DARTEL toolbox. I am using a longitudinal dataset of 3D images that were saved as nii. I registered the images at different time points for each animal using the serial longitudinal toolbox of SPM. This rendered avg 3D file for each animal, that I masked using AMIRA 5.4. I used the old segment tool to get the c1, c2 and c3 images. Then I use the seg_sn files for the initial import to get rc1, rc2 and rc3 images. While this initial import works well in half of my dataset, in the other half of my dataset this initial import step resulted in a large misalignment compared to the c1 files (see attachment). The upper two rows show animals where the rc1 image seems to be shifted upward. While in the bottom 2 rows the initial import did not give this problem. This large shift in my images gives of course problems when I try to create a segment template.

I am puzzled how the initial import works well in some of my data, but not in my other data. I cannot find a clear reason why this would be so different between animals. The prior steps were all the same in all animals.
I repeated the old segment and initial import step several times in the ones that failed, using spm to mask my data instead of amira, changing the bounding box in the initial import step... but this still gave me misaligned images. It seems that there is an extra transformation of my images in the initial import step. 
Is there a way to solve this problem? 
- can I use the c* files instead of the rc* files to calculate the segment template? 
As I only have this problem with the rc* files. 
- re-slicing my images to isotropic voxels before the old segmentation?
My images are not isotropic (voxelsize 0.9x2.5x2.5 mm) when I run the old segment, during the initial import I chose isotropic voxels of (0.9x0.9x0.9 mm). However, I am curious why this anisotropy would not be a problem in half of my dataset.  

kind regards 
Jasmien