The reason for the discrepancy is because the _seg_sn.mat files get corrected using a Procrustes method. This decomposes the spatial transformation into a rigid-body transform and a nonlinear deformation. The idea of this hidden feature was to allow morphometry to be done more easily in the nonlinear deformations. The affine transform matrices are actually rigid-body, but taking account of the voxel-to-world mappings of the source image and template, so they become a voxel-to-voxel mapping. Best regards, -John On Friday 14 April 2006 00:29, Hedok Lee wrote: > Dear SPMers > > I’ve been trying to normalize SPGR T1W images (256x256x106 > 0.94mmx0.94mmx1.50mm) using SPM5 VBM segmentation function > (native+modulated+unmodulated). I use matlab 7.1 in Linux with the update > 456. > > I wanted to compare the differences between “with” and “without” non-linear > transformation, so I turned on non-linear transformation by setting DCT > cutoff=25 and turn it off by setting it to DCT cutoff=inf. All of other > parameters are set to the default values. I was expecting affine > component, which was extracted from ”_seg_sn.mat”, to be the same between > the two ways, but they turn out to be different. > > I appreciate if anyone has an explanation for this. > > ”_seg_sn.mat” > > DCT cutoff=25 > 2.1310 -0.0608 0.0777 28.1888 > 0.0796 2.0519 -0.5784 15.4785 > -0.0364 0.3629 1.2825 -13.4118 > 0 0 0 1.000 > > DCT cutoff=inf > 2.1315 -0.0478 0.0745 27.6491 > 0.0660 2.0546 -0.5704 15.9588 > -0.0369 0.3577 1.2839 -12.8985 > 0 0 0 1.000 > > Thanks, > > Hedok Lee