 |
 |
The number of cosine transform bases depends on the FOV of the image to be segmented - rather than the FOV of the template. This is because it estimates a mapping from the native image to the tissue probability maps - rather than the other way around - which is what the Normalise module does. Because the mapping is the other way around from the mapping required to write spatially normalized images, the deformation needs to be inverted. These inverted warps are stored according to their DCT coefficients. Because the number of basis functions required to generate the inverse mapping is not identical to the number of bases needed for the original, I add a few more coefficients so that a bit less of the information is lost. This is why the seg_sn.mat files have more coefficients than the seg_inv_sn.mat files. Best regards, -John -----Original Message----- From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Joao Pereira Sent: Wednesday, May 23, 2007 7:00 PM To: [log in to unmask] Subject: [SPM] Number of DCT basis functions Hi, I'm trying to better understand the implementation of the Unified Segementation rotine in SPM5 and I've been unable to figure out how the number of DCTs is computed. I understand that the default is 7x8x7, which fits the 91x109x91 dimensions of the prior volumes. Nonetheless, when I run spm_preproc.m, I get a Twarp of 9x7x9(x3), which fits (given the cutoff of 25mm) the 256x124x256 (raw) volume of input. Is this supposed to be like this? Is the reference 7x8x7 outdated when talking about SPM5? Finally, when I run spm_prep2sn (and then spm_preproc_write), the resulting sn.mat file will have a transformation Tr 11x9x11(x3) because of the reparametrisation. What does this final value mean? Thank you for your help! Regards, Joao Pereira