 |
 |
Hi all, I have a few questions regarding how FSL performs registration and also about Flame and would really appreciate any help you could offer! 1) When we do registrations in FSL, lets say aligning T1 scans to an MNI template, we are aligning all images to to establish voxel-wise correspondence, correct? So in the case of a T1 being smaller then the MNI template, a nonlinear registration is done and the image is stretched and fitted to the template. In this case, since the T1 was smaller and had less voxels then the MNI template, am I creating new voxels when I register the T1 to the template? To be able to do a voxel-wise comparisons between subjects, do we make it so that every subject have to have the exact same number of voxels? 2) In FMRI analysis, in order to control the FWER we must approximate the tail of the maxT distribution using Random Field Theory. In the FSL lecture slides and other tutorials the maxT distribution is shown to have a certain shape ( https://www.researchgate.net/profile/Anders_Eklund2/publication/267228188/figure/fig33/AS:[log in to unmask]). Is the maxZ distribution something we empirically measure during fmri/flame analysis or is it something that we do not know, but can approximate with RFT? Are we creating permutations during flame to get the maxT distribution? 3) For VBM, I understand that we use the t distribution when determining if voxels/clusters have significantly different gray matter between groups. If this is the case, then why do we use nonparametric permutation methods to approximate the null distribution? Are we assuming the null t distribution is not normal in the case of VBM and if so why? Thank you very much for your time and help! Cheers Sheena