Dear John and SPM users.
I would like to inquire if there is any available data showing the average registration error by DARTEL?
I am currently running VBM using DARTEL-processed modulated images. I found that in group comparison, the smaller extra smooth kernel I put on the modulated images (e.g. 8mm --> 4mm --> 2mm --> 1mm --> no extra smooth), the more significant clusters identified using topological FDR. Note that even with no extra smoothing, the effective smoothness of the images is about 3.8mm (voxel size is 1mm isotropic). I am not sure what smoothness should I go for so would like choose it according to the registration error by DARTEL (in mm), if there's any data available. Any idea or suggestion regarding an appropriate smoothness for DARTEL-processed modulated images with 1mm isotropic voxels when running VBM?
Thanks so much.
Meng-Chuan Lai
|