The files have the same name but differ slightly in content. There are perhaps a few subjects for who the files differ greatly in content (see example in third paragraph), but I'm not sure how I can determine exactly what the extent of this difference is. I guess we could say that its a large enough difference to cause me some grief?
As an example of a "slight" difference: in the highres2standard_warp images, the same voxel (11 13 10) has an intensity of 3.00976 in one analysis (with multiple inputs) and an intensity of 2.96305 in the other analysis (with the single subject input). In contrast, if I re-run a single subject first-level FEAT analysis because I want to change a parameter outside of the registration (the registration parameters stay the same), the reg/ files are identical between the new FEAT analysis and the original. I used the linux tool md5sum to determine that the files are "identical", and also looked at a few images manually to compare the intensity values.
A little more background might be helpful to understand why this causes me grief: The reason that I noticed this difference in registration in the first place was because I was trying to extract signal change information from each subject's functional data with featquery, using a mask from standard space. When the mask was transformed back into native space, it appeared in a noticeably different location when the featquery was run on one feat directory versus the other. For some subjects the difference was extreme: for example, a mask from standard space V1 being transformed into native space V1 in one featquery, but into the cerebellum in the other featquery. I deduced that the problem was in the registration, since the reg/ files differed between the two Feat directories that corresponded to the same subject.
Please let me know if you require more information,
Thanks,
Martha
|