Hi,
> I have a question regarding registration from functional scans to anatomical in native space.
> I would like to register my functional scans to my structural scan in native space, because I'm going to draw out retinotopic regions. Fsl seems like the best program to use for registration. However, when I use epi_reg or flirt with the bbr option, my functional files become very big (2.6 GB for 180 volumes, before registration they are around 35 MB).
This is to be expected given the higher resolution of your structural scans.
> One reason might be that the functionals, which are 2.5 mm isotropic, might be upsampled to 1 mm isotropic to match the structural. However, an additional problem might arise that I don't understand. I have tried to do single subject statistics on these big data files. It takes a very long time, even for one run, and the outcome is strange: a lot of extra 'empty' data is added to the beginning and end of my functionals. I would like to add a screenshot here, but I don't know how to do this, since it will only paste text.
I’m not quite sure what you are referring to, but running on huge files is very challenging and is not necessary. Why not transfer information from your anatomical image into functional space? Or transfer the results of the native-space functional analysis into anatomical space? That’s what our tools are setup to do.
> What it looks like is as follows: normally, on runs that are not aligned to the structural (small files, 35 MB), I can view the report.html of my post-stats image in the screen easily: that is, the whole brain fits the screen and I don't have to scroll down to see all the slices.
> When I do my stats on the large, registered, epi, however, the zstat images cannot be displayed on one screen: if I open report.html, I have to scroll down to see the whole image. It seems like an additional 20 empty slices are added to the top and bottom of my functional.
Oh, now I understand. These are just blank slices due to the extra FOV that is in your anatomical space. As I said, we would normally not recommend upsampling your functional data, but instead transforming the analysis results.
> Does anybody have an idea what is going on and how I might fix this? I am scripting my whole pipeline, and so would like to avoid using the GUI at all. Below is the line of script I'm using for registration. Before registration, functional files are around 35 MB, and structural 2MB.
If you stick to doing the analysis in the native functional space then I think things will be simpler and you can then either transform results to the anatomical space, or information from the anatomical space into the functional space.
All the best,
Mark
> Thank you in advance,
> Annelinde
>
> CODE:
> epi_reg -v --epi= func_prealign --t1= struct --t1brain = struct_bet --out=func_align
>
> flirt gives the same outcome:
> flirt -in func_prealign -ref struct_bet -out func_align-interp spline -wmseg struct_fast_wmseg -cost bbr -dof 6 - applyxfm
>
> ps. If anybody wants to see the screenshots I'd be happy to email them to you directly
|