Hi fsl experts:
I don't know if it is my fault or not when running fsl, but it spend too much time on doing the same thing several times, so are there any way to reduce this wasted time?
Here is the details. You know, I have one high resolution T1 image, one low resolution T1 image, and several functional images for one subject. Then within FEAT, firstly I did the `first-level preprocessing` on all those functional images, in the `Registration` tab, I choose the low resolution T1 image as the `expanded functional image`, the high resolution T1 image as the main structural image.... Then FEAT will do the following registration for all those functional images:
+ low_res to high_res (inital_highres2highres)
+ high_res to standard (highres2standard)
+ inverse, (standard2highres, highres2inital_highres)
+ functional to low_res (example_func2initial_highres)
+ then apply different transforms on this functional image, there are example_func2highres, example_func2standard et al.
OK, my question is, because for all those functional images, the inital_highres and highres and standard image are all the same, why must we run those steps again and again while it CAN be done only once? I've checked the output (.mat, .nii.gz), they are identical.
I've tried to copy those files into the .feat/reg folder, but FEAT will generate a `.feat+` folder instead of reusing the `.feat` folder, so this direct way just don't work.
And BTW, FSL runs too slow, multi-core does NOT help, fslmaths, flirt et al., can use only one thread.... Any suggestion on speed up FSL?
Thanks.
|