Dear Team FSL,
I came across the below post while trying to find suggestions for a transformation I want to run. Do the same rules apply if I am trying to transform a T2 FLAIR into diffusion space. Originally, I was going to try to transform the T2 to b0 space, but after reading the message board, it appears that using the FA map is the better option?
To give you a bit more context, I am looking to:
1. Transform T2 FLAIR into diffusion space (i.e., b0 or FA map)
2. Use the transformation matrix from (1) to move a binary lesion mask into diffusion space.
Thanks in advance for your time and wisdom!
Cristina
_____________________________________________________________________________________________
Hi again Anastasia,
I played a little with your data, and the following strategy seems to work
alright.
#
# Erode FA image to get rid of bright edge.
#
fslmaths dti_FA.nii.gz -kernel 3D -ero eroded_FA
#
# Crop T1-file to reduce FOV to where there is brain.
#
fslroi s03-2.3danat_flip_brain cropped_T1 14 140 45 180 40 145
#
# Register FA image to T1 image. Intuitively this would make sense
# since we want to be in undistorted space, i.e. in T1 space.
# Note that the associated .cnf file
# a. Considers zero as "missing value" in both --ref and --in
# b. Ends with subsampling 2 since the ~1mm res in the T1 is overkill
#
# Start with finding an affine transform
flirt -ref cropped_T1 -in eroded_FA -omat FA_2_T1_Affine -out
for_checking_how_it_went
# And non-linear
fnirt --config=FA_2_same_subject_T1 --aff=FA_2_T1_Affine --ref=cropped_T1
--in=eroded_FA --cout=FA_2_T1_warpcoef --iout=FA_warped_2_T1
where I have attached the .cnf file to this mail. The crucial thing here
(I think) is that by eroding the FA image I have removed the rim of high
values.
When looking at the warped FA image, it looks like it has done a quite
good job of correcting the distortions without introducing any new ones
(which would be a concern).
Note that if you want the "warped" version of the original FA (with the
rim intact) you need to run applywarp with the field you obtained above on
a cropped (to get same FOV) version of the FA image.
Having said that, this is NOT what fnirt is really written for. The fact
that it happens to work is just a side effect of the intensity modelling,
and I strongly suspect that this particular application will always be a
bit "hit and miss" and that one will frequently have to resort to "little
fixes" like the one above.
The better approach to correcting distortions is to use fieldmaps to
correct all the dwi's before further processing.
Good luck Jesper
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
|