On Jan 18, 2008 10:02 AM, Jiansong Xu <
[log in to unmask]> wrote:
Dear Friends:
I'm learning how to analyze DTI data with FSL(FDT), and I'm wondering
if any one can make comments about my procedures described below.
My DTI data were acquired with a Simens Trio 3T scanner. 32
directions. b=1000
(I) Use MRIconvert to convert the DICOM data into NIFTI format for
DTI. The conversion created
1. one textile file with bvalues
2. one textile file with bvecs
3. one image file with b0
4. 32 image files with b1000
My question here is: Do I need to do any thing with the number in the
bvalues and bvecs file before using the two files, e.g., normalize?
With MRIconvert you need to change all the bvalues except the first one from zero to 1000. I believe the program does not do this for you, so you probably have 32 zeros in the original bvalue file. Depending on your scanner, you may need to flip the signs in a row of your bvecs. You will have to check the output after fitting the tensors. There are a few posts on this list about this if you need help with it, I ran into the same problem a while back.
(II) Use fslmerge to merge all image files into 4D file. The 4D file
has 33 images, and the b0 image is the first one. Now I remove all 3D
image files from the directory, so that the directory has 3 files
left, one for bvalues, one for bvecs, one for 4D images.
(III) Rename the three files, so that their names are: set00_bvalus,
set00_bvecs, set00_data.nii.gz
My question here is: how strict is this rule for file naming?
(IV) Use FDT to do Eddy current correction. The input file is
set00_data.nii.gz, the output is set00_corr.nii.gz
(V) Use BET to create binary mask. The input file is set00_corr.nii.gz
(VI) Use DTIFIT Reconstruct diffusion tensors to create FA image. The
input file is set00_corr.nii.gz.
Sorry, I am not familiar with this naming system and I don't know if bedpostX will recognize those files automatically, but I believe it would be easier if you used the following convention. Otherwise, you could specify the individual files using the fdt GUI.
bvecs
bvals -Remember to change those zeros to 1000s
data.nii.gz -This is your eddy current corrected diffusion image
nodif_brain_mask.nii.gz -This is the mask from the BETed no diffusion image
Now I get dti_FA.nii.gz file. Is this file have the FA values?
Yes, this is your FA map for that subject. You can open it in FSLview and see the FA intensities at each voxel. You can check this by looking at the ventricles, which should be close to zero, and the corpus callosum, which should be in the 0.7-0.8 range.
If I want to do ROI analysis, how can I extract the FA values after
the ROI is defined, which FSL should I use?
Draw a binary mask in FSLview and then multiply it by your FA image. This will zero all the voxels in the FA image that are not included in your mask. The command would be something like this: fslmaths FA_image -mul mask_image output_name
You can then do various things with the output, such as find the mean FA value: fslstats output_name -M
Your comments are highly appreciated.
Jiansong