Hi,

Please find comments in the text.

2014-09-24 13:38 GMT+02:00 Francesco Puccettone <[log in to unmask]>:

Hi Chris; thanks for your answers. Some follow-ups are below.

1. see screenshot attached, in which (part of) the brain's outline is visible in the skeleton

This is the threshold which you gave in the tbss_4_prestats

You can read here: http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/TBSS/UserGuide (under tbss_3_postreg)

Normally 0.2 is proposed but depending on the data it might be higher or lower.

3. I know. My question was: why are they not in a single file and instead are split in 3 files of each type, corresponding to 21+22+25 diffusion directions separately. For instance, I have DTI21.bval, DTI22.bval and DTI25.bval files, and, for images, DTI21.nii.gz etc

I am a little confused by what you describe. Normally you have 3 files: 1) data.nii.gz and 2) bval and 3) bvecs.

How did you get your data? Did you do the dicom conversion yourself? It does not make sense to have three different bvecs files...

By the way, I think the problem of the rim at the brain border is also related to incorrect preprocessing. Normally tbss_1_preproc will crop the rim away.

Did you run all steps of tbss?

5. That's what I thought as well, but it seems to me from the output that after step 2, the registrations are already done.

I suggest you start form scratch (=dicom conversion) and follow each step carefully!

Best,

Markus

--Francesco

On 23 September 2014 21:21, Chris Watson <[log in to unmask]> wrote:

1. Post screenshots; I don't know what you mean.
2. FA is never negative. I don't think MD can be, either.
3. The bvecs file is the only one containing the directions (I think). bvals contains the b-values, and the nii file is the image itself.
4. Read papers by Behrens et al. and Jbabdi et al. to learn what bedpost does.
5. I could be wrong, but step 2 might just compute the registration, whereas step 3 actually applies them to the data.

On 09/23/2014 01:52 PM, Francesco Puccettone wrote:

Hi everybody,

There are some things which I don't understand from my DTI analyses:

1. Why does my all_FA_skeletonised image have a gray line outlining the brain's outer contour, when clearly the edge of the brain does not contain white matter, let alone skeletonised white matter? Are these values ignored in the analyses, or are they artefacts in how all_FA_skeletonised was created? If so, what causes this artefact, and how to get rid of it?

2. I know that FA values close to zero reflect isotropy of water diffusion, i.e. non-coherent fibre bundles, whereas the opposite (anisotropy) is true for FA values close to 1. However, I know that the range of FA values is from -1 to 1. What do negative values of FA mean, and does it make sense to ignore the sign and just think of the *absolute value* of the FA index as a measure of white matter integrity? Also, how does this observation hold for non-FA measures such as MD?

3. Why aren't all diffusion directions contained in a single BVAL, BVEC and NII file, instead being split into 3 files of each type (with 21+22+25 diffusion directions)?

4. If the diffusion model per se is generated by DTIfit, and the specific bundles are later tracted by ProbTrackX, then what exactly does the BedPostX step of the pipeline do that is not either of these previous operations?

5. What is the difference between step 2 of TBSS, which "runs the nonlinear registration, aligning all FA images to a 1x1x1mm standard space", and step 3, which "applies the nonlinear transforms found in the previous stage to all subjects to bring them into standard space"?

Thank you very much for your kind help!

--Francesco