Hi Miguel,
you should use for each subject the corresponding gradient table, rather than a common one across all subjects. The different gradient tables across sessions reflect differences in slice angulation.
Cheers,
Stam
On 4 Aug 2010, at 18:15, Miguel Burgaleta wrote:
>
> Hello FSLers,
>
> This question is not strictly related to the FSL usage so extra thanks to responders in advance.
>
> I have DTI data acquired twice for each subject. I extracted the gradient table with dcm2nii and compared a few outputs to make sure that everything is OK. What I found is that, within subjects, the two gradient tables from the two acquisitions are identical, but there are differences between subjects both in sign and in "magnitude". My understanding is that differences in sign shouldn't affect FA, MD and L1,2,3 measures -is that correct? Regarding differences in "magnitude", see an example below:
>
> Subject1:
>
> 0 -0.999035299 -0.489039928 0.215717688 0.527241349 -0.647644281 -0.373112321 ...
> 0 0.006396742 0.757913172 0.209543079 0.84421742 -0.139462292 0.862912297 ...
> 0 0.043448307 0.431749523 -0.953707278 0.096507609 -0.749071002 -0.340836525 ...
>
> Subject2:
>
> 0 0.999370098 0.50899297 0.264023662 0.521039307 0.609656811 0.355761737 ...
> 0 -0.006128726 -0.792563558 0.117028512 0.850839913 0.215570584 -0.823399842 ...
> 0 -0.005823052 -0.332863569 -0.945997536 0.04023619 0.759996772 0.436671495 ...
>
>
> Sometimes the difference is very small, but sometimes it is around 0.1 (which sounds a lot to me). The question is: assuming that this is the general pattern across the sample, is it safe to use a single gradient table for the tensor estimation or should I do it subject-by-subject?
>
> Thanks again,
> Miguel
>
>
|