Hi - for sanity, I would start by applying the below sets of commands
to a single subject's L1,2,3 in native space and compare the result
with FA - if you've got the equations right then they should match
very well.
Assuming the above is ok, then your differences are probably explained
by the fact that when you apply the FNIRT warps to the different FA/
L1,2,3 images the images get interpolated, and because FA is not a
linear function of the eigenvalues then the equation relating all
these functions will no longer exactly hold any more.
Cheers, Steve.
On 8 Apr 2009, at 19:52, David Harper wrote:
> Greetings,
>
> I am running TBSS on a data set of 16 controls and 16 patients. We
> have significant findings using TBSS and randomise on our FA data.
> So, in order to understand these findings better, we explored the
> eigenvalues (l1 and the RMS of l2, l3). We got some results we
> didn't expect, so we decided to see if we could reconstruct
> (voxelwise) the FA skeletonized data using the three sleletonized
> eigenvalue images (l1, l2, and l3). Our FA reconstruction was
> according to this voxelwise formula (lambda without subscript is
> mean):
>
> <FA_eq.gif>
>
>
> In case the formula doesn't go into the forum, I'm adding it here in
> text:
>
> FA=sqrt(3/2) * sqrt( (l1-l)^2+(l2-l)^2+(l3-l)^2 ) /
> sqrt(l1^2+l2^2+l3^3)
>
> When we compared skeletonized points using fslview our
> reconstruction appeared to systematically vary from what fsl
> produced. We appeared to have a positive error that increased in
> magnitude as FA increased above 0.5 and a negative error that
> increased in magnitude as FA decreased below 0.5. We were wondering
> if you folks had any idea why our reconstruction would be off in
> this way. I'm attaching a log of our fslmaths calculations.
>
> fslmaths all_L1_skeletonised.nii.gz -add all_L2_skeletonised.nii.gz
> -add all_L3_skeletonised.nii.gz all_trace_skel
> fslmaths all_trace_skel.nii.gz -div 3 all_meanL_skel
> fslmaths all_L1_skeletonised.nii.gz -sqr all_L1sqr_skel
> fslmaths all_L2_skeletonised.nii.gz -sqr all_L2sqr_skel
> fslmaths all_L3_skeletonised.nii.gz -sqr all_L3sqr_skel
> fslmaths all_L1sqr_skel.nii.gz -add all_L2sqr_skel.nii.gz -add
> all_L3sqr_skel.nii.gz all_Lsumsqr_skel
> fslmaths all_Lsumsqr_skel.nii.gz -sqrt all_Lrms_skel
> fslmaths all_L1_skeletonised.nii.gz -sub all_meanL_skel.nii.gz
> all_L1dm_skel
> fslmaths all_L2_skeletonised.nii.gz -sub all_meanL_skel.nii.gz
> all_L2dm_skel
> fslmaths all_L3_skeletonised.nii.gz -sub all_meanL_skel.nii.gz
> all_L3dm_skel
> fslmaths all_L1dm_skel.nii.gz -sqr all_L1dmsqr_skel
> fslmaths all_L2dm_skel.nii.gz -sqr all_L2dmsqr_skel
> fslmaths all_L3dm_skel.nii.gz -sqr all_L3dmsqr_skel
> fslmaths all_L1dmsqr_skel.nii.gz -add all_L2dmsqr_skel.nii.gz -add
> all_L3dmsqr_skel.nii.gz all_Ldmsqr_skel
> fslmaths all_Ldmsqr_skel.nii.gz -sqrt all_Ldmrss_skel
> mv all_Lrms_skel.nii.gz all_Lrss_skel.nii.gz
> mv all_Ldmsqr_skel.nii.gz all_Ldmsumsqr.nii.gz
> fslmaths all_Ldmrss_skel.nii.gz -div all_Lrss_skel.nii.gz
> all_Lratiorss_skel
> fslmaths all_Lratiorss_skel.nii.gz -mul 1.22474487 all_FArecon_skel
>
>
> Any help you can provide would be appreciated.
>
> Best,
>
> David
>
>
>
> David G. Harper, Ph.D.
>
> Assistant Professor of Psychology
> Department of Psychiatry
> Harvard Medical School
>
> Associate Psychologist
> McLean Hospital
> 115 Mill Street
> Belmont, MA 02478
>
> ph. 617-855-3160
> fax 617-855-3246
> [log in to unmask]
>
>
>
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|