Hi Eddy
See Jesper's recent comment in the list on FA>1 (or negative
eigenvalues).
If you want to build an average of many subjects, you can e.g. ignore
those voxels with negative eigenvalues (mask them out).
[although if you want to use your atlas for quantitative stuff, such
as statistical testing, you should probably keep the info on how many
subjects you've used to calculate the mean in each voxel].
Saad.
On 20 Mar 2009, at 18:34, Eddy D.L. wrote:
> Hi Saad,
> I have a question to negative eigenvalues L1, L2, L3 (build with
> tbss_non_FA, for
> reconstruction of DTI tensor image). This value are in the order of
> 10^-5,
> can it due software or hardware approximation?
>
> cheers,
> Eddy
>
Saad Jbabdi
Oxford University FMRIB Centre
JR Hospital, Headington, OX3 9DU, UK
+44 (0) 1865 222545 (fax 717)
www.fmrib.ox.ac.uk/~saad
|