Dear John
I have a question concerning the Jacobian calculation in SPM8.
I would like to compare a local VBM analysis (using MedInria and ITK)
and a global analysis using SPM8 (DARTEL and New Segmentation).
For doing this, I need to compare the jacobian determinants obtained by
the two approaches.
I computed deformation fields using MedInria and implemented a jacobian
determinant calculation using ITK.
I got SPM jacobian determinants using Tools->DARTEL Tools->Jacobian
Determinant.
Results obtained are really different.
I suspected an error in my determinant calculation but it sounds good.
Then I checked SPM code and I didnt really understand the method.
It appears that, for the jacobian determinant calculation, weights are
introduced that correspond to the time points chosen.
In function smalldef_jacdet (in diffeo3d.c)
J0[o] = j00*(j22*j11-j21*j12)
+ j01*(j12*j20-j10*j22)
+ j02*(j21*j10-j20*j11);
and each jii is something like j10 = (v1[op1]-v1[om1])*sc2; (sc2 being
what I called a weight...).
This calculation is followed by a composition of this jacobian
determinant in function composition_jacdet.
Why is it not a direct jacobian calculation ? Did I missunderstood
something ?
Thanks in advance for your precious help.
Sebastien CELLE
|