This all depends on the underlying model that you wish to use for
understanding the data, and whether or not you plan to smooth it afterwards.
The deformations estimated using DARTEL are pretty extreme, with lots of
expansion and contraction. Without some compensation for variability in the
Jacobian determinants, some of the original voxels will contribute more
heavily to the smoothed images than others. For example, a voxel that
doubles in volume during warping will contribute about twice as much signal
as a neighbouring voxel that retains its original volume. This does not make
optimal use of the data.
There is a common view within the neuroimaging field, that more precise
spatial normalisation does not improve sensitivity to differences. I suspect
that empirical studies have not compensated for such expansion/contraction,
and therefore do not make the most of the signal. I have not personally
performed any empirical studies to determine what works best, so I won't say
more.
Best regards,
-John
On Friday 16 January 2009 14:09, Marco Lorenzi wrote:
> Dear SPM users,
> I would like to use Dartel flow field obtained from MR processing in order
> to normalize the pertinent co-registered PET images.
> Can I simply apply these transformation or would be preferred any kind of
> modulation?
> Does anyone have previously tried this procedure?
> Thank you very much!
>
> Marco
|