Dear Elena,
Just to add to the comments of Simon and John; In principle,
the template will have no effect on group differences for 'smooth'
spatial normalizations of the sort using in SPM2/99. This is because
the template determines where the data are mapped to, in template-
space but does not affect differences among the data at any voxel.
This is born out empirically (see the Salmond et al analysis cited
by Simon).
John's observation is that differences induced by template
changes have to be caused by a difference in the mapping of
data (from native to template space) between the groups. This
could occur if one group needs substantially more warping than
the other, of a sort that is penalized by basis functions or
priors. Changes in the template could induce changes in the
differential warps and affect the VBM comparison. As John says,
the only way to asses this is empirically, using the template
as a treatment factor in a conventional VBM analysis (looking
for an interaction between template and group).
Although you are unlikely to find this interaction it is
worth remembering that even the interaction has to be caused
by differences in anatomy. Modelling the effect of template
explicitly in the analysis allows you to characterise VBM
results in terms of those differences that depend on gross
anatomical changes (the interaction) and those that do not
(the main effect of group).
I hope this helps - Karl
At 10:29 15/08/2003 +0100, Simon Keller wrote:
>Dear Elena,
>
>Whilst theoretically advantageous, I've found that incorporating a
>study-specific customised template constructed from patients and
>controls doesn't provide more information than using the default SPM
>template (ie controls only) on grey matter differences between
>patients and controls. I obtained the same results, significance
>levels, and spatial extent across SPMs.
>
>If I remember correctly, a couple of papers have shown that template
>type doesn't significantly alter SPMs (Claire Salmond, NeuroImage,
>2002; Tina Good, NeuroImage, 2002).
>
>Best wishes,
>Simon.
>
>__
>Simon S Keller
>The Magnetic Resonance and Image
>Analysis Research Centre (MARIARC)
>University of Liverpool, Pembroke Place
>Liverpool, L69 3BX.
>
>[log in to unmask]
>Tel: 0151 794 5634
>Fax: 0151 794 5635
>
>
|