Hello Jenny,
as you are using a VBM like approach to analyse your DTI data and there
seems to be an anatomical pattern in the fwhm image,
I cannot see any reason why not to report cluster p-values that are adapted
for local smoothness.
Indeed, as you say, the local smoothness influences the sensitivity getting
a significant cluster p-value, so if you in addition have seen
some pattern of your clusters fitting onto the fwhm pattern, this even more
points to the fact that in your data set local smoothness should be
considered.
The darker signal of the WM tracts (indirectly) probably indicates that the
normalisation/registration procedure between your cases works best (or:
better)
in these areas when compared to more peripheral areas (the fwhm image is
quasi "inverse" to the RPV roughness image).
So, as for DTI when analyses in VBM style local smoothness differences are
a major problem as mostly coregistration is not fine coarse enough to also
match the more distal tracts onto eachother, the use of permutation methods
in general and the consideration of non-stat. smoothness is a good idea.
Aside from that, methods as TBSS seem to be more reliable with regard to
the coreg problem for anisotropy maps - so may be you should also consider
switching to that.
best regards,
Philipp
Max-Planck-Institute of Psychiatry
NMR Research Group
Kraepelinsr. 2-10
80804 Munich
Mail: [log in to unmask]
Phone: 0049-89-30622-413
|