Dear Karen,
> This topic has been discussed recently, and I was told to look into a
> Friston, 1997 technical note (Human Brain Mapping 5:133-136) for
> determining significance of apriori locations of interest. My question was
> 'can I use uncorrected p-values for voxels in an apiriori region of
> interest?' It appears that this article is geared specifically for
> determining the significance of apriori regions based on the spatial extent
> of significant regions nearby, when that apriori region does not appear on
> the output to be significant. However, my apriori area of interest is broad
> (dorsomedial prefrontal cortex) and indeed, I do find significant voxels
> (uncorrected) in that region that upon correction, do not remain
> significant. Therefore, I don't need to worry about spatial extent of sig.
> voxels nearby since I do have sign. voxels (uncorrected) in my region
> already. I hope that was clear- anyway, the article does not answer my
> question- which remains, can I use uncorrected p-values for voxels in the
> dorsomedial pfc?
Strictly speaking no. You would have to correct for the volume of the
DLPFC that constitutes your anatomical hypothesis. This correction
would use Keiths's small volume correction and will be implemented in
SPM98. At the moment however I would reconsider the simpler approach
based on spatial extent. Take the cluster of activated voxels nearest
the centre of the DLPFC region in question (using a 0 extent
threshold). This cluster will probably be in the DLPFC. If the
uncorrected p value based on spatial extent is less than p=0.05 then
you can consider this cluster to be signficant because (under the null
hypothesis) it was selected at random from all clusters. (the spatial
extent approach does not require the cluster to be outside a VOI it
just requires that the cluster is selected uniquely on anatomical
grounds without biasing its extent.
With best wishes - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|