Helena,
One of of the subtleties of cluster inference is that you can't
actually say which voxels in the cluster are active. You can only say
that one or more voxels within the cluster are active, but you are not
sure exactly where.
A printed reference for this is the great book chapter in "fmri
techniques and protocols" (ed. by Filippi) written by Woolrich,
Beckmann, Nicols and Smith, p. 221.
So, having a large cluster that extends across multiple regions is not
very useful, because you can't actually say where the activation
occurs.
If you raise the threshold so your cluster only extends across one
region, then you can say that the activation occurs in that region.
Hope this helps!
Jim Lee
On 2/11/18, Helena Lee <[log in to unmask]> wrote:
> Hello,
>
> I am currently using spm8-fmri to perform 2nd level analysis and about to
> report the mni coordinate of the peaks in each activation cluster .
>
> However, some clusters size I found in the result table are quite big.
>
> For example, one of the clusters has 43892 voxels and contains both left and
> right fusiform and parts of parahippocampal region.
>
> Therefore, I am not sure about how to report this result.
>
> I tired to change the significance threshold from uncorrected p<0.001 to FWE
> p<0.05, but sometimes it turned out to leave almost no activation cluster.
>
> Any suggestion for reporting SPM results?
> Could I manually select the peaks I am interested in within the cluster to
> report? Will this manual selection procedure seem like a forgery?
>
> Thanks you so much.
>
> Best,
> Helena
>
|