There must be a simple answer to this... unfortunately we can't figure it out and any help anyone can provide would be greatly appreciated.
I computed an F-test comparing 3 groups on a specific contrast. I thresholded the entire brain at p=.05, then ran a small volume correction using an anatomical mask of the left putamen. There is a 288 voxel cluster that is significant after FWE correction. I would like to extract the values from this cluster in order to conduct post-hocs to see which groups differ. Unfortunately I can't seem to do this.
It seems as if the 288 voxel cluster in the L putamen is linked to a 1270 voxel cluster that spans several regions of the brain (b/c there's lots of activity at p=.05). Consequently, whenever I try to extract data from the L putamen cluster (using the MarsBar tool), I end up with data from the entire 1270 voxel cluster and not from the 288 cluster of interest.
I tried using the imcalc tool to create a new image file consisting of only the voxels in the L putamen. Although this limits the cluster size to the putamen and reduces the overall cluster size significantly, it includes all of the voxels of the putamen that fall below p=.05 - not just the ones that are part of the 288 cluster.
I run into similar problems when I do a small volume extraction using a 3mm sphere around the peak activation x y z coordinates identified in the earlier steps above.
Thus far, the only way I can remove extra voxels is by modifying the significance thresholds I set up before the small volume corrections. But there is no way to identify only the 288 voxels here.
This seems like the type of analysis that many people have done before so any help that you can provide would be greatly appreciated!
best wishes
Christen M. Deveney, Ph.D.
Section on Bipolar Spectrum Disorders
Emotional Development and Affective Neuroscience Branch
Mood and Anxiety Disorders Program
National Institute of Mental Health
Bldg 15K; MSC 2670
Bethesda, MD 20892
phone: 301.451.9754
e-mail: [log in to unmask]
|