Hi there,
we would like to do % signal change and correlation analyses on our activation fMRI study using ROIs that have been derived from the parcellation of resting state data and made available to the public (http://www.nitrc.org/projects/cluster_roi/). The data come in different number of parcellations, for the sake of this discussion say we used the parcellation of the brain in 400 ROIs.
Some words of the procedure:
1. We are interested in brain regions interested in working memory, so we re-created an ALE cluster based on a meta-analysis (see GingerALE website; actually we use the FDR-corrected p value map). The result is a mask covering large brain areas - it's too big.
2. We mask the independent functional ROIs with the ALE cluster to restrain the number of ROIs to regions only involved in working memory.
Our concerns are:
1. Because the parcellations are functionally derived, they do not really correspond to anatomy.
2. The size of the ROIs vary, not only because of the masking process (with the ALE cluster)
3. it is not clear to what resting state networks these ROIs belong to - these can be checked but we are not sure whether this makes sense as usually the resting state networks described are big compared to the ROIs, thus there may be several ROIs withing one network.
While we think the ROI selection is as good as any other, there are some concerns as outlined that may cause problems when submitting the paper.
What do you guys think, any suggestions?
Best wishes,
Torsten
|