Identifying location of clusters of activation

Hi all,

I’ve been trying to decide how to identify the location of clusters of activation. For example, I have a contrast which has identified the areas in which a group of hypertensive subjects (HPTN) have shown more activation than the diabetic (DM) subjects.

My initial approach was in FSL. I had multiplied the HPTN > DM contrast with each atlas label for the Harvard-Oxford atlas to create contrasts that would isolate the significant voxels in each region. Then, extract these activation values using Featquery. My rationale for this approach is that it would break apart the contrast in term of the anatomical regions it covers. Then, I could combine these into networks or whatever else I might be interested at looking at.

I’ve been asked instead to use the AFNI viewer to view the contrast. Then, to play with the thresholding and use clusterize to break the contrast apart into regions. I would then use SaveMsk in the Cluster Results window, and extract the stats from each cluster for each subjects using 3dROIstats. But I don’t entirely understand how to use this approach to properly identify the regions. What is my reasoning for using a particular clustering of regions over another? Isn’t the thresholding value rather arbitrary with this approach? How do I accurately identify the cluster regions… by pressing Jump on the Cluster Results window, then right-clicking for whereami? In which case do I use what is listed as the focus point and ignore what is within 2 mm, etc?


In the clusterize report, the clusterize plug calls the “whereami” program with the “WamI” button. That calls the whereami program with the “-omask” option for all clusters and shows the report of overlap with regions in the same space atlas.

Many thanks, Daniel. If I understand correctly, clicking the “WamI” button seems to provide the same information as that contained in the AFNI Cluster Results window. The issue I’m struggling more with is choosing the cluster threshold and correctly identifying the regions. This paper ( seems to suggest using a primary threshold that reduces the extent of clusters so that each cluster is contained within a single anatomical region, and to use the term for the larger encompassing anatomical region in your results if a cluster still spans more than one region.

But it would seem that clusters will often span more than one region. How do I get the name of the larger region that encompasses two smaller regions using whereami? Or, even if it is contained within a single region, do I use what is identified as the “Focus point” by clicking the “Jump” button in the AFNI Cluster Results window and ignore the other labels provided, even if they are within 2 mm? If I understand correctly, the other labels could be within the extent of the cluster, given that “Jump” simply moves to the Peak or CMass.

The omask option for whereami shows the amount of overlap across regions and across atlases. That same omask option is what is used in the WamI from the Clusterize plugin. The overlap then depends on how your dataset got to a standard space and how the atlas was made too. I wouldn’t expect one specific region for any particular atlas usually. Additionally, each atlas defines regions differently, so one atlas will call a particular region one thing, and another atlas will use another name or subdivide the brain differently. The atlases distributed within AFNI do not have hierarchical formats yet, so there is no definition within an atlas for an encompassing region. You will have to find that out on your own for now. That will likely be in a future release though with the development of the HAWGS atlas format and atlases made for that.