Dear Bob,
Yes, what i meant is the contiguous set of 'x' voxels that are above threshold, and have for instance, the largest total activation of such contiguous sets and/or include the most active voxel (and yes, ideally both). In my case, I would like to create seed regions based on data from a localizer task, but I would like these seed regions to be similar in size for each subject. I see how it would be difficult to efficiently compute such a subset of voxels. I could just select the voxel with the largest activation and draw a sphere around it or your new interactive clusterizing feature might do the trick, as some of my lab members have pointed out.
Thanks much for your thoughts,
Heleen