Once a height threshold is chosen, the intensity information is
discarded in a cluster thresholding -- all that matters is the number
of contiguous voxels and whether that extent is significantly greater
than you would expect by chance given:
1. the height threshold chosen
2. the search volume (and surface area, but I think FSL ignores that
one)
3. the smoothness of your data
So there is a minimum extent value and FSL calculates it but doesn't
explicitly provide that figure. SPM, in contrast, puts the minimum
extent in the report, which is probably why the reviewer has asked for
it. Sometimes you can figure it out, though, by looking at your
cluster chart if you happen to have clusters with p-values around
0.05. But as long as you report the extent of your activations, you
shouldn't need to report the minimum extent.
When a reviewers ask this it is typically because cluster stats are
routinely misused to hide uncorrected inference (things like a voxel-
wise p<0.001 and a cluster extent of p<0.05 was used and then
reporting activations with Z=2.3 and 10 voxels). If you report your
extents either in voxels (if so, also report voxel size) or mm3, then
one can normally see whether the activations are approximately the
right order of magnitude. So even if you can't figure it out, you can
probably convince your reviewer you're doing valid stats by providing
extent info.
Joe
Joseph T. Devlin
Cognitive, Perceptual and Brain Sciences
[log in to unmask]
|