help > Prob with images corrected for multiple comp
Showing 1-2 of 2 posts
Display:
Results per page:
Nov 9, 2017  07:11 AM | Jorge Morales
Prob with images corrected for multiple comp
Hi Chris,

Is there a way of overlaying an image created in SPM with only the clusters that survive multiple comparisons? The problem with using unthresholded images is that clusters that do not survive correction for multiple comparisons may still show up if the unthresholded .nii file is used--even after setting a threshold on Surfice. Naturally, overlaying the thresholded imaged with just significant clusters saved in SPM gives the 'wrong' impression of much larger clusters, unless a much higher threshold is applied (but then some significant clusters may disappear). I understand why thresholded images might not be rendered correctly in a 2D mesh like the one that Surfice uses. But, is there any work around this?

Thanks!
Dec 5, 2017  05:12 AM | Chris Rorden
RE: Prob with images corrected for multiple comp
The Preferences choice "Smooth voxel-based images" has an impact on these images. The issue is that the voxels in a volume do not have a one-to-one correspondence with the vertices in a mesh. In general, if your data is pre-thresholded (artificially surrounded by zeros) you must either use nearest-neighbor interpolation (which makes surviving regions look smaller and with jagged edges) or trilinear smoothing that can blur the edges. Personally, I prefer using unthresholded images so the non-significant voxels are able to accurately influence their surviving neighbors allowing a better interpolation. The MediaWiki describes this in more detail.