Like so much else in this world, the choice of smoothing kernel is a choice, with no clear right or wrong answer. By defauly, many people use 8mm. I have also used 6mm and 10mm, as needed.
A larger smoothing kernel will often give you 'better' group statistics. The peak of activity across people is never exactly the same, and higher smoothing can make it easier to capture the spatial variability in the data. It can also reduce noise.
A larger kernel does, however, cost you spatial resolution. So, if you are looking for smaller clusters in smaller structures, you should use a smaller kernel. If you are interested in larger clusters, a 10mm kernel is not such a bad choice.
I generally use 10mm when I am concerned my data my have poorer signal to noise ratio for some reason.
Good luck,
Colin Hawco, PhD
Neuranalysis Consulting
Neuroimaging analysis and consultation
www.neuranalysis.com
[log in to unmask]
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Wei, Shau-Ming (NIH/NIMH) [F]
Sent: October-17-16 10:37 AM
To: [log in to unmask]
Subject: [SPM] Defining smoothing kernel
Dear SPM users,
We used a 10mm3 smoothing kermel in our fMRI analysis, and the reviewer is asking for justifications for using that particular smoothing kernel. Other than invoking the matched filter theorem, are there other ways to justify smoothing kernels?
Thanks,
Shau-Ming Wei
|