Dear SPM list,
I want to get a measure of the smoothness of various MRI images.
At first sight, the calculation of Resels carried out by spm_est_smoothness.m
would be a good tool to use for this.
However, from that script's comments and the Worsley tech report
that it is derived from, it appears that resels are intended to be calculated
only from images which are composed of the residuals from GLM analyses.
The images whose smoothness I want to measure are, however,
not residual images. Instead, they are percentage-correct images
which are the outputs of a pattern-based fMRI "searchlight" analysis.
If I make sure that these images have zero mean and unit variance,
is that enough for a meaningful resels-estimate to be produced?
Or is spm_est_smoothness.m simply an inappropriate tool to use in this case?
It would be possible to cook up some alternative way of measuring smoothness,
e.g. using power at different spatial frequencies or local spatial autocorrelation,
but using a resels-measurement would seem to be better grounded.
But only if calculating resels on non-residuals images is a sensible thing to do.
Any help or advice greatly appreciated.
Raj
|