Hi there,
I am comparing regions of activation from two different sets of functional data, and would like to make a statement about the likelihood that the area of overlap (i.e., the common area of significant activation for both contrasts) is due to chance (or, more precisely, my confidence that it is not due to chance).
It seems to me that an extension of the Monte Carlo approach used with mri_glmfit to correct for multiple comparisons might work well for this, so I'm wondering whether this might be possible with the current set of freesurfer/fs-fast tools.
Specifically, I'd like to:
- use mri_glmfit to twice run the Monte Carlo simulator (mc-full-- input noise, smooth, and analyze) for each of n-thousand iterations.
- submit the two resulting activation maps to something like mergecontrasts-sess to obtain a map of the clusters of overlapping activation (-conjunction and)
- use these clusters-of-overlap to provide cluster simulation data (CSD) to mri_surfcluster along with my original map of clusters of overlapping activation in order to calculate the corresponding cluster-wise p-values (CWP).
If my reasoning is correct, the CWP values following this procedure would be indicative of the probability associated with areas of overlapping activation, rather than the likelihood of obtaining a cluster of certain size/significance, per se.
The main outstanding questions before doing this seem to concern:
- getting the output of each of the two Monte Carlo simulations so that extent of overlap in the different sets of clusters can be computed -- Is there a way to pipe the simulation output to a glmdir?
- computing the areas of overlap for clusters from each pair of simulations. It's not clear to me how to feed these two maps as input to mergecontrasts-sess. -- Is there a way to provide direct paths for the inputs to mergecontrasts, rather than rely on the -analysis, -contrast etc. flags?
- assessing the list of clusters resulting from mergecontrasts-sess in order to get a CSD file that can be fed into mri_surfcluster
Any comments on this approach, and especially suggestions to help in its implementation would be greatly appreciated.
Thanks, Mark
----- Mark J. Fenske, Ph.D. MGH Martinos Center for Biomedical Imaging - Harvard Medical School Tel: (617) 726-9034 | Fax: 726-7422 | http:// barlab.mgh.harvard.edu/fenske.htm