On 31 Aug 2012, at 23:35, SUBSCRIBE FSL Nucleos wrote:
We are using TFCE to find differences in FA between two study groups using FSL. Similar to another question where they found more than 70000 voxels in a cluster, we are getting a cluster of 23000 voxels with good corrected p. However, when we visualize this cluster within FSLview, they are far apart. One significant cluster is in the forceps minor, another in the temporal part of the superior longitudinal fasciculus and another one in the anterior thalamic radiation. However, FSL assigns a single index number to a cluster of >23000 voxels that encompasses all these regions making it hard to extract FA values for further analysis outside of FSL. The documentation states that FSL accounts for global FA
Where? I didn't think it does do this, or say that it does it in the doc!
But yes if you want to put in global mean FA as a covariate into randomise then that's fine.
while estimating regional differences in FA.
How can we create masks to extract the FA values from three regions of interest?
Or - in general if your clusters are too large wrt how you want to interpret them, you can just raise the corrp threshold so that they start to split up….?
Cheers.