 |
 |
Hello experts! I am analyzing a case study. This patient has very large ventricles so we planned to mask them out. We thought this would improve our sensitivity by decreasing the number of multiple comparisons. We are using a cluster-based threshold. At the threshold we've chosen (and several others that I've played with to try to understand the problem better), it seems that all of the activated voxels belong to one large cluster, despite the presence of reasonable-looking local maxima in brain tissue. These local maxima are lost when we include the pre-threshold mask, presumably because they are attached to a larger cluster that is being partially masked out (but they are only attached by a few voxels). It seems like this approach is given more credence to the narrow path of activated voxels that are connecting my local maxima than to the local maxima themselves. I am wondering if I am interpreting these results correctly, and if it would be reasonable to apply the mask before the analysis to avoid this problem, or if we should move away from cluster-based thresholding in this case? Thanks in advance for your insight, Erin