Clusterwise, TFCE Vs. voxelwise inference

[Guillermina Griffa]

Hi Tom, 

I conducted a longitudinal experiment (n=28) in which we acquired Diffusion Magnetic Resonance Imaging (dMRI) images at 4 time points. After applying the Diffusion Tensor Model (DTI), I obtained a mean diffusivity (MD) image for each subject and each session. 

I'm interested in running an analysis to find differences in MD between sessions using the SwE toolbox. I'm using a bilateral hippocampal mask (binary) to assess differences only within that region. 

My question is, when I use a mask, can I still run an analysis with clusterwise inference or TFCE, or would it be more appropriate to do voxelwise analysis only? 

I tried performing a non-parametric ANOVA contrast among the 4 sessions using voxelwise analysis, and I obtained a cluster with 27 significant voxels (corrected for family-wise error rate). However, using clusterwise analysis, I obtained the same cluster but slightly larger (40 voxels). This raised the question of whether it is appropriate to use clusterwise analysis in this case.

Thanks in advance,

Best,

Guillermina