Please note, Qihong and I have been discussing off-line. We are now talking about denoising 3D data, and I'm recommending that PESTICA is unlikely to work for 3D acquisitions. FSL Melodic may be the only "accepted" method, but please note that spatial whole-brain ICA does NOT remove physiologic noise anywhere near as well as a model-based regression using either pulse/resp or PESTICA (see http://www.ncbi.nlm.nih.gov/pubmed/20600... for a very careful evaluation of Infomax and FastICA in spatial whole-brain ICA at removing physio noise).

Erik