On Friday, October 14, 2011 15:01:37 you wrote: > Hi Benjamin > Thanks a lot for all your suggestions! > 1. Conveniently I ignored this step because I thought mellodic does > registration, But like you said I will try to register them to standard > brain before I run melodic. Indeed, the GUI version does if you ask it to. > 3. That's exactly what I am trying to > do...Thank you I didn't know about this. But if I'm using fsl_glm what > will be my design? Johnson You might try just running melodic on your data as-is first and see if you get anything that looks good. Otherwise, if you search the mailing list archives you will find threads about using fsl_glm to remove nuisance regressors (e.g. motion parameters) prior to ICA. You will also find a discussion of how to incorporate nuiscance regressors into dual regression. As Stephen Smith recently pointed out, if you still really want to use fsl_regfilt, you could consider doing ICA in two stages. First do ICA on each subject individually, identify individual noise components, and remove them with fsl_regfilt. Then take the filtered data and do your tcICA on that. This is a lot of work, though!