 |
 |
That's a lot of questions! To address a few points: 1. When doing multiple-subject ICA, you should probably get your data in standard space *before* doing ICA with melodic. 2. fsl_regfilt is typically run on each subject individually so as to best model each individual's noise components. 3. If you are doing resting state analysis in the sense of, "I wonder what brain regions are active at rest in this group of subjects?", then you do not need to "denoise" your components after tcICA. (You might consider removing nuisance regressors with fsl_glm before you do ICA, though.) Rather, identify which of the 83 components are noise and which are interesting, and then do what you want to do with the interesting ones. Hope this helps you out! Benjamin On Friday, October 14, 2011 14:00:58 you wrote: > Hello everyone > I am quite new to FSL and trying to run my first resting state melodic > analysis. I was able to successfully run Multisession temporal > concatenation on 10 patients and got 83 identified components. I know my > next step is to denoise them but this were I am stuck. I tried to run > fsl_regfilt command but kept getting 'NEWMAT: > IncompatibleDImensionsException' error. So I was trying to run flirt to > match melodic_IC.nii.gz to standard brain, but couldn't figure out how to > apply tranform (applyxfm) because FSL created different > .../reg/example_func3standard.mat file under each subject. If I just pic > one subjects example_func3standard.mat as transform for flirt the brains > dont register properly. Should I be running melodic individualy for each > subject instead of group? If so is there a way to put them together in the > end? And also how to set the design matrix for resting state analysis, > when running feat after denoising in a single subject level? > > Thanks a lot for your time. > Johnson