Oh, I didn't know it slipped out by accident. I haven't actually run it yet, but it seems like you could use flameo or perhaps randomise to calculate thresholded group maps. I don't know the output of fsl_sbca or the specifics of how it would be done, but it seems possible. Jeanette On Wed, Mar 21, 2012 at 7:28 PM, bettyann <[log in to unmask]> wrote: > I am under the impression that fslf_sbca has not been officially released > yet: > https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=fsl;391e9b62.1107 > > I was playing around with fsl_sbca but was unsure of how to enter the > results from fsl_sbca into a group analysis. Is there documentation > showing how to use fsl_sbca? > > Thanks, > - BettyAnn > > > > > Hi, > > > > I would try using the fsl_sbca function instead of going through feat. > It > > may be a bit simpler. > > > > As for what you did in feat, you wouldn't want to prewhiten the data, but > > that probably didn't cause your issue. You would also need to normalize > > your seed after extracting from the res4D image *even if you already > > normalized the res4d. > > > > Most likely, if you just getting speckled results, Feat couldn't find the > > brain. Check the mask that it created for your analysis and verify that > it > > actually covers the whole brain. If it didn't, you'll need to double > check > > how you added 100 to the residual image. The purpose of that was to make > > the mean nonzero so feat could find the brain, if you just added 100 to > the > > res4d image, that won't work. You'd need to multiply the mask by 100, > add > > it to the res4d and be sure to use -odt float in your fslmaths command to > > save the output in floating point. > > > > It may be easier to just use fsl_sbca. > > > > Cheers, > > Jeanette > > > > On Wed, Mar 21, 2012 at 7:35 AM, Dr. David Watson < > [log in to unmask]>wrote: > > > > > Dear FSL > > > I am currently engaged in processing rs-fMRI data from a group of > neonates. > > > I am trying to get to grips with the steps involved in removing > nuisance > > > signals and then processing the resulting residual 4D images for seed > based > > > connectivity analysis. > > > I am having some difficulty and wonder if I am making some basic error > > > somewhere. > > > When I do the final FEAT analysis on the Res4D signals I tend to get > > > little statistically significant clusters even in and around the seed > > > regions. So far I have tried Auditory (Left) and Motor (left) seeds > (cubes > > > of 5mm side) located at typical MNI spatial locations employed > elsewhere. > > > The nuisance ts's I am using are: > > > Whole brain - average ts inside brain mask > > > WM - 5mm radius masks based on two (R/L) deep frontal white matter > areas > > > (averaged ts) - identified from WM segmentation > > > CSF - masks based on lateral ventricles (R/L) (average ts) - identified > > > from CSF segmentation > > > These masks are formed for each individual subject > > > > > > After running the initial feat with motion signals and the three > nuisance > > > ts above (intensity normalisation on, FILM prewhitening off) the Res4D > is > > > normalised (subtract mean and divide by std and add 100) and submitted > to > > > another feat analysis using one EV based on a chosen seed region mask > ts, > > > extracted from the Res4D image. > > > The settings I am using for this final stage are: > > > Motion Correction - off > > > Slice Timing - off > > > BET extraction - off > > > Intensity Normalisation - off > > > > > > FILM prewhitening - on > > > Model: Motion correction off; EV based on seed region added > > > > > > Does this look appropriate. Glad for any suggestions. > > > > > > David >