 |
 |
> thanks for your suggestions. i've done what you suggested about the gray > matter segmentation mask for functionals. > now i'd like to put together results from more than 1 session and > compare the analysis from other subjects. > so i need all subjects to be in the same space and segmentation mask to > be the same for all subjects I'm afraid that spatial normalisation using only about 1000 parameters would not be able to spatially transform all the GM masks to be completely identical. They should be fairly close though. > > either > 1- calculate normalization parameters for my T1 structural to T1 template > 2- segment Gray matter in my T1 structural > 3- normalize gray matter segmentation mask to T1 template using > parameters estimated at step 1 above If you plan to segment in native space, then you could use the grey matter to estimate the spatial transformation that would best match it to the GM image in the apriori directory. i.e. use the "optimised VBM" strategy. > 4- normalize all (preprocessed=slice tim. + realign to 1 image) sessions > from each subject to T1 template > 5- apply normalized segmentation mask from step 3 to normalized > functionals from step 4 It should work. > > otherwise I could do the following: > 1- conduct analysis from not segmented functional images from all subjects > 2- write my results about activations in results.img analyze image for > each session from each subject > 3- normalize results.img from step 3 using parameters calculated on T1 > structural > 4- segment result.img using segmentation mask calculated on normalized > T1 structural > > I was thinking the second way could avoid some error propagation for > voxels signal, using normalization just for representation in the same > space do you think that makes sense? Do you plan to smooth your fMRI before doing the stats? If not, then the scheme should (in theory) work, but you are likely to lack a lot of the tools for interpreting the results with GRF. Also, one of the really annoying things about the stats part of SPM is the masking of the results. This is likely to have negative consequences for what you plan to do, as well as meaning that for random effects studies, it is not possible to process by: realign the fMRI images, statistical analysis to get a contrast image, spatially normalise the contrast images, smooth it, and finally enter it into the second level analysis. Instead, the processing has to be: realign the hundreds of fMRI images, spatially normalise hundreds of fMRI images, smooth hundreds of fMRI images, statistical analysis to get a contrast image and finally do the second level analysis. Best regards, -John > John Ashburner ha scritto: > >> i'd like to obtain a gray matter segmentation binary mask for my > >> functional volumes. > >> > >> 1- I was thinking of coregistering T1 sagittal to functional mean image > >> and segment the T1. My original structural T1 images are sagittal, > >> while functionals are axial: could that be a problem? > > > >I would suggest using the Display button to reorient all your images to be > >axial. See the mailing list for clues about this (keywords Display and > >Reorient). > > > >> 2- Is it possible to obtain a binary mask out of Gray Matter segmented > >> T1 analyze file at the functionals resolution? > > > >1) Coregister the structional and functional data (reslicing not > > necessary). 2) Segment the structural image. > >3) Use the Coregister button to reslice the seg1 image to match one of > > your functional images (Space defining image: functional, Images to > > reslice: seg1). > > > >Best regards, > >-John