Thanks Brian, I tried bias field correction prior to FIRST as you described. There is some improvement but there is still a significant amount of tissue missing. Best Deniz On Wed, 26 Aug 2009 23:47:54 +0100, Brian Patenaude <[log in to unmask]> wrote: >Hi, > >There is actually many children in the training set used to create the >model. Is there any difference in presence of bias field effects between >the adults and children? You could try running > >brain extracting -> bias field estimation with FAST -> apply correction to >whole head image -> the run FIRST > >Cheers, > >Brian > >> Hi, >> >> I am trying to segment cerebellum using FIRST on small group of subjects >> that >> consists of adults and children. I am really impressed by the segmentation >> that >> I get on adults, but the segmentation on children images are >> underestimating >> the cerebellum by about 5 mm towards the bottom and sides. >> >> I was wondering if there are any parameters I can adjust, or different >> models >> that are available for children. >> >> I am using the following commands to get the segmentation. (I believe >> somebody has posted these here a few months ago, thanks :) ). "T1" refers >> to >> an MPRAGE sequence, it is *not* brain stripped. >> >> Thanks in advance for any input you can provide. >> >> Deniz. >> >> run_first_all -i T1.nii.gz -o T1_out >> >> first_flirt T1 T1_to_std_sub -cort >> run_first -i T1 -t T1_to_std_sub_cort.mat -n 40 -o T1_R_Cereb -m >> ${FSLDIR}/data/first/models_336_bin/intref_puta/R_Cereb.bmv -intref >> ${FSLDIR}/data/first/models_336_bin/05mm/R_Puta_05mm.bmv >> first_boundary_corr -s T1_R_Cereb -i T1 -b fast -o T1_R_Cereb_corrected >> >>