RfMRI

antsr reproducibility study of fmri

slides are here

analyze data from:

Single subject fMRI test–retest reliability metrics and confounding factors, Krzysztof J. Gorgolewski, Amos J. Storkey c, Mark E. Bastin b, Ian Whittle d, Cyril Pernet

article

by running an organization script then a processing script

# this script uses wget which you install via homebrew on osx
# on linux wget is native or its alternative, yum 
homedir=${PWD}  # root RfMRI directory 
${homedir}/scripts/download_task_data.sh
# now process each subject to group space 
tem=${homedir}/template/template.nii.gz
for task in task002 task003  ; do 
  for myrun in run001 run002 ; do 
    for x in sub001 sub002 sub003 sub004 sub005 sub006 sub007  sub008 sub009 sub010  ; do 
      cd $homedir 
      mysub=${homedir}/${x}/BOLD/${task}_${myrun}/avg.nii.gz
      mybold=${homedir}/${x}/BOLD/${task}_${myrun}/bold.nii.gz
      if [[ -s $mybold ]] ; then 
        antsMotionCorr -d 3 -a $mybold -o $mysub # average bold 
    gdir=${homedir}/group_analysis/${x}/${task}/${myrun}
        if [[ ! -s ${gdir}/${x}_group.nii.gz ]] ; then 
          mkdir -p $gdir 
          ${homedir}/scripts/ants_2_template.sh $tem $mysub $mybold ${gdir}/${x}_group  
          cd ${gdir}
          ${homedir}/scripts/process_bold.R  -d $task -t 2.5 -n ${myrun} -s $x --bold ${x}_group.nii.gz
        fi
      fi 
    done                     
  done
done

Now we have first-level statistics done and a map to template space.

Next do group level statistics.

# then for each task / run 
./scripts/process_bold_group.R --tr 2.5 --design task002 \
  --run run001 --templatemask ./template/aal.nii.gz \
  --bold group  --output GroupTask002Run001

see the pdf presentation and scripts directory for processing description.

several key-points: groupwise analysis, nuisance variables, look at globalsignal for subset-selection, random effects, slice-timing, hrf-model selection, whitening ...