Dear AFNI experts,
I am trying to implement @SSwarper into my pre-processing and whilst the defaults (@SSwarper -input {T1 file} -subid {subject} -odir {out_dir} -base {template}) worked well for most subjects, some subjects had clippings in axial inferior slices. To combat this, I have updated AFNI (now running Precompiled binary linux_ubuntu_16_64: Nov 12 2020 (Version AFNI_20.3.02 ‘Vespasian’)) and consulted the message board where I found suggestions that the lack of alignment between the anatomical scan and the template image could be the reason for clippings (see this post). Hence, I have changed my pipeline to (a) run @Align_centers to align the template and scan before running @SSwarper and (b) add settings (-giant_move -SSopt ‘-push_to_edge’) for the skull stripping (see code below).
#!/bin/bash
# define path
DIR="/storage/shared/research/cinn/2018/MAGMOT"
# change directory to BIDS folder
BIDS_dir="$DIR"/MAGMOT_BIDS
# change directory to the raw NIFI files
cd $BIDS_dir
subjects=($(ls -d sub*))
subjects=(sub-control003)
# define derivatives directory
deriv_dir="$DIR"/derivatives
# define which template to use and where to find them
template=MNI152_2009_template_SSW.nii.gz
template_path=`@FindAfniDsetPath $template`
# for each subject in the subjects array
for subject in "${subjects[@]}"; do
echo "###################################################################################"
echo "@SSwarper for subject $subject"
echo "###################################################################################"
# create output folder
out_root=$deriv_dir/afniproc
out_dir=$out_root/$subject/SSwarper
mkdir $out_dir
# define BIDS anat folder
anat_dir=$DIR/MAGMOT_BIDS/"${subject}"/anat
anat="$subject"_rec-NORM_T1w.nii.gz
prefix="$subject"_shft.nii.gz
cd $anat_dir
# copy original anatomical scan
3dcopy $anat $out_dir/$anat
cd $out_dir
# align centers
@Align_Centers \
-cm \
-prefix $prefix \
-dset $anat \
-base $template_path/$template
# specify SSwarper
@SSwarper -input $prefix \
-subid $subject \
-odir $out_dir \
-base $template_path/$template \
-giant_move -SSopt '-push_to_edge'
done
Although the adaptations have improved the output of @SSwarper significantly (compared to default ran on an older version), it is still not as good as the output of most default subjects and some cortex is still clipped (see attachment). When checking the file init_qc_00_overlap_usrc_obase.jpg I noticed that even though the output of @Align_centers is used as input for @SSwarper, the alignment is still poor. I have tried to play with the alignments a bit to improve the results (e.g. by running 3dMC before running @Align_centers; see code below), but the result still does not seem satisfying (see attachment).
#!/bin/bash
# define path
DIR="/storage/shared/research/cinn/2018/MAGMOT"
# change directory to BIDS folder
BIDS_dir="$DIR"/MAGMOT_BIDS
# change directory to the raw NIFI files
cd $BIDS_dir
# define subjects based on folder names in the BIDS directory
subjects=($(ls -d sub*))
subjects=(sub-control003)
# define derivatives directory
deriv_dir="$DIR"/derivatives
# define which template to use and where to find them
template=MNI152_2009_template_SSW.nii.gz
template_path=`@FindAfniDsetPath $template`
# for each subject in the subjects array
for subject in "${subjects[@]}"; do
# create output folder
out_root=$deriv_dir/afniproc
out_dir=$out_root/$subject/align_centers
mkdir $out_dir
# define BIDS anat folder
anat_dir=$DIR/MAGMOT_BIDS/"${subject}"/anat
anat="$subject"_rec-NORM_T1w.nii.gz
anat_rf="$subject"_refit.nii.gz
cd $anat_dir
# define prefix
prefix_shft="$subject"_shft.nii.gz
prefix_shft_rf="$subject"_shft_refit.nii.gz
prefix_3dCM="$subject"_3dCM.nii.gz
prefix_3dCM_rf="$subject"_3dCM_refit.nii.gz
prefix_3dCM_shft="$subject"_3dCM_shft.nii.gz
prefix_3dCM_shft_rf="$subject"_3dCM_shft_refit.nii.gz
# copy original anatomical scan
3dcopy $anat $out_dir/$anat
# go to output directory
cd $out_dir
# change center of mass
3dcopy $anat $prefix_3dCM
3dCM -automask -set 0 0 0 $prefix_3dCM
# align centers between anatomical and template
@Align_Centers \
-cm \
-prefix $prefix_shft \
-dset $anat \
-base $template_path/$template
# align centers between anatomical and template after changing center of mask of anatomical
@Align_Centers \
-cm \
-prefix $prefix_3dCM_shft \
-dset $prefix_3dCM \
-base $template_path/$template
# refit to MNI for visualisation purposes: first copy files, then refit
3dcopy $anat $anat_rf
3dcopy $prefix_shft $prefix_shft_rf
3dcopy $prefix_3dCM $prefix_3dCM_rf
3dcopy $prefix_3dCM_shft $prefix_3dCM_shft_rf
3drefit -space MNI $anat_rf
3drefit -space MNI $prefix_shft_rf
3drefit -space MNI $prefix_3dCM_rf
3drefit -space MNI $prefix_3dCM_shft_rf
# copy template into directory for visualisation purposes
3dcopy $template_path/$template $template
done
I have noticed that even after running 3dCM the center of mass is still not close to the ACPC line (though closer than before) and also that it seems as if the center of mass moves back to its original position if I run 3dCM + @Align_centers. Is there any way how I can improve the alignment between my anatomical scan and the MNI template further to improve the skull stripping performed by @SSwarper?
Also, because this has been discussed previously, I have added the obliquity information of the original anatomical image and the output of @Align_centers.
3dWarp -disp_obl_xform_only -deoblique $anat
++ 3dWarp: AFNI version=AFNI_20.3.02 (Nov 12 2020) [64-bit]
++ Authored by: RW Cox
mat44 Obliquity Transformation ::
0.998696 0.015065 -0.048780 -5.739861
-0.026233 0.971115 -0.237165 -13.290672
0.043798 0.238135 0.970244 -36.265778
3dWarp -disp_obl_xform_only -deoblique $prefix_shft
++ 3dWarp: AFNI version=AFNI_20.3.02 (Nov 12 2020) [64-bit]
++ Authored by: RW Cox
mat44 Obliquity Transformation ::
1.000000 0.000000 0.000000 0.000000
0.000000 1.000000 0.000000 0.000000
0.000000 0.000000 1.000000 0.000023
Any pointers on how to improve the alignment between anatomical and MNI would be really helpful!
Many thanks in advance,
Stef
