Skip to content

17. DTI analysis

Jump to bottom
Fa-Hsuan Lin edited this page Mar 13, 2023 · 13 revisions

File arrangement

Place following files in the sub-folders:

  • DTI:
  • dti.nii.gz (DTI in AP phase encoding with diffusion weightings; a larger file)
  • P2A_b0.nii.gz(DTI in PA phase encoding without diffusion weightings; a smaller file)
  • acqp.txt
  • index.txt
  • bvecs
  • bvals

NOTE The above four files are used in our conventional protocol (b-value= 1,000; 60 diffusion directions)

  • T1: (MPRAGE in 1-mm iso. resolution)
  • T1.nii.gz (T1 images)

The following shell script includes all processes above

#!/bin/bash
# fsl runs the dcm2nii for DTI and T1 images
# output DTI.nii.gz file associated with bvecs, bvals to folder DTI
# output T1.nii.gz files to folder T1
# copy acqp,txt and index.txt to folder DTI, (need to specify source path )

dcm2niipath=./mricron
datapath=./baseball
aipath=./

filelist=$(ls ${datapath})
for ff in ${filelist}; do
echo ${datapath}/${ff};
mkdir ${datapath}/${ff}/DTI
mkdir ${datapath}/${ff}/T1

${dcm2niipath}/dcm2nii ${datapath}/${ff}/dcm_DTI
mv ${datapath}/${ff}/dcm_DTI/*.bvec ${datapath}/${ff}/DTI/bvecs
mv ${datapath}/${ff}/dcm_DTI/*.bval ${datapath}/${ff}/DTI/bvals
mv ${datapath}/${ff}/dcm_DTI/*.nii.gz ${datapath}/${ff}/DTI/dti.nii.gz

if [ -d ${datapath}/${ff}/dcm_DTI_P2A ];then
${dcm2niipath}/dcm2nii ${datapath}/${ff}/dcm_DTI_P2A
mv ${datapath}/${ff}/dcm_DTI_P2A/*.nii.gz ${datapath}/${ff}/DTI/P2A_b0.nii.gz
fi

cp ${aipath}/acqp.txt ${datapath}/${ff}/DTI/
cp ${aipath}/index.txt ${datapath}/${ff}/DTI/
${dcm2niipath}/dcm2nii ${datapath}/${ff}/dcm_T1
rm ${datapath}/${ff}/dcm_T1/o*
rm ${datapath}/${ff}/dcm_T1/co*
mv ${datapath}/${ff}/dcm_T1/*.nii.gz ${datapath}/${ff}/T1/T1.nii.gz
rm -r ${datapath}/${ff}/dcm*

done;

Process DTI data by FSL

  1. Skull stripping by bet (default threshold = 0.5, needs tuning).
bet DTI/dti DTI/nodif_brain  -f 0.5 -g 0 -m

bet T1/T1 T1/T1_brain  -f 0.5 -g 0 -m
  1. EPI distortion correction by topup with non-diffusion weighted images with EPI blipped-up and blipped-down along the anterior-posterior direction.
fslroi DTI/dti.nii.gz DTI/A2P_b0 0 2

fslmerge -t DTI/A2P_P2A_b0 DTI/A2P_b0.nii.gz DTI/P2A_b0.nii.gz

topup --imain=DTI/A2P_P2A_b0 --datain=DTI/acqp.txt --subsamp=1 --config=b02b0.cnf --out=DTI/topup_results --fout=DTI/topup_field --iout=DTI/upwarp_b0
  1. Eddy current correction by eddy.
eddy_openmp --imain=DTI/dti.nii.gz --mask=DTI/nodif_brain_mask.nii.gz --bvals=DTI/bvals --bvecs=DTI/bvecs --acqp=DTI/acqp.txt --index=DTI/index.txt --topup=DTI/topup_results --out=DTI/data --ref_scan_no=0 --ol_nstd=4 --verbose
  1. (optional) Convert masks by flirt or flnrt + convert_xfm.
mkdir DTI/xfms
$flirt -in DTI/nodif_brain -ref T1/T1_brain.nii.gz -omat DTI/xfms/diff2str.mat -searchrx -90 90 -searchry -90 90 -searchrz -90 90 -dof 6 -cost corratio

convert_xfm -omat DTI/xfms/str2diff.mat -inverse DTI/xfms/diff2str.mat

flirt -in T1/T1_brain.nii.gz -ref $FSLDIR/data/standard/MNI152_T1_2mm_brain -omat DTI/xfms/str2standard.mat -searchrx -90 90 -searchry -90 90 -searchrz -90 90 -dof 12 -cost corratio

convert_xfm -omat DTI/xfms/standard2str.mat -inverse DTI/xfms/str2standard.mat

convert_xfm -omat DTI/xfms/diff2standard.mat -concat DTI/xfms/str2standard.mat DTI/xfms/diff2str.mat

convert_xfm -omat DTI/xfms/standard2diff.mat -inverse DTI/xfms/diff2standard.mat
  1. DTI metrics, such as FA and MD, by dtifit.
dtifit --data=DTI/data --out=DTI/dti --mask=DTI/nodif_brain_mask --bvecs=DTI/bvecs --bvals=DTI/bvals --sse --save_tensor
  1. Probabilistic tractography by bedpostx.
bedpostx DTI --nf=2 --fudge=1  --bi=1000

NOTE Sometimes the folder DTI.bedpostx needs to be deleted and run the command again, when an error occured.

  1. (optional) Seed ROI for fiber tracking

The following shell script includes all processes above

!/bin/bash
# run the fsl process for DTI data
# {1} : path of case folder
# dti file in ./DTI/dti.nii.gz
#             ./DTI/P2A_b0.nii.gz
# T1 file in ./T1/T1.nii.gz
# bvecs, bvals, acqp.txt, index.txt in ./DTI/
# run the bash with input of subject folder,
# for example : sh r_fsldti_nissen.sh ./sub01

fslpath=/usr/local/fsl
# bet for dti and t1 : dti, t1
${fslpath}/bin/bet ${1}/DTI/dti ${1}/DTI/nodif_brain  -f 0.5 -g 0 -m
${fslpath}/bin/bet ${1}/T1/T1 ${1}/T1/T1_brain  -f 0.5 -g 0 -m

# eddy with topup for dti : bvecs bvals acqp index
${fslpath}/bin/fslroi ${1}/DTI/dti.nii.gz ${1}/DTI/A2P_b0 0 2
${fslpath}/bin/fslmerge -t ${1}/DTI/A2P_P2A_b0 ${1}/DTI/A2P_b0.nii.gz ${1}/DTI/P2A_b0.nii.gz
${fslpath}/bin/topup --imain=${1}/DTI/A2P_P2A_b0 --datain=${1}/DTI/acqp.txt --subsamp=1 --config=b02b0.cnf --out=${1}/DTI/topup_results --fout=${1}/DTI/topup_field --iout=${1}/DTI/upwarp_b0

${fslpath}/bin/eddy_openmp --imain=${1}/DTI/dti.nii.gz --mask=${1}/DTI/nodif_brain_mask.nii.gz --bvals=${1}/DTI/bvals --bvecs=${1}/DTI/bvecs --acqp=${1}/DTI/acqp.txt --index=${1}/DTI/index.txt --topup=${1}/DTI/topup_results --out=${1}/DTI/data --ref_scan_no=0 --ol_nstd=4 --verbose

# registration among diff, structure(str) and standard
mkdir ${1}/DTI/xfms
${fslpath}/bin/flirt -in ${1}/DTI/nodif_brain -ref ${1}/T1/T1_brain.nii.gz -omat ${1}/DTI/xfms/diff2str.mat -searchrx -90 90 -searchry -90 90 -searchrz -90 90 -dof 6 -cost corratio

${fslpath}/bin/convert_xfm -omat ${1}/DTI/xfms/str2diff.mat -inverse ${1}/DTI/xfms/diff2str.mat

${fslpath}/bin/flirt -in ${1}/T1/T1_brain.nii.gz -ref ${fslpath}/data/standard/MNI152_T1_2mm_brain -omat ${1}/DTI/xfms/str2standard.mat -searchrx -90 90 -searchry -90 90 -searchrz -90 90 -dof 12 -cost corratio

${fslpath}/bin/convert_xfm -omat ${1}/DTI/xfms/standard2str.mat -inverse ${1}/DTI/xfms/str2standard.mat

${fslpath}/bin/convert_xfm -omat ${1}/DTI/xfms/diff2standard.mat -concat ${1}/DTI/xfms/str2standard.mat ${1}/DTI/xfms/diff2str.mat

${fslpath}/bin/convert_xfm -omat ${1}/DTI/xfms/standard2diff.mat -inverse ${1}/DTI/xfms/diff2standard.mat

# dtifit for typical diff parameters such as MD, FA: bvecs bvals
${fslpath}/fsl/bin/dtifit --data=${1}/DTI/data --out=${1}/DTI/dti --mask=${1}/DTI/nodif_brain_mask --bvecs=${1}/DTI/bvecs --bvals=${1}/DTI/bvals --sse --save_tensor

# bedpost for probabilistic diffusion model
${fslpath}/bin/bedpostx ${1}/DTI --nf=2 --fudge=1  --bi=1000

Render DTI result (FA map in this case) in Matlab

close all; clear all;

targ=MRIread('/Users/fhlin_admin/workspace/seeg/subjects/s057/mri/orig.mgz');
subject='s057';

targ_reg=eye(4);

dti=MRIread('DTI/dti_FA.nii.gz');


% do registration between pre- and post-OP by the following command:
%
% cd /Users/fhlin_admin/workspace/seeg/subjects/s057/tmp
% bbregister --s s057 --mov ../../../s057/DTI/nodif_brain.nii.gz --init-coreg --reg register.dat --t2
% 
% %check registration
% tkregisterfv --mov ../../../s057/DTI/nodif_brain.nii.gz --reg register.dat --surfs

dti_reg=etc_read_xfm('file_xfm','/Users/fhlin_admin/workspace/seeg/subjects/s057/tmp/register.dat'); %for MAC/Linux

dtim=MRIvol2vol(dti,targ,dti_reg);

etc_render_fsbrain('surf','orig','hemi','rh','subject',subject,'vol',targ,'overlay_vol',dtim,'overlay_threshold',[0.3 0.5]); 

return;

Clone this wiki locally