Protocol for applying TBSS skeletonizations from FA analysis to diffusivity and obtaining ROI measures using the ENIGMA-DTI template

May 27, 2015

support.enigmaDTI@ini.usc.edu

The following steps will allow you to skeletonize diffusivity measures including mean, axial and radial diffusivity (denoted by MD, L1, and RD respectively) and extract relevant ROI information from them according to the ENIGMA-DTI template, and keep track of them in a spreadsheet.

***Before you get started, you must perform all the ENIGMA- FA analyses!! This protocol will follow the same naming conventions ***

Make sure you have already performed the FA analyses here:
http://enigma.ini.usc.edu/wp-content/uploads/DTI_Protocols/ENIGMA_TBSS_protocol_USC.pdf
and
http://enigma.ini.usc.edu/wp-content/uploads/DTI_Protocols/ENIGMA_ROI_protocol_USC.pdf

INSTRUCTIONS

1. Setup

a. From the previous TBSS protocol (linked above), we will assume the parent directory is: /enigmaDTI/TBSS/run_tbss/,
which we will define this through the variable parentDirectory but you should modify this according to where your images are stored.
b. Also as before, we will assume your ENIGMA template files are
/enigmaDTI/TBSS/ENIGMA_targets/ENIGMA_DTI_FA.nii.gz
/enigmaDTI/TBSS/ENIGMA_targets/ENIGMA_DTI_FA_mask.nii.gz
/enigmaDTI/TBSS/ENIGMA_targets/ENIGMA_DTI_FA_skeleton.nii.gz
/enigmaDTI/TBSS/ENIGMA_targets/ENIGMA_DTI_FA_skeleton_mask.nii.gz
/enigmaDTI/TBSS/ENIGMA_targets/ENIGMA_DTI_FA_skeleton_mask_dst.nii.gz

**Note: if you had to re-mask the template your paths will be to the edited versions, so remember to use these instead!**
/enigmaDTI/TBSS/ENIGMA_targets_edited/mean_FA.nii.gz
/enigmaDTI/TBSS/ENIGMA_targets_edited/mean_FA_mask.nii.gz
/enigmaDTI/TBSS/ENIGMA_targets_edited/mean_FA_skeleton.nii.gz
/enigmaDTI/TBSS/ENIGMA_targets_edited/mean_FA_skeleton_mask.nii.gz
/enigmaDTI/TBSS/ENIGMA_targets_edited/mean_FA_skeleton_mask_dst.nii.gz

2. Copy all necessary diffusivity images (from TBSS’s DTIFIT, for example) into designated directories in your run_tbss/ folder.

We will assume your diffusivity files are located in dtifit_folder but make sure to correct this to reflect your naming convention

a. We will assume your diffusivity files are located in dtifit_folder but make sure to correct this to reflect your naming convention

b. Mean and axial diffusivities are output as part of DTIFIT, but here we will compute the radial diffusivity as a mean of the second and third eigenvalue images.

c. The following is written in a loop so all subjects run in series, however this can be parallelized

d. Use the latest FSL version 5.0.7 if you have it, but the same code will work for older versions as well.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53

FSLDIR=/usr/local/fsl-5.0.7/
 
ENIGMAtemplateDirectory=/enigmaDTI/TBSS/ENIGMA_targets/
parentDirectory=/enigmaDTI/TBSS/run_tbss/
dtifit_folder=/enigmaDTI/DTIFIT/
 
mkdir ${parentDirectory}/MD/
mkdir ${parentDirectory}/AD/
mkdir ${parentDirectory}/RD/
 
cd $parentDirectory
 
 
for subj in subj_1 subj_2 … subj_N
do
   cp ${dtifit_folder}/${subj}*_MD.nii.gz ${parentDirectory}/MD/${subj}_MD.nii.gz
   cp ${dtifit_folder}/${subj}*_L1.nii.gz ${parentDirectory}/AD/${subj}_AD.nii.gz
   $FSLDIR/bin/fslmaths ${dtifit_folder}/${subj}*_L2.nii.gz –add ${dtifit_folder}/${subj}*_L3.nii.gz \\
       -div 2 ${parentDirectory}/RD/${subj}_RD.nii.gz
 
 
   for DIFF in MD AD RD
   do
   mkdir -p ${parentDirectory}/${DIFF}/origdata/
   mkdir -p ${parentDirectory}/${DIFF}_individ/${subj}/${DIFF}/
   mkdir -p ${parentDirectory}/${DIFF}_individ/${subj}/stats/
 
   $FSLDIR/bin/fslmaths ${parentDirectory}/${DIFF}/${subj}_${DIFF}.nii.gz -mas \\
      ${parentDirectory}/FA/${subj}_FA_FA_mask.nii.gz \\ 
      ${parentDirectory}/${DIFF}_individ/${subj}/${DIFF}/${subj}_${DIFF}
 
   $FSLDIR/bin/immv ${parentDirectory}/${DIFF}/${subj} ${parentDirectory}/${DIFF}/origdata/
 
   $FSLDIR/bin/applywarp -i ${parentDirectory}/${DIFF}_individ/${subj}/${DIFF}/${subj}_${DIFF} -o \\ 
      ${parentDirectory}/${DIFF}_individ/${subj}/${DIFF}/${subj}_${DIFF}_to_target -r \\ 
      $FSLDIR/data/standard/FMRIB58_FA_1mm -w ${parentDirectory}/FA/${subj}_FA_FA_to_target_warp.nii.gz
 
##remember to change ENIGMAtemplateDirectory if you re-masked the template
 
  $FSLDIR/bin/fslmaths ${parentDirectory}/${DIFF}_individ/${subj}/${DIFF}/${subj}_${DIFF}_to_target -mas \\
       ${ENIGMAtemplateDirectory}/ENIGMA_DTI_FA_mask.nii.gz \\
       ${parentDirectory}/${DIFF}_individ/${subj}/${DIFF}/${subj}_masked_${DIFF}.nii.gz	
 
   $FSLDIR/bin/tbss_skeleton -i ${ENIGMAtemplateDirectory}/ENIGMA_DTI_FA.nii.gz -p 0.049 \\
       ${ENIGMAtemplateDirectory}/ENIGMA_DTI_FA_skeleton_mask_dst.nii.gz $FSLDIR/data/standard/LowerCingulum_1mm.nii.gz \\         
       ${parentDirectory}/FA_individ/${subj}/FA/${subj}_masked_FA.nii.gz  \\ 
       ${parentDirectory}/${DIFF}_individ/${subj}/stats/${subj}_masked_${DIFF}skel -a \\ 
       ${parentDirectory}/${DIFF}_individ/${subj}/${DIFF}/${subj}_masked_${DIFF}.nii.gz -s \\ 
       ${ENIGMAtemplateDirectory}/ENIGMA_DTI_FA_skeleton_mask.nii.gz
 
 
   done
done

Now you should have your diffusivity skeletons!
**Check to make sure all skeletons cover the identical set of voxels, for example:

1
2
3
4
5

FSLDIR=/usr/local/fsl-5.0.7/
 
${FSLDIR}/bin/fslview  ${parentDirectory}/MD_individ/${subj}/stats/subj_1_masked_MDskel.nii.gz \\
     ${parentDirectory}/FA_individ/subj_1/stats/subj_1_masked_FAskel.nii.gz \\
     ${ENIGMAtemplateDirectory}/ENIGMA_DTI_FA_skeleton.nii.gz

Now we can extract ROI measures from these skeletons!
Remember you have already done this with FA
SCRIPT HERE

Here the runDirectory represents the directory where all your downloaded scripts and codes are located.
The descriptions of the commands (from the protocol) are below the code

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

parentDirectory=/enigmaDTI/TBSS/run_tbss/
runDirectory=/enigmaDTI/TBSS/run_tbss/
 
for DIFF in MD AD RD
do
   mkdir ${parentDirectory}/${DIFF}_individ/${DIFF}_ENIGMA_ROI_part1
   dirO1=${parentDirectory}/${DIFF}_individ/${DIFF}_ENIGMA_ROI_part1/
 
   mkdir ${parentDirectory}/${DIFF}_individ/${DIFF}_ENIGMA_ROI_part2
   dirO2=${parentDirectory}/${DIFF}_individ/${DIFF}_ENIGMA_ROI_part2/
 
   for subject in subj_1 subj_2 … subj_N
     do
 
     ${runDirectory}/singleSubjROI_exe ${runDirectory}/ENIGMA_look_up_table.txt \\ 
         ${runDirectory}/mean_FA_skeleton.nii.gz ${runDirectory}/JHU-WhiteMatter-labels-1mm.nii.gz \\ 
         ${dirO1}/${subject}_${DIFF}_ROIout ${parentDirectory}/${DIFF}_individ/${subject}/stats/${subject}_masked_${DIFF}skel.nii.gz
 
     ${runDirectory}/averageSubjectTracts_exe ${dirO1}/${subject}_${DIFF}_ROIout.csv ${dirO2}/${subject}_${DIFF}_ROIout_avg.csv
 
# can create subject list here for part 3!
     echo ${subject},${dirO2}/${subject}_${DIFF}_ROIout_avg.csv >> ${parentDirectory}/${DIFF}_individ/subjectList_${DIFF}.csv
   done
 
   Table=${parentDirectory}/ROIextraction_info/ALL_Subject_Info.txt 
   subjectIDcol=subjectID
   subjectList=${parentDirectory}/${DIFF}_individ/subjectList_${DIFF}.csv
   outTable=${parentDirectory}/${DIFF}_individ/combinedROItable_${DIFF}.csv
   Ncov=3  #2 if no disease
   covariates="Age;Sex;Diagnosis" # Just "Age;Sex" if no disease
   Nroi="all" 
   rois="all"
 
#location of R binary 
  Rbin=/usr/local/R-2.9.2_64bit/bin/R
 
#Run the R code
  ${Rbin} --no-save --slave --args ${Table} ${subjectIDcol} ${subjectList} ${outTable} \\ 
         ${Ncov} ${covariates} ${Nroi} ${rois} < ${runDirectory}/combine_subject_tables.R
done

1. The first command – singleSubjROI_exe uses the atlas and skeleton to extract ROI values from the JHU-atlas ROIs as well as an average diffusivity value across the entire skeleton

• It is run with the following inputs
• ./singleSubjROI_exe look_up_table.txt skeleton.nii.gz JHU-WhiteMatter-labels-1mm.nii.gz OutputfileName Subject_FA_skel.nii.gz
• The output will be a .csv file called Subject1_ROIout.csv with all mean FA values of ROIs listed in the first column and the number of voxels each ROI contains in the second column (see ENIGMA_ROI_part1/Subject1_ROIout.csv for example output)

2. The second command – averageSubjectTracts_exe uses the information from the first output to average relevant (example average of L and R external capsule) regions to get an average value weighted by volumes of the regions.

• It is run with the following inputs
• ./averageSubjectTracts_exe inSubjectROIfile.csv outSubjectROIfile_avg.csv
• where the first input is the ROI file obtained from Step 4 and the second input is the name of the desired output file.
• The output will be a .csv file called outSubjectROIfile_avg.csv with all mean FA values of the new ROIs listed in the first column and the number of voxels each ROI contains in the second column (see ENIGMA_ROI_part2/Subject1_ROIout_avg.csv for example output)

3. The final portion of this analysis is an ‘R’ script R that takes into account all ROI files and creates a spreadsheet which can be used for GWAS or other association tests. It matches desired subjects to a meta-data spreadsheet, adds in desired covariates, and combines any or all desired ROIs from the individual subject files into individual columns.

• Input arguments as shown in the bash script are as follows:
  • Table=./ALL_Subject_Info.txt
    • A meta-data spreadsheet file with all subject information and any and all covariates
  • subjectIDcol=subjectID
    • the header of the column in the meta-data spreadsheet referring to the subject IDs so that they can be matched up accordingly with the ROI files
  • subjectList=./subjectList.csv
    • a two column list of subjects and ROI file paths.
    • this can be created automatically when creating the average ROI .csv files – see sh on how that can be done
  • outTable=./combinedROItable.csv
    • the filename of the desired output file containing all covariates and ROIs of interest
  • Ncov=2
    • The number of covariates to be included from the meta-data spreadsheet
    • At least age and sex are recommended
  • covariates=”Age;Sex”
    • the column headers of the covariates of interest
    • these should be separated by a semi-colon ‘;’ and no spaces
  • Nroi=”all”
    • The number of ROIs to include
    • Can specify “all” in which case all ROIs in the file will be added to the spreadsheet
    • Or can specify only a certain number, for example 2 and write out the 2 ROIs of interest in the next input
  • rois= “all” #”IC;EC”
    • the ROIs to be included from the individual subject files
    • this can be “all” if the above input is “all”
    • or if only a select number (ex, 2) ROIs are desired, then the names of the specific ROIs as listed in the first column of the ROI file
      • these ROI names should be separated by a semi-colon ‘;’ and no spaces for example if Nroi=2, rois=”IC;EC” to get only information for the internal and external capsules into the output .csv file
    • (see csv for example output)

Congrats! Now you should have all of your subjects ROIs in one spreadsheet per diffusivity measure with only relevant covariates ready for association testing!