Preprocessing Diffusion Images

T1

T2

DWI

DWI vs. DTI

Diffusion weighted imaging

  • Measures water move within a voxel of tissue
  • Intensity value represents rate of water diffusion

Improvidence tensor imaging

  • Specific type of modeling of diffusion weighted images based on the theory that water molecules diffuse differently within unlike types of tissue

Verify Volumes

It is very piece of cake to have missing volumes in the DWI. It is critical that the numbers match between the NIFTI image and the bvec and bval files.

                                                $                  c4d dwi.nii.gz                  -info                                                  Image                  #one: dim = [104, 104, 69, 63];  bb = {[-103.445 91.427 -36.6199 0], [104.555 299.427 101.38 819]};  vox = [2, 2, 2, 13];  range = [0, 2910];  orient = Oblique, closest to RPI

This NIFTI prototype has 63 volumes with a FOV of 104 x 104 ten 69. Make sure in that location's 63 entries in the bval file:

                                                $                  cat dwi.bval                  |                  wc                  -w                                                  63

Finally, make certain you have the same number of volumes in bvec file:

                                                $                  head dwi.bvec                  |                  wc                  -w                                                  63

Pad Paradigm

If your FOV box is right upwards against skull or even cutting off part of skull merely not brain, you will need to pad the image to give background space:

                                                $                  c4d                  \                                dwi.nii.gz                  \                                -pad 10% 10% 0                  \                                -o dwi_padded.nii.gz

Mask B0

Must exist equally authentic as possible.

pnlNipype

Using nifti_bet_mask that packages FSL:

                                                      ~/build/pnlNipype/exec/nifti_bet_mask                    \                                    -i dwi_bse.nii.gz                    \                                    --bvals dwi.bval                    \                                    -o dwi

Using nifti_atlas that packages ANTs:

                                                      ~/build/pnlNipype/exec/nifti_atlas                    \                                    -t dwi_bse.nii.gz                    \                                    --train ~/apps/pnlbwh/pnlNipype/trainingDataT2Masks-12a14d9/trainingDataT2Masks-hdr.csv                    \                                    -o dwi

Make clean mask

Sometimes all that's needed is to erode / shrink the mask a little to make it better

                                                      c3d                    \                                    dwi_mask.nii.gz                    \                                    -erode 1 1x1x1vox                    \                                    -o dwi_mask_cleaned.nii.gz

Eddy Current Correction

This correction volition help set up eddy electric current-induced distortions and subject movements during the scan.

Set acqparams.txt

All data can be derived / found in the DWI JSON file. For your acqparams.txt file, each row consists of a vector (three values) that specify what the phase-encode (PE) axis is. The fourth chemical element in each row is the time (in seconds) between reading the middle of the start echo and reading the center of the concluding repeat. The fourth element is also the reciprocal of the PE bandwidth/pixel. Remember that to calculate the reciprocoal you lot take: (1 BandwidthPerPixelPhaseEncode)

                                                      printf                    "0 1 0 0.095"                    >                    acqparams.txt

Forward

  • Posterior to Inductive
  • Phase Encoding is j
  • acqparams.txt is 0 1 0

Contrary

  • Anterior to Posterior
  • Phase Encoding is j-
  • acqparams.txt is 0 -1 0

Looking at the JSON file:

  • "PhaseEncodingDirection": "j-,"
  • "BandwidthPerPixelPhaseEncode": 13.935,
    • (one = 0.07176175099)
                                                      printf                    "0 -ane 0 0.072"                    >                    acqparams.txt

Fix index.txt

Create an index file that tells eddy which line of the "acqparams.txt" file to employ for each book of data. In our case above, the index file has a single line with 63 i's separated by spaces:

                                                      myVar                    =                    ($(                    wc                    -w                    dwi.bval))                                                        indx                    =                    ""                                                        for                    ((                    i                    =                    1                    ;                    i                    <=                    $myVar                    ;                    i                    +=                    1                    ));                    do                    indx                    =                    "                    $indx                                          1"                    ;                    done                                                        echo                    $indx                    >                    index.txt

Eddy

                                                      fourth dimension                    eddy_cuda8.0                    \                                    --imain=dwi_xc.nii.gz                    \                                    --acqp=acqparams.txt                    \                                    --index=index.txt                    \                                    --mask=dwi_bse_mask.nii.gz                    \                                    --bvecs=dwi_xc.bvec                    \                                    --bvals=dwi_xc.bval                    \                                    --out=dwi_eddy                    \                                    --cnr_maps                    \                                    --repol                    \                                    --mporder=xvi                    \                                    --estimate_move_by_susceptibility                    \                                    --verbose

One mode to check eddy correction is to create a gif image that scans through each volume forth the same slice. I've create an R script that can salve a gif for each participant in your report:

                                                      library(oro.nifti)                                      library(neurobase)                                      setwd('/fslhome/intj5/compute/information/BRAINSCOPE/pnlNipype/')                  dirs=                    list.files(path=                    ".")                                      for                    (ten                    in                    ane                    :                    length(dirs)){                                      setwd(paste('/fslhome/intj5/compute/data/BRAINSCOPE/pnlNipype/', dirs[x],                    sep =                    ""))                                      getwd()                                      img                    <-                    readNIfTI("dwi_eddy.nii.gz")                                      png(file=                    "eddy%02d.png",                    width=                    200,                    height=                    200)                                      for                    (i                    in                    1                    :img@dim_[5]){                                      oro.nifti::                    prototype(robust_window(img),z =                    35,due west =                    i,plot.blazon =                    "single")                                      }                                      dev.off()                                      system("catechumen -delay 8 *.png eddy.gif")                                      file.remove(list.files(pattern=                    ".png"))                  }

Eddy Quality Control

For full information: https://fsl.fmrib.ox.ac.united kingdom/fsl/fslwiki/eddyqc/UsersGuide

Participant

                                                      for                    i                    in                    $(                    ls                    <path-to-dir>/pDWI/                    |                    grep                    "sub-"                    )                    ;                    do                                                        SUBJ_DIR                    =<path-to-dir>/pDWI/${i}                                                        eddy_quad                    \                                                        ${SUBJ_DIR}/dwi_eddy                    \                                                        -idx                    ${SUBJ_DIR}/index.txt                    \                                                        -par                    ${SUBJ_DIR}/acqparams.txt                    \                                                        -chiliad                    ${SUBJ_DIR}/dwi_b0_mask.nii.gz                    \                                                        -b                    ${SUBJ_DIR}/dwi.bval                    \                                                        -5                                                        done

Study

                                                      eddy_squad                    \                                    participants.txt                    \                    # Listing with full-path proper noun to dwi_eddy.qc directories                                      -g                    grouping.txt                    \                                    -u                    \                                    -o squad_site

The correct way to format the group.txt file:

                                                      VarName                                                        0                    [1 for continuous]                                      v_1                                                        v_2                                                        .                                                        .                                                        .                                                        v_n

Notation that the variable must always be numeric regardless if the variable is categorical or continuous.

EPI Baloney Correction

Create a skull stripped B0 image

It is easier to register the B0 and T1 if they are appropriately skull stripped. The skull will cause registration errors.

Fix Displacement between B0 and T1

Do a rigid alignment between preprocessed T1 and boil corrected B0 image to eliminate whatsoever displacement / motion differences betwixt the two images.

                                                      antsRegistrationSyNQuick.sh                    \                                    -d iii                    \                                    -f b0.nii.gz                    \                                    -one thousand t1.nii.gz                    \                    # skull stripped image from FreeSurfer                                      -o                    t1tob0-                    \                                    -t a

Fix Distortions

Non-linearly register B0 to T1 to fix distortions (stretching and squishing). You are not moving the b0 image, only adjusting voxels out of place. You do not need to adjust bvec file afterwards.

                                                      antsRegistrationSyNQuick.sh                    \                                    -d 3                    \                                    -f t1tob0-Warped.nii.gz                    \                                    -k b0.nii.gz                    \                                    -o epi-                    \                                    -t so

Employ Transformation

Utilise the registration to the entire eddy electric current corrected paradigm.

                                                      antsApplyTransforms                    \                                    --dimensionality 3                    \                                    --input-image-type 3                    \                                    --input eddy.nii.gz                    \                                    --output epi.nii.gz                    \                                    --reference-image t1tob0-Warped.nii.gz                    \                                    --interpolation Bspline                    \                                    --transform epi-1Warp.nii.gz

Topup

Create a Symlink for all your files:

                                                ln                  -fs                  ${BIDS_DIR}/dwi/${i}_${2}_acq-AP_dwi.nii.gz AP_scan.nii.gz                                  ln                  -fs                  ${BIDS_DIR}/dwi/${1}_${2}_acq-PA_dwi.nii.gz PA_scan.nii.gz                                  ln                  -fs                  ${BIDS_DIR}/dwi/${1}_${2}_acq-AP_dwi.bval AP_scan.bval                                  ln                  -fs                  ${BIDS_DIR}/dwi/${1}_${2}_acq-AP_dwi.bvec AP_scan.bvec                                  ln                  -fs                  ${BIDS_DIR}/dwi/${1}_${ii}_acq-PA_dwi.bval PA_scan.bval                                  ln                  -fs                  ${BIDS_DIR}/dwi/${1}_${2}_acq-PA_dwi.bvec PA_scan.bvec

Extract B0 for each scan and combine together:

                                                fslroi                  AP_scan b0_AP_scan 0 1                                  fslroi                  PA_scan b0_PA_scan 0 1                                  fslmerge                  -t                  b0_AP_PA_scan b0_AP_scan b0_PA_scan

Create an acqparams.txt file:

                                                printf                  "0 -i 0 0.0805839"                  >                  acqparams.txt                                  printf                  "\n0 1 0 0.0805839"                  >>                  acqparams.txt

Run topup:

                                                topup                  \                                --imain=b0_AP_PA_scan                  \                                --datain=acqparams.txt                  \                                --out=topup                  \                                --config=b02b0.cnf                  \                                --iout=b0_unwarped

Create a brain mask:

                                                fslmaths                  b0_unwarped                  -Tmean                  b0_unwarped_mean                                  bet                  b0_unwarped_mean b0_unwarped_brain                  -yard                  -f                  0.1

Create a single 4D image:

                                                fslmerge                  -a                  AP_PA_scan AP_scan PA_scan

Combine the bvec and bval files:

                                                paste                  -d                  " "                  AP_scan.bvec PA_scan.bvec                  >>                  bvecs                                  paste                  -d                  " "                  AP_scan.bval PA_scan.bval                  >>                  bvals

Create index.txt:

                                                myVar                  =                  ($(                  wc                  -west                  AP_scan.bval))                                                  indx                  =                  ""                                                  for                  ((                  i                  =                  1                  ;                  i                  <=                  $myVar                  ;                  i                  +=                  1                  ));                  do                  indx                  =                  "                  $indx                                      1"                  ;                  done                                                  for                  ((                  i                  =                  1                  ;                  i                  <=                  $myVar                  ;                  i                  +=                  1                  ));                  exercise                  indx                  =                  "                  $indx                                      ii"                  ;                  done                                                  echo                  $indx                  >                  index.txt

Run eddy:

                                                eddy_openmp                  \                                -v                  \                                --imain=AP_PA_scan                  \                                --mask=b0_unwarped_brain_mask                  \                                --index=alphabetize.txt                  \                                --acqp=acqparams.txt                  \                                --bvecs=bvecs                  \                                --bvals=bvals                  \                                --topup=topup                  \                                --out=eddy                  \                                --cnr_maps                  \                                --repol                  \                                --mporder=sixteen                  \                                --estimate_move_by_susceptibility                  \

At this point you do not demand to run any epi distortion corrections since topup was included with eddy to set up the distortions.

Generate New Mask

A concluding mask needs to be generated earlier tensor fitting.