Skip to content

Latest commit

 

History

History
121 lines (76 loc) · 5.06 KB

Process_T1_anatomical.md

File metadata and controls

121 lines (76 loc) · 5.06 KB
Raw

Processing the T1 anatomical

The standard location for T1 data is /NHP_MRI/Data_proc/SUBJECT/YYYYMMDD/anat/T1. Example scripts are at: /NHP_MRI/Data_proc/EDDY/20170614/anat/T1/process_t1-step*.sh. Alternatively, after running the initial minimal_processing and isotropic_resampling steps with nipype in NHP-BIDS you can find them in /NHP-MRI/NHP-BIDS/derivatives/resampled-isotropic-06mm/sub-<subject>/ses-<yyyymmdd>/anat.

Pre-processing the indivisual T1's

If you take the data from Data_proc they still need to be preprocessed. If you are combining different types of T1's, use something like the following:

#!/bin/bash

set -e # exit if a command fails

# --- First process the T1's individually
FILES="T1_07.nii.gz
T1_10.nii.gz
T1_12.nii.gz"

preprocess_indiv () {
  local f=$1
  b=${f%.nii.gz}
  mri_convert -i ${b}.nii.gz -o ${b}_ro.nii.gz --sphinx -vs 0.5 0.5 0.5
  fslreorient2std ${b}_ro.nii.gz ${b}_ro.nii.gz
  mri_nu_correct.mni --i ${b}_ro.nii.gz --o ${b}_nu.nii.gz --distance 24
}

for f in $FILES; do
  preprocess_indiv $f &
done

wait # wait for individual file preprocessing

# --- Next, average together with motion correction
mri_motion_correct.fsl -o T1_avg_nu.nii.gz \
    -i T1_10_nu.nii.gz -i T1_15_nu.nii.gz

Note that N4BiasFieldCorrection could be used instead of mri_nu_correct.mni, if mri_nu_correct.mni is difficult to get working on your system.

T1's that are already pre-processed in nipype, don't need to be re-oriented, but you might still want to perform the bias field correction.

Brain extraction / skull-stripping

Brain extraction can be done several ways:

Using the BET command

Get the approximate middle coordinate (somewhere in the pons) and write them down ()

fslview T1_image &

JW: BET might work better a little higher up, like this:

BET center

Extract the brain using fsl’s BET routine (you may need to tweak the optional parameters a bit for the best result)

bet inputfile outputfile(preferably ‘inputfile_brain’) -f 0.55 -c <x y z>

ME note:

  • running bet with -R (robust option) gives generally favorable results
  • tuning the -f parameter around the range f=0.11-0.9 gives results that include the frontal pole (but also include some non-brain tissue anterior to the pons; in my experience that had no negative effect for the registration).

You can also use the gui:

Bet &

Performing Brain extraction with Freesurfer

Brain extraction sometimes works better with Freesurfer

(JW) from 20170524/anat/T1/process_t1-step2_freesurfer.sh

# -- Convert for freesurfer
# Not performing crop, b/c it doesn't seem needed and I want to put back in the
#   coordinates.
mri_convert -i T1_avg_nu.nii.gz -o orig.mgz --in_orientation RAS -iis 1 -ijs 1 -iks 1

# Should give different subjid for each run or delete existing folder (or not use the -i option below...)
subjid=EDDY_20170524-2
recon-all -i orig.mgz -subjid $subjid -autorecon1 -skullstrip -no-wsgcaatlas -wsthresh 15 -notal-check -notalairach -clean-bm -gcut

# Lowering threshold will reduce the brain region, could be reduced further
#recon-all -i orig.mgz -subjid $subjid -autorecon1 -no-wsgcaatlas -wsthresh 20 -notal-check -notalairach

#recon-all -subjid $subjid -skullstrip -no-wsgcaatlas -wsthresh 15 -notal-check -notalairach -clean-bm -gcut

# Copy file from freesurfer subject directory to current directory
# will need to be modified, freesurfer must provide a better way to do this...
cp /big/freesurfer-subjects/$subjid/mri/brainmask.mgz .

# Return to original scale and coordinates
mri_convert -i brainmask.mgz -o brainmask.nii.gz --out_orientation RAS -iis 0.5 -ijs 0.5 -iks 0.5

# "-n" prevents overwritting (in my version)
cp -n brainmask.nii.gz brainmask.manual.nii.gz

# Convert into mask (0 or 255)
fslmaths brainmask.manual.nii.gz -bin -mul 255 brainmask.manual.nii.gz

# Edit with freeview
freeview T1_avg_nu.nii.gz brainmask.manual.nii.gz:colormap=Heat:opacity=0.4

Using the NMT template as a prior

An alternative way to perform skullstripping, registration to template & atlas, and segmentation is by using the NIMH Macaque Template in /NHP-MRI/Template/NMT/NMTv1.2.

Create a folder for your subject in <>/NMTv1.2/single_subject_scans and copy thealign_and_process.sh script to it together with your (averaged) T1 image. Adjust the script to your needs (documentation in the script) and run it.

This will:

  • warp your T1 to the NMT template
  • provide non-linear warps for going back-and-forth in single-subject and template space
  • warp the D99 atlas to your subject
  • segment your T1, using the template segmentation as priors.
  • skullstrip your T1

If for some reason the brainmask is off, you can now warp the template mask to single-subject space and apply it to create the skullstripped brain volume:

fslmaths <Subject_BrainWithSkull.nii.gz> -mas <WarpedBrainMask.nii.gz> <Subject_Brain.nii.gz>