Skip to content

Latest commit

 

History

History
136 lines (109 loc) · 3.59 KB

morphometry.md

File metadata and controls

136 lines (109 loc) · 3.59 KB
Raw

How to Process MRI Images:

  1. Beginner's Guide to Command-Line Interface
  2. Installation Instructions
  3. Preprocessing T1 images
  4. Hippocampus Segmentation

Reading Materials and Lecture Slides

For reference, your fixed image is your template and your moving image is your participant.

ANTs Nonlinear Registration

In your command window, type:

antsRegistrationSyn.sh \
-d 3 \
-f <fixedImage>.nii.gz \
-m <movingImage>.nii.gz \
-o <outputPrefix> \
-t s

To run a quicker version, type:

antsRegistrationSynQuick.sh \
-d 3 \
-f <fixedImage>.nii.gz \
-m <movingImage>.nii.gz \
-o <outputPrefix>
-t s

Both of these commands are simplified shell scripts of the actual command antsRegistration. You can view the code in these scripts by running either:

cat /usr/local/antsbin/bin/antsRegistrationSyn.sh
cat /usr/local/antsbin/bin/antsRegistrationSynQuick.sh

One of the differences between the two commands is the number of iterations in the last iteration step. For antsRegistrationSynQuick.sh there are 0 iterations and for antsRegistrationSyn.sh there are 100 iterations. The other difference is the metric used for registration. For antsRegistrationSynQuick.sh, the registration metric is based on mutual information (great for multimodal images). For antsRegistrationSyn.sh, the registration metric is based on cross correlation (great for T1 weighted image registration).

Generate Warped Images

# Make the participant image look like the template
WarpImageMultiTransform 3 \
<movingImage>.nii.gz \
<outputImage>.nii.gz \
<outputPrefix>Warp.nii.gz \
<outputPrefix>Affine.txt \
-R <fixedImage>.nii.gz

# Make the template look like the participant
WarpImageMultiTransform 3 \
<fixedImage>.nii.gz \
<outputImage>.nii.gz \
-i <outputPrefix>Affine.txt \
<outputPrefix>InverseWarp.nii.gz \
-R <movingImage>.nii.gz

Generate the Grid Image

CreateWarpedGridImage 3 \
<outputPrefix>Warp.nii.gz \
<outputGrid>.nii.gz \
1x1x1 \ 
5x5x5

Generate the Jacobian

# Jacobian
CreateJacobianDeterminantImage 3 \
  <outputPrefix>Warp.nii.gz \
  <outputJac>.nii.gz \
  0 \ # Default
  1
  
# Log Jacobian
CreateJacobianDeterminantImage 3 \
  <outputPrefix>Warp.nii.gz \
  <outputLogJac>.nii.gz \
  1 \ # Take the log of the Jacobian  
  1

Modulate the Gray Matter Image

c3d \
<inputImage>.nii.gz \
<jacobianImage>.nii.gz \
-multiply \
-o <modulatedImage>.nii.gz

If you have your tissue segmentation ROIs (see Preprocessing T1 example), then split the ROIs, so you just have a gray matter ROI. You will extract just the modulated GM from the brain image using this mask.

c3d \
<modulateImage>.nii.gz \
<gmMask>.nii.gz \
-multiply \
-o <outputImage>.nii.gz

Evaluate Image Similarity

ImageMath \
3 \
<outputFile>.ext \
NormalizedCorrelation \
<image1>.nii.gz \
<image2>.nii.gz

Further Readings