• LeftRightReversal

This page describes the steps taken to Left-Right reverse a subject in order to validate that there was a RH dominant asymmetry in the hippocampal volumes and not a bias introduced by the segmentation atlas.

To Left-Right reverse your subject:

(all commands run from within your subjects mri/ subdirectory)

First you need to get the orientation string of the volume:

mri_info --orientation orig.mgz

This will return either LIA or RIA. Then run mri_convert with that orientation string, but changing L to R:

mri_convert --in_orientation RIA orig.mgz orig.lrflipped.mgz

You can check visually to see that these two volumes are actually a reversal of each other:

tkmedit -f orig.mgz -aux orig.lrflipped.mgz

When you are satisfied that you have a reversal you should make a new subject directory within your SUBJECTS_DIR and copy this reversed volume in as the new subject's orig volume:

mkdir -p $SUBJECTS_DIR/LRflipped-subject/mri
cp orig.lrflipped.mgz $SUBJECTS_DIR/LRflipped-subject/mri/orig.mgz

Launch recon-all as normal, running only through the skull strip:

recon-all -autorecon1 -nomotioncor -s LRflipped-subject

*the -nomotioncor flag is used because we've copied in an orig, but have not 001 volume.

Running the aseg on your LR flipped subject

Before you can run the aseg you need to fix the RAS coordinates, otherwise the aseg will recognize and correct for the inversion. To do this, convert the brainmask.mgz and nu.mgz volumes to COR-format (because these are the two inputs for the aseg).

mri_convert brainmask.mgz -ot COR brainmask

Now, using the text editor of your choice, open the brainmask/COR-.info file.

imnr0 1
imnr1 256
ptype 2
x 256
y 256
fov 0.256
thick 0.001
psiz 0.001
locatn 0
strtx -0.128
endx 0.128
strty -0.128
endy 0.128
strtz -0.128
endz 0.128
tr 0.000000
te 0.000000
ti 0.000000
flip angle 0.000000
ras_good_flag 1
x_ras -1.000000 -0.000000 -0.000000
y_ras -0.000000 -0.000000 -1.000000
z_ras 0.000000 1.000000 0.000000
c_ras 5.186159 -15.249554 -0.080511

Remove the last four lines from the COR-.info file and set the ras_good_flag to zero:

imnr0 1
imnr1 256
ptype 2
x 256
y 256
fov 0.256
thick 0.001
psiz 0.001
locatn 0
strtx -0.128
endx 0.128
strty -0.128
endy 0.128
strtz -0.128
endz 0.128
tr 0.000000
te 0.000000
ti 0.000000
flip angle 0.000000
ras_good_flag 0

Now, conver the COR files back to mgz format:

mri_convert brainmask/ brainmask.mgz

and run the aseg as normal:

recon-all -subcortseg -s LRflipped-subject"

Our Results

When we first looked at the volumes of the hippocampus we saw a right-hemisphere dominant asymmetry. Using these two subjects as an example, the volumes were:

• orig_010103_61864:

  • Left-Hippocampus_vol = 2627.0
    Right-Hippocampus_vol = 3195.0
    

  orig_030324_vc11522:

  • Left-Hippocampus_vol = 3137.0
    Right-Hippocampus_vol = 3771.0
    

  orig volumes show a RH dominant asymmetry

After left-right reversal of our subjects the hippocampal volumes show a left-hemisphere dominant asymmetry, indicating that the asymmetry is not due to introduced bias but a reflection of the MR data.

• LR_flipped-010103_61864:

  • Left-Hippocampus_vol = 3023.0
    Right-Hippocampus_vol = 2688.0
    

  LR_flipped-030324_vc11522:

  • Left-Hippocampus_vol = 3685.0
    Right-Hippocampus_vol = 3218.0
    

  LR_flipped volumes show a LH dominant asymmetry