Motion and BOLD correction in SS-SI-VASO

Intro

This is a description on how to use SPM and/or AFNI for high-resolution SS-SI-VASO data evaluations.

Here it is explained what to do with the raw fMTR data from the scanner to obtain CBV and BOLD signal time courses .

Sequence

Every other image has a different contrast.

Screenshot 2017-11-25 15.27.14.png

SPM (AFNI and FSL work too):

  • spline interpolation
  • weighting of interesting areas
  • correct individual contrasts separately
  • manual check for “usual suspects”:
  • task-correlated motion
  • different motion in BOLD and VASO
  • amount of translation:
  • disregard dataset, when > voxel size
  • dropout rate approx. 30%

dominant in phase encoding direction.

image10.png

always check that the motion parameters are the same in BOLD and VASO.

Screenshot 2017-11-25 15.29.39.png

Code:

https://github.com/layerfMRI/repository/tree/master/motion

start_moco_*.sh:

  • gets rid of non-steady state first time points
  • grabs the SPM moco script: *batch2run.m
  • runs the SPM moco script in Matlab
  • plots motion traces. ALWAYS LOOK AT THEM

afni_VASO_eval.sh:

  • combines the SPM output and calculared separate BOLD and VASO time courese
  • It also caclulated MEAN, TSNR, T1-stability images:
#!/bin/bash

echo "It starts now:    I expect two files Not_Nulled_Basis_a.nii and Nulled_Basis_b.nii that are motion corrected with SPM"

NumVol=`3dinfo -nv Nulled_Basis_b.nii`
3dcalc -a Nulled_Basis_b.nii'[3..$]' -b  Not_Nulled_Basis_a.nii'[3..'`expr $NumVol - 1`']' -expr 'a+b' -prefix combined.nii -overwrite
3dTstat -cvarinv -prefix T1_weighted.nii -overwrite combined.nii
rm combined.nii

3dcalc -a Nulled_Basis_b.nii'[1..$(2)]' -expr 'a' -prefix Nulled.nii -overwrite
3dcalc -a Not_Nulled_Basis_a.nii'[2..$(2)]' -expr 'a' -prefix BOLD.nii -overwrite

3drefit -space ORIG -view orig -TR 3 BOLD.nii
3drefit -space ORIG -view orig -TR 3 Nulled.nii

3dTstat -mean -prefix mean_nulled.nii Nulled.nii -overwrite
3dTstat -mean -prefix mean_notnulled.nii BOLD.nii -overwrite

# Then calculate the mean, stdev, & cvarinv (detrended mean/stdev) for BOLD & Nulled volumes
# I was mainly doing this as a check to make sure motion correction worked reasonably.
# A TSNR calculation should probably be done after non-linear drift removal.

# Calculate VASO!

  # The first vaso volume is first nulled volume divided by the 2nd BOLD volume
  3dcalc -prefix tmp.VASO_vol1.nii \
         -a      BOLD.nii'[1]' \
         -b      Nulled.nii'[0]' \
         -expr 'b/a' -overwrite

  # Calculate all VASO volumes after the first one
  # -a goes from the 2nd BOLD volume to the 2nd-to-last BOLD volume
  # -b goes from the 3rd BOLD volume to the last BOLD volume
  # -c goes from the 2nd Nulled volume to the last Nulled volume
  NumVol=`3dinfo -nv BOLD.nii`
  3dcalc -prefix tmp.VASO_othervols.nii \
         -a      BOLD.nii'[0..'`expr $NumVol - 2`']' \
         -b      BOLD.nii'[1..$]' \
         -c      Nulled.nii'[1..$]' \
         -expr 'c*2/(a+b)' -overwrite

   # concatinate the first VASO volume with the rest of the sequence
   3dTcat -overwrite -prefix VASO.nii tmp.VASO_vol1.nii tmp.VASO_othervols.nii

# Remove the temporary seprate files for the first VASO volume and the rest of the VASO volumes
rm tmp.VASO_vol1*.nii tmp.VASO_othervols*.nii

# BOLD and VASO now have 3s or 312s TRs
# Use 3drefit to make this change in the file headers

#echo ' remove last time point in BOLD'
#NumVol=`3dinfo -nv BOLD.nii`
#echo "NumVol of BOLD is '$Numvol'  "
#3dcalc -prefix BOLD.nii \
#         -a      BOLD.nii'[1..'`expr $NumVol - 1`']' \
#         -expr 'a' -overwrite

  3dTstat  -overwrite -mean  -prefix BOLD.Mean.nii \
     BOLD.nii'[1..$]'
  3dTstat  -overwrite -cvarinv  -prefix BOLD.tSNR.nii \
     BOLD.nii'[1..$]'
  3dTstat  -overwrite -mean  -prefix VASO.Mean.nii \
     VASO.nii'[1..$]'
  3dTstat  -overwrite -cvarinv  -prefix VASO.tSNR.nii \
     VASO.nii'[1..$]'

Example Data

Accronis access: Sync & Share  >> Example_data >>  example_from_Neuron_paper >> tapping_with_touch

 

Other ways of doing it

There are multiple different versions of this:

  • from Renzo: Using SPM and NeuroDebian <- recommended and explained here.
  • important files for SPM motion correction are: start_moco_VASO*.sh, and *mocobatch*.sh
  • important files for BOLS correction are: antiBold on Neurodebian or afni_VASO_eval_SPM.sh

Alternative software versions