ANTsX / ANTs

Advanced Normalization Tools (ANTs)
Apache License 2.0
1.21k stars 381 forks source link

Feature Idea: antsApplyTransforms, combining timeseries, and antsMotionCorr warp fields #465

Open gdevenyi opened 7 years ago

gdevenyi commented 7 years ago

Taking a look at https://stnava.github.io/fMRIANTs/ it seems if one wants to combine the motion correction resampling step along with a coregistration and template build transform than a number of very large 4D warp files need to be generated and applied.

If I'm willing to use the output timeseries of antsMotionCorr instead, I can apply my transforms with antsApplyTransforms -d 3 -e 3.

It comes to mind that antsApplyTransforms might be able to handle the 4D warp field for the "-e 3" case internally, avoiding these issues.

I'm not sure how hard it would be to get working but it would be a huge improvement to the workflow for timeseries data.

gdevenyi commented 5 years ago

Same problem here https://github.com/ANTsX/ANTs/issues/700

wschelle commented 5 years ago

I am looking for a solution to the same problem. I have high spatial resolution MR BOLD data and wish to interpolate as little as possible. Ideally, the antsApplyTransforms function would also accept the 'MOCO_params.csv' file generated by antsMotionCorr.

Additionally, the example mentioned in the first post doesn't seem to be adjusting the original BOLD timeseries as the comments say happens. Am I missing something? Is this currently the only way to apply motion and registration transforms simultaneously? Thanks for your time!

stnava commented 5 years ago

for the most part, we use ANTsR or ANTsPy for such things as they both allow easy split/combine functionality which will let you perform whatever transformation combination(s) that you want.

brian

On Fri, Mar 8, 2019 at 11:53 AM wschelle notifications@github.com wrote:

I am looking for a solution to the same problem. I have high spatial resolution MR BOLD data and wish to interpolate as little as possible. Ideally, the antsApplyTransforms function would also accept the 'MOCO_params.csv' file generated by antsMotionCorr.

Additionally, the example mentioned in the first post doesn't seem to be adjusting the original BOLD timeseries as the comments say happens. Am I missing something? Is this currently the only way to apply motion and registration transforms simultaneously? Thanks for time!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ANTsX/ANTs/issues/465#issuecomment-470997835, or mute the thread https://github.com/notifications/unsubscribe-auth/AATyfup91AcP0BrbEZQB9e4dqQMXbXERks5vUpV8gaJpZM4OP6cM .

renilmathew commented 4 years ago

Hey! I successfully used the antsRegistrationSyNQuick on mean EPI image to get the files for antsApplyTransforms for high resolution fMRI BOLD data. I'm having trouble getting through the antsApplyTransform. Using a 4D file as input gave an error segmentation fault (core dumped).

I'm using ANTs on windows platform via Ubuntu 16.04. Anyway to get it done on this ubuntu subsystem

Thanks, R

cookpa commented 4 years ago

@renilmathew you can open your own issue to ask about this, please include a reproducible example if possible but if you can't share the data, at least the full ANTs version information and the commands you ran plus their terminal output.

renilmathew commented 4 years ago

@renilmathew you can open your own issue to ask about this, please include a reproducible example if possible but if you can't share the data, at least the full ANTs version information and the commands you ran plus their terminal output.

Sure!

dangom commented 3 years ago

I was looking for the requested feature as well. I came up with this wrapper function based on the split/merge suggestion mentioned above. Here's the code, in the hope that it may be useful (note that I use some of the FSL tools for split/merge for lack of knowledge about their ANTs equivalents):

#!/usr/bin/bash
# concat_moco_affine.sh

# Safety precautions
# Using a variable that is not set will raise an error
set -o nounset
# Exit if a command fails.
set -o errexit

function echo_help {
    echo "Humane wrapper to concatenate ANTs moco transforms with a generic affine"
    echo "Usage: concat_moco_affine.sh input.nii.gz warp.nii.gz affine.mat reference.nii.gz"
    echo "Concatanates and applies motion correction and affine in a single step."
}

if (($# < 1))
then
    echo_help
    exit 0
fi

fmri=$1
mocowarp=$2
affine=$3
ref=$4

tr=$(PrintHeader $fmri | grep "Voxel Spac" | cut -d ',' -f 4 | cut -d ']' -f 1)
nvols="$(($(fslnvols $fmri)-1))"

tmpfmri=$(mktemp -d)
fslsplit $fmri ${tmpfmri}/ -t

tmpwarp=$(mktemp -d)
fslsplit $mocowarp ${tmpwarp}/ -t

tmpout=$(mktemp -d)
parallel -j 4 antsApplyTransforms -d 3 -r $ref \
         -o $tmpout/{1}.nii.gz -n LanczosWindowedSinc -i $tmpfmri/{1}.nii.gz \
         -t $tmpwarp/{1}.nii.gz -t $affine -v 1 ::: $(seq -f "%04g" $nvols)

fslmerge -tr ${fmri/.nii.gz/space-fov.nii.gz} ${tmpout}/* $tr
exit 0
gdevenyi commented 3 years ago

The alternative would be probably from ImageMath:

 TimeSeriesDisassemble : Outputs n 3D image volumes for each time-point in time-series 4D image.
    Usage        : TimeSeriesDisassemble 4D_TimeSeries.nii.gz 

 TimeSeriesAssemble : Outputs a 4D time-series image from a list of 3D volumes.
    Usage        : TimeSeriesAssemble time_spacing time_origin *images.nii.gz