Summary of what happened:
I am having some difficulty with FSL’s mcflirt and flirt and would love some advice! I am trying to register a T1w image into DCE space and then apply that transform to other ROIs, but the motion correction seems to misalign the images, and the coregistration to the mean image results in distorted, misaligned images. Can anybody please advise?
Command used (and if a helper script was used, a link to the helper script or the command generated):
# Create 4D file and save path
fslmerge -t "$OUTPUT_DIR"/dce_timeseries.nii.gz "$INPUT_DIR"/dce_*.nii
TIMESERIES="$OUTPUT_DIR/dce_timeseries.nii.gz"
# Motion correction
# mcflirt -in "$TIMESERIES" -cost mutualinfo -out "$MOTION_CORRECTED" -plots
# Compute mean image
fslmaths "$MOTION_CORRECTED" -Tmean "$MEAN_IMAGE"
# Register ROI image to motion-corrected DCE image using FSL FLIRT
flirt -in "$ROI_IMAGE" -ref "$MEAN_IMAGE" -cost mutualinfo -searchcost mutualinfo -schedule $FSLDIR/etc/flirtsch/simple3D.sch -out "$ROI_REGISTERED" -interp nearestneighbour -omat "$OUTPUT_DIR/ROI_to_MEAN.mat"
# Apply transformation from ROI registration to AIF image
flirt -in "$AIF_IMAGE" -ref "$MEAN_IMAGE" -out "$AIF_REGISTERED" -applyxfm -init "$OUTPUT_DIR/ROI_to_MEAN.mat" -interp nearestneighbour
# Apply transform to T1 map
flirt -in "$T1_IMAGE" -ref "$MEAN_IMAGE" -out "$OUTPUT_DIR/T1map_to_DCE.nii.gz" -applyxfm -init "$OUTPUT_DIR/ROI_to_MEAN.mat" -interp nearestneighbour
Version:
FSL version 6.0.7.15
Environment (Docker, Singularity / Apptainer, custom installation):
Custom installation using HomeBrew. Test tasks run appropriately.
Relevant log outputs (up to 20 lines):
No relevant errors in the logs. All commands run appropriately, but motion correction and registration aren’t performing as expected.
Screenshots / relevant information:
Other steps I’ve taken to troubleshoot: I’ve tried coregistering to the first time point to rule out intensity issues, but this did not resolve the problem. I have experimented with different cost functions, including mutualinfo
, corratio
, and normcorr
, but have not noticed any improvement. I am trying to register a T1w image into DCE space and then apply that transform to other ROIs.
Thank you!!