BOLD Brain Mask too wide besides T1 Brain Mask being fine

Software Support
1

Dear Experts,

we have been running into an odd problem with the BOLD-based brain mask in some participants. This id always located ventral and prefrontally, to varying grades of severity. I appended a picture of the brain masks generated on the T1-weighted and BOLD-based masks below.
I found one post in which this has happened before, which has sadly not been resolved to date.
As I am somewhat new to fMRI analyses, I am sorry if I missed something obvious as to why this might happen.

Thanks and regards,
Sebastian

Command used:

singularity run -B ${DIR_PARENT} /opt/software/fmriprep-20.2.7/fmriprep-20.2.7.sif \
  ${DIR_BIDS} ${DIR_FMRIPREP} \
  participant \
  --participant-label $1 \
  --skip-bids-validation \
  --fs-license-file ${DIR_SCRIPTS}/license.txt \
  --fs-no-reconall \
  --output-spaces MNI152NLin6Asym:res-2 \
  --stop-on-first-crash \
  -w ${DIR_FMRIPREP}  2>&1 | tee -a $fmriprep_log

Version:

20.2.7 (We had some issues with our sequences and SDC with newer versions)

Environment:

Singularity container accessed by bash script.
Scanner was a Siemens Prisma Fit 3T.

Screenshots / relevant information:

combinedMasks
combinedMasks1864×1568 317 KB

1 Like
2

Based on a suggestion by @LoefflerMartin I have created a work-around which at least improves on this issue.

  1. Use FSL’s BET to extract the brain from the functional data as precisely as possible by playing around with the threshold value
  2. Use fslmaths -bin to create a binarized mask of the extracted brain
  3. Use fslmaths -dilM 2-3 times to create a binarized mask of increased size
  4. Subtract the binarized mask from it’s dilated counterpart and invert the result to get a black outline around the cortex
  5. Multiply the original input file with the black outline (halo) you created to get nii.gz file that has a very clear outline around the brain
  6. Use these images as your input for fmriprep

I packaged the whole thing into a bash script that takes input, output and the threshold value for bet as inputs and then does all the bothersome fslmaths stuff for you:

#!/bin/bash

# Check if arguments are provided
if [ $# -lt 3 ]; then
    echo "Usage: $0 <input_file> <output_file> <bet_threshold>"
    echo "Provide identical input and output to overwrite original file"
    exit 1
fi

# Get input and output filenames from command line arguments
input_file=$1
output_file=$2
bet_threshold=$3

echo "Starting bordering for $input_file"
# Run computation
echo "Extracting brain at $bet_threshold threshold"
bet $input_file tmp_brain -f $bet_threshold -g 0
echo "Creating binarized mask"
fslmaths tmp_brain -bin tmp_bin
echo "Dilating binarized mask"
fslmaths tmp_bin -dilM -dilM -dilM tmp_dil
echo "Creating halo mask"
fslmaths tmp_dil -sub tmp_bin tmp_halo_inverted
fslmaths tmp_halo_inverted -mul -1 -add 1 tmp_halo
echo "Subtracting halo mask from original image"
fslmaths $input_file -mul tmp_halo $output_file

# Clean-Up
echo "Removing temporary files"
rm tmp*

# Check if the output file was created successfully
if [ $? -ne 0 ]; then
    echo "Error: Failed to create output file."
else
    echo "fslmaths command completed successfully."
fi

I appended on of my first results below, I still have to tinker with the size of the outline to possibly get better results, but it’s a start. If you have any comments or suggestions, I would be glad to hear them!

fmriprep_beforeafter

1 Like