Fmriprep - wired brain extraction

bear_sun · July 1, 2020, 1:12am

Hi, there,

I got some weird brain mask from fmriprep. Could you please help me to see what is going on? Thx!

bear_sun · July 1, 2020, 1:12am

White is the mask from FSL bet, red is the mask from fmriprep:

alexlicohen · July 2, 2020, 10:35am

If you look at the sag slices in your fmriprep output, you will notice how much brighter the parietal/occipital regions are compared to the inf. frontal regions. This requires ‘intensity normalization’ which IS a part of fmriprep, but occurs after brainmask creation. I’m not sure exactly why it failed, but if it is just one subject, you could try something like the linked code (you’ll need to download the whole set of scripts for it to work, or read through the code and just use the underyling FSL tool you need) to just prepare the T1 image first, then hand it back to fmriprep:

github.com

bchcohenlab/bids_lesion_code/blob/master/prep_T1w.sh

#!/bin/bash

: <<COMMENTBLOCK
Call fsl_anat_alt.sh and optiBET.sh to reorient, bias correct and crop the T1w image.
And we generate a brain mask that can be used to limit the spread of the lesion we will draw.
Assume we are in the directory with the images we need.
COMMENTBLOCK

# If there are no arguments
if [ $# -lt 1 ] ; then
  echo "Usage: $0 <T1w_image>"
  echo "e.g. prep_T1w.sh sub-001_T1w.nii.gz"
  echo "This script prepares the T1w image for lesion mapping."
  echo "It calls fsl_anat_alt.sh and optiBET.sh to reorient, "
  echo "bias correct, and crop a T1w image and also to produce a "
  echo "quality brain mask that can be used to limit leakage of a lesion mask."
  echo "output:"
  echo "1) A subdirectory containing the results of running the processing steps"
  echo "2) brain mask, e.g., sub-001_T1w_brain_mask.nii.gz"
  echo "3) backup of original head: e.g., sub-001_T1w.nii.gz->sub-001_T1w_orig.nii.gz"

This file has been truncated. show original

bear_sun · July 2, 2020, 10:30pm

Thank you very much for you suggestion!

The original T1w volume doesn’t seem to have this brightness problem. So I guess it could be some side effect due to the incorrect brain extraction.

What I am doing now is to bypass the antsBrainExtraction.sh. I use FSL bet to do brain extraction first and then run fmriprep without brain extraction. It works well, though I still have no idea why the antsBrainExtraction.sh doesn’t work for my dataset.

oesteban · July 3, 2020, 1:51pm

I believe this is another instance of this issue: https://github.com/nipreps/smriprep/issues/125

We are currently working on this