Strange Linear Artifacts on fMRI Scan Images

[ejdavid]

Hi all,

I have run my scans through preprocessing using fMRIPrep but I am seeing some strange linear artifacts on them (see images below). I’m wondering if anyone has any idea what might be causing these artifacts? Is it something to do with the scan sequence or is it something that came up during the preprocessing? Any information anyone might have would be incredibly useful. Thanks again!

This is the code I used in fMRIPrep if that helps:

!/bin/bash

# CSH: created Sept2022

# set subject list and loop
for subj in `cat sub.txt`; do
echo "Starting subject ${subj}"

# set paths
bids_root_dir=/lbc/lbc1/PREVENT_AD/BIDS_ya
bids_deriv_dir=/lbc/lbc1/PREVENT_AD/derivatives/rsFC_ya/fmriprep

# set subject specific paths
mkdir -p /lbc/lbc1/PREVENT_AD/derivatives/rsFC_ya/work/sub-${subj}_wd
workdir=/lbc/lbc1/PREVENT_AD/derivatives/rsFC_ya/work/sub-${subj}_wd

# begin fmriprep:

export TEMPLATEFLOW_HOME=/home/bic/collhugh/.cache/templateflow
export FS_LICENSE=/lbc/lbc1/PREVENT_AD/derivatives/rsFC_ya/license.txt
export FILTER_FILE=/lbc/lbc1/PREVENT_AD/derivatives/rsFC_ya/bids_filter_file.json

docker run --rm --user $(id -u):$(id -g) \
-v $bids_root_dir:/data \
-v $bids_deriv_dir:/out \
-v $FS_LICENSE:/license.txt \
-v $FILTER_FILE:/bids_filter_file.json \
-v $workdir:/work \
nipreps/fmriprep:23.2.1 /data /out participant \
--participant-label $subj \
--skip-bids-validation \
--fs-license-file /license.txt \
--bids-filter-file /bids_filter_file.json \
--md-only-boilerplate \
--output-spaces func T1w MNI152NLin6Asym \
--n_cpus 5 \
--skull-strip-t1w force \
--me-output-echos \
--low-mem \
-w /work

done

rm -r $workdir

[jsein]

Hi @ejdavid ,

I think that what you are seeing may be just a consequence of your display settings in FSLeyes. You could use interpolation in the “general settings” to smooth your views and attenuate this effect. You could also change the display space of your data in the “scene settings” to either your T1w space, your bold space, or world coordinates. Tell us if that changes your views in an expected manner.

[neurolabusc]

@jsein is correct. These are interpolation artifacts observed when sampling voxel-based data off of its grid. With FSLeyes, you have two solutions

1. press the Wrench icon and set the display space to the image space rather than the world coordinates.
2. Press the Gear icons and choose a higher-order interpolation from the drop down menu. The No interpolation uses nearest neighbor which exhibits severe interpolation artifacts, the Linear interpolation reduces these at the cost of high frequencies (it can look blurry), the Spline interpolation does a great job at fighting interpolation artifacts while preserving higher frequencies but it can create ringing artifacts.

You may also want to try NiiVue. While it also allows you to choose between voxel and world space, the world space does not constrain the depth direction so it can select a coherent slice from your volume. This allows you to add functions like drawing regions of interest (where FSLeyes forces editing to be done in voxel space).