Overview

This PR adds native FLAIR support to MRIQC, allowing _FLAIR BIDS-named files to be processed directly. FLAIR images are routed through the existing anatomical QC workflow and produce the same reports and IQMs as T1w and T2w.

Related Issue

• Using 3DFLAIR? #871

Requirements

These instructions assume Docker is already installed on your system. See https://docs.docker.com/get-docker/ for installation guidance.

Building the Docker image

After cloning this branch, build the development image from the repo root:

docker build --no-cache -f Dockerfile_devel -t mriqc_devel --build-arg VERSION=24.0.2 .

Usage

mriqc /data /out participant -m FLAIR T1w

No additional environment variables are required.

Files Changed

Key Implementation Details

N4 bias correction fallback

FLAIR images frequently cause N4 to converge to a near-empty output due to the inverted CSF contrast. When modality == 'FLAIR' and the N4-corrected image is effectively empty, the pipeline substitutes the original uncorrected image and logs
a warning. For all other modalities (T1w, T2w) the original RuntimeError is raised as before. This fallback is strictly FLAIR-only.

Spatial normalization

ANTs RobustMNINormalization does not accept 'FLAIR' as a valid reference value. FLAIR is mapped to 'T2w' for the purposes of normalization only, via _get_norm_mod. The modality label remains 'FLAIR' everywhere else in the pipeline (datasink, provenance, IQM output).

Testing

The following steps reproduce a minimal end-to-end test using a single subject from
OpenNeuro dataset [ds007486](https://openneuro.org/datasets/ds007486), which contains FLAIR acquisitions.

1. Download the dataset

If you have DataLad installed:

datalad clone https://github.com/OpenNeuroDatasets/ds007486.git /path/to/ds007486
cd /path/to/ds007486
datalad get sub-10562/anat/

If you do not have DataLad, use the AWS CLI instead (no credentials required):

# Install if needed
pip install awscli

# Create the destination directory
mkdir -p /path/to/ds007486/sub-10562/anat

# Download the subject's anatomical data
aws s3 sync \
  --no-sign-request \
  s3://openneuro.org/ds007486/sub-10562/anat/ \
  /path/to/ds007486/sub-10562/anat/

2. Create output directory

mkdir -p /path/to/ds007486/derivatives/mriqc/logs

3. Run MRIQC

docker run --rm \
  -v /path/to/ds007486:/data \
  -v /path/to/ds007486/derivatives/mriqc:/out \
  -v /tmp/scratch:/scratch \
  -e TEMPLATEFLOW_HOME=/tmp/templateflow \
  mriqc_devel \
  /data /out participant \
  --participant_label 10562 \
  --no-datalad-get \
  --no-sub \
  -m FLAIR \
  --bids-database-wipe \
  -w /scratch

Replace /path/to/ds007486 with the actual path on your system.

4. Expected output

On success you should see a FLAIR IQM JSON file and visual report at:

/path/to/ds007486/derivatives/mriqc/sub-10562/anat/sub-10562_FLAIR.json
/path/to/ds007486/derivatives/mriqc/sub-10562/anat/sub-10562_FLAIR.html