In the Search for Truth: Navigating Variability in Neuroimaging Software Pipelines

Youenn Merel-Jourdan, Mathieu Acher, Camille Maumet

Submitted to SPLC'25, September 01-September 05, 2025 Coruña, Spain

Data

Data used in the paper is available in the data directory

UVL feature model

The UVL feature model expressing the explored analytical space is at data/model/full_pipeline.uvl

fMRI data

Task-fMRI dataset is available in data/auditory. It was downloaded from https://www.fil.ion.ucl.ac.uk/spm/data/auditory/

Configurations

The configurations used in the experiment are available in the data/configs

The 1000 randomly sampled configuration are divided between 20 CSV files (50 rows + header) named config_[n].csv

The expert-crafted configuration is config_ref.csv.

Data descriptor

data/auditory/data_desc_auditory.json is a JSON file describing the experiments data and some execution paths.

Sampling

The code used for this part is available at https://github.com/Inria-Empenn/fmri_feature_model

Pull & install project

git clone https://github.com/Inria-Empenn/fmri_feature_model.git
cd fmri_feature_model
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Sample

Randomly sample 1000 configurations divided into 20 files (+ reference configuration)

python sample.py --nconfig 1000 --parts 20

Pipelines execution & postprocessing

The code used for this part is available at https://github.com/Inria-Empenn/fmri-conf-runner

Pull & install project

git clone https://github.com/Inria-Empenn/fmri-conf-runner.git
cd fmri-conf-runner
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Build docker image

docker build . -t fmri-conf-runner

Final fmri-conf-runner image size is approximately 6.5 GB

Alternatively, you can directly pull latest image from GitHub : docker pull ghcr.io/inria-empenn/fmri-confs-runner:latest

Pipelines execution

Change /local/path/to/... to your local paths

• /local/path/to/data : Will be mapped to /data in the container. This folder must contains
  • the auditory dataset/subfolder
  • data_desc.json file
• /local/path/to/results : This folder must exists. Will be mapped to /results in the container.
• /local/path/to/workdir : This folder must exists. Will be mapped to /workdir in the container.
• /local/path/to/configs : This folder must contains configuration CSV files (in this example config.csv and config_ref.csv). Will be mapped to /configs in the container.

docker run -u root -v "/local/path/to/data:/data" -v "/local/path/to/results:/results" -v "/local/path/to/workdir:/work" -v "/local/path/to/configs:/configs" fmri-conf-runner python -u run.py --configs "/configs/config.csv" --data /data/data_desc.json --ref /configs/config_ref.csv

On Abaca (Inria cluster), use run_configs.sh

oarsub -S -n fmri-conf-runner ./run_configs.sh

Postprocessing

Change /local/path/to/... to your local paths

• /local/path/to/results : This folder must contains the outputs of the pipeline execution step. Will be mapped to /results in the container.

docker run -u root -v "/local/path/to/results:/results" fmri-conf-runner python -u postprocess.py --results "/results"

On Abaca (Inria cluster), use postprocess.sh

oarsub -S -n postprocess ./postprocess.sh

Analysis

Analysis code and results can be found in analysis

analysis/auditory_analysis.ipynb is the Jupyter notebook used to process and analyse the data.

analysis/classifier contains the output from classifier decision tree training (dot and pdf files)

analysis/regression contains the output from regression decision tree training (dot and pdf files)