neuralset

A Python toolkit for building neuroimaging data pipelines — from raw data to AI-ready batched tensors.

• Event-driven processing with typed, validated DataFrames
• Multi-modal extractors for MEG, EEG, fMRI, EMG, iEEG, text, image, audio, video
• Caching and remote compute via exca

Install

pip install neuralset

Quick example

# First install study dependencies:
# pip install neuralfetch
import neuralset as ns

# Load a study and download its data (MNE sample, ~1.6GB on first run).
study = ns.Study(name="Mne2013Sample", path=ns.CACHE_FOLDER)
study.download()
events = study.run()

# Configure extractors
meg = ns.extractors.MegExtractor(frequency=100.0, filter=(0.5, 30))

# Segment around triggers and build a dataset
segmenter = ns.Segmenter(
    extractors=dict(meg=meg),
    trigger_query='type == "Stimulus"',
    start=-0.1, duration=0.5,
)
dataset = segmenter.apply(events)
batch = dataset.load_all()

Citation

If you use NeuralSet in your research, please cite NeuralSet: A High-Performing Python Package for Neuro-AI:

@article{king2026neuralset,
  title   = {NeuralSet: A High-Performing Python Package for Neuro-AI},
  author  = {King, J-R. and Bel, C. and Evanson, L. and Gadonneix, J. and Houhamdi, S. and L{\'e}vy, J. and Raugel, J. and Santos Revilla, A. and Zhang, M. and Bonnaire, J. and Caucheteux, C. and D{\'e}fossez, A. and Desbordes, T. and Diego-Sim{\'o}n, P. and Khanna, S. and Millet, J. and Orhan, P. and Panchavati, S. and Ratouchniak, A. and Thual, A. and Brooks, T. and Begany, K. and Benchetrit, Y. and Careil, M. and Banville, H. and d'Ascoli, S. and Dahan, S. and Rapin, J.},
  year    = {2026},
  url     = {https://kingjr.github.io/files/neuralset.pdf},
  note    = {Preprint; URL will be updated when the paper lands on arXiv}
}

Third-Party Content

References to third-party content from other locations are subject to their own licenses and you may have other legal obligations or restrictions that govern your use of that content.

License

This project is licensed under the MIT License. See LICENSE for details.