A Roadmap for MEG Foundation Models
Abstract
Foundation models are beginning to reshape brain-signal analysis by moving the field beyond task-specific decoding pipelines toward reusable models pretrained on broad neural datasets. Magnetoencephalography (MEG) is a compelling but still underdeveloped target for this shift: it captures human cortical dynamics at millisecond resolution while offering stronger spatial interpretability than EEG, making it especially valuable for source-resolved studies of perception, language, cognition, and clinical brain function. Yet MEG foundation models remain at an early stage, with only a small number of MEG-specific and MEG-inclusive multi-modal models, modest pretraining corpora, and emerging but still limited benchmarks. This perspective lays down the basic concepts needed to understand MEG foundation models and provides a didactic overview of the field's key design choices, including tokenization, sensor- versus source-space representations, sensor-geometry encoding, backbone architectures, self-supervised objectives, and pretraining data. We then offer a roadmap for future development, organized around native MEG pretraining, adaptation of EEG foundation models, transfer from generic time-series models, and multi-modal integration with EEG, fMRI, MRI, behaviour, and stimulus features. We highlight the need for coordinated infrastructure, including diverse and reusable MEG datasets, rigorous evaluation across subjects, sites, tasks, and clinical settings, and responsible data-sharing practices that address consent, privacy, access, and governance.
Related
- An updated roadmap for brain-mental health associations in the UK Biobank
- A research roadmap for grazinglands: leveraging scientific evidence to drive beef sustainability outcomes.
- Innovation in the U.S. beef value chain: a sustainable beef research roadmap.
- The case for a sustainable beef research roadmap: building consensus on beef sustainability.
- Algorithmic Circuit Psychiatry: A Roadmap for Precision Treatment of Serious Mental Illness
BibTeX
@misc{tholke2026roadmap,
title = {{A Roadmap for MEG Foundation Models}},
author = {Philipp Thölke and Hamza Abdelhedi and Yorguin Mantilla-Ramos and Fouad Lbakali and Oumayma Gharbi and Catherine Duclos and Annalisa Pascarella and Vanessa Hadid and Oiwi Parker Jones and Karim Jerbi},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.04461},
eprint = {2609.04461},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.04461},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - A Roadmap for MEG Foundation Models
AU - Philipp Thölke
AU - Hamza Abdelhedi
AU - Yorguin Mantilla-Ramos
AU - Fouad Lbakali
AU - Oumayma Gharbi
AU - Catherine Duclos
AU - Annalisa Pascarella
AU - Vanessa Hadid
AU - Oiwi Parker Jones
AU - Karim Jerbi
PY - 2026
DA - 2026/09/21
PB - arXiv
DO - 10.48550/arxiv.2609.04461
UR - https://arxiv.org/abs/2609.04461
AB - Foundation models are beginning to reshape brain-signal analysis by moving the field beyond task-specific decoding pipelines toward reusable models pretrained on broad neural datasets. Magnetoencephalography (MEG) is a compelling but still underdeveloped target for this shift: it captures human cortical dynamics at millisecond resolution while offering stronger spatial interpretability than EEG, making it especially valuable for source-resolved studies of perception, language, cognition, and clinical brain function. Yet MEG foundation models remain at an early stage, with only a small number of MEG-specific and MEG-inclusive multi-modal models, modest pretraining corpora, and emerging but still limited benchmarks. This perspective lays down the basic concepts needed to understand MEG foundation models and provides a didactic overview of the field's key design choices, including tokenization, sensor- versus source-space representations, sensor-geometry encoding, backbone architectures, self-supervised objectives, and pretraining data. We then offer a roadmap for future development, organized around native MEG pretraining, adaptation of EEG foundation models, transfer from generic time-series models, and multi-modal integration with EEG, fMRI, MRI, behaviour, and stimulus features. We highlight the need for coordinated infrastructure, including diverse and reusable MEG datasets, rigorous evaluation across subjects, sites, tasks, and clinical settings, and responsible data-sharing practices that address consent, privacy, access, and governance.
ER -