# MSBraM: A Multi-scale Self-supervised Brain Foundation Model for Hierarchical EEG Dynamics Learning

arXiv: https://arxiv.org/abs/2607.21402 · HTML: https://mlsanity.com/paper/2607.21402/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 21 of its 54 references that are in the map.
- First posted on arXiv 23 Jul 2026.

## Builds on (most cited first, 21 of 21)

- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [Feature Pyramid Networks for Object Detection](https://mlsanity.com/paper/1612.03144/) (2016)
- [Deep High-Resolution Representation Learning for Visual Recognition](https://mlsanity.com/paper/1908.07919/) (2019)
- [EfficientDet: Scalable and Efficient Object Detection](https://mlsanity.com/paper/1911.09070/) (2019)
- [Path Aggregation Network for Instance Segmentation](https://mlsanity.com/paper/1803.01534/) (2018)
- [EEGNet: A Compact Convolutional Network for EEG-based Brain-Computer Interfaces](https://mlsanity.com/paper/1611.08024/) (2016)
- [Large Brain Model for Learning Generic Representations with Tremendous EEG Data in BCI](https://mlsanity.com/paper/2405.18765/) (2024)
- [NAS-FPN: Learning Scalable Feature Pyramid Architecture for Object Detection](https://mlsanity.com/paper/1904.07392/) (2019)
- [BENDR: using transformers and a contrastive self-supervised learning task to learn from massive amounts of EEG data](https://mlsanity.com/paper/2101.12037/) (2021)
- [HRFormer: High-Resolution Transformer for Dense Prediction](https://mlsanity.com/paper/2110.09408/) (2021)
- [Transformer-based Spatial-Temporal Feature Learning for EEG Decoding](https://mlsanity.com/paper/2106.11170/) (2021)
- [SeqSleepNet: End-to-End Hierarchical Recurrent Neural Network for Sequence-to-Sequence Automatic Sleep Staging](https://mlsanity.com/paper/1809.10932/) (2018)
- [BrainBERT: Self-supervised representation learning for intracranial recordings](https://mlsanity.com/paper/2302.14367/) (2023)
- [Transformer Convolutional Neural Networks for Automated Artifact Detection in Scalp EEG](https://mlsanity.com/paper/2208.02405/) (2022)
- [EEG2Rep: Enhancing Self-supervised EEG Representation Through Informative Masked Inputs](https://mlsanity.com/paper/2402.17772/) (2024)
- [A Multimodal Approach to Estimating Vigilance Using EEG and Forehead EOG](https://mlsanity.com/paper/1611.08492/) (2016)
- [Brant-X: A Unified Physiological Signal Alignment Framework](https://mlsanity.com/paper/2409.00122/) (2024)
- [Prediction of Reaction Time and Vigilance Variability from Spatiospectral Features of Resting-State EEG in a Long Sustained Attention Task](https://mlsanity.com/paper/1910.10076/) (2019)
- [U-HRNet: Delving into Improving Semantic Representation of High Resolution Network for Dense Prediction](https://mlsanity.com/paper/2210.07140/) (2022)
- [CEDNet: A Cascade Encoder-Decoder Network for Dense Prediction](https://mlsanity.com/paper/2302.06052/) (2023)
- [Tyee: A Unified, Modular, and Fully-Integrated Configurable Toolkit for Intelligent Physiological Health Care](https://mlsanity.com/paper/2512.22601/) (2025)

## Built on by (most cited first, 1 of 1)

- [NeurDuo-EEG: A Long-Sequence EEG Foundation Model with Persistent State and Explicit Memory](https://mlsanity.com/paper/2609.38587/) (2026)

## Abstract (arXiv)

Self-supervised foundation models have recently shown strong potential for electroencephalogram (EEG)-based analysis. However, existing approaches struggle to capture the inherently multi-scale temporal structure of EEG signals, where local neural patterns and long-range dependencies jointly encode task-relevant information. This limitation hampers cross-scale representation learning and generalization across diverse downstream tasks. To address this challenge, we propose MSBraM, a Multi-Scale self-supervised Brain foundation Model designed to learn hierarchical EEG representations. MSBraM follows a two-stage pretraining framework. First, a multi-scale neural tokenizer discretizes raw EEG signals into semantic codes at different temporal resolutions via vector-quantized reconstruction. Second, the model is pretrained to predict masked codes using a curriculum multi-scale masking strategy, progressively integrating fine-grained local patterns with global temporal context. We pretrain MSBraM on over 2,400 hours of EEG data and evaluate it across 10 downstream tasks on 12 public datasets. Extensive experiments show that MSBraM achieves superior performance on other state-of-the-art pretrained models, demonstrating strong generalization and transferability. These results indicate that explicitly modeling multi-scale temporal dynamics is critical for effective EEG foundation models.