Exploring brain network cycling in real time

Brain activity continually shifts between large-scale cortical networks. Although individual transitions are variable, these networks tend to activate in a preferred cyclical order. Previous research has shown that this cyclical organisation emerges across independent datasets, with cycling characteristics linked to age, cognition, and behaviour [1, 2]. Visualising this organisation as it unfolds offers an intuitive way to explore the temporal structure of brain activity and how it varies across individuals and recording conditions.

We present an interactive framework for estimating and visualising network cycling using streaming electroencephalography (EEG), without source reconstruction during online inference. Building on a hidden Markov model representation of cortical network activity, the framework estimates brain-state probabilities and decodes cycling rate as EEG samples arrive. Visualising this organisation as it unfolds offers an intuitive way to explore the temporal structure of brain activity and how it varies across individuals and recording conditions.

Demo Experience The demo uses prerecorded resting-state EEG from the publicly available LEMON dataset, streamed to simulate incoming data. Attendees can select recordings, navigate between eyes-open and eyes-closed periods, and observe the model’s evolving estimates. A circular display places brain states according to their preferred cyclical ordering, while accompanying time courses show state probabilities and estimated cycling rate. This presentation makes both the overall organisation and the variability of individual transitions visible: network activity can revisit or skip states rather than progressing through a rigid sequence. Attendees will explore how cycling estimates change across recording conditions and individuals, and how these changes relate to the underlying state activity. The demonstration also provides an opportunity to discuss what is preserved when transferring a cortical network representation to scalp EEG, and the distinction between recovering individual state identities and recovering broader temporal dynamics.

The framework offers a concrete example of translating learned representations of neural activity into interpretable outputs that can be updated during streaming inference. It provides a prototype for future investigations of network dynamics in neurofeedback and adaptive experimental paradigms. The demonstration runs on a laptop and requires table space and a power outlet; an external monitor would improve shared viewing. All recordings are stored locally, and no EEG acquisition equipment is required.

[1] van Es, M. W. J., Higgins, C., Gohil, C., Quinn, A. J., Vidaurre, D., & Woolrich, M. W. (2025). Large-scale cortical functional networks are organized in structured cycles. Nature neuroscience, 28(10), 2118–2128. https://doi.org/10.1038/s41593-025-02052-8 [2] Forster, C., Gohil, C., Burgher, B., Kuzovkin, I., van Es, M. W. J., Woolrich, M., Vidaurre, D., van den Heuvel, M. P., Higgins, C., & Cocchi, L. (2026). Anterior default mode brain state dynamics predict depressive symptom severity before and during TMS treatment [Preprint]. bioRxiv. https://doi.org/10.64898/2026.05.27.728312