Reproduce “ODEBrain: Continuous-Time EEG Graph for Modeling Dynamic Brain Networks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Haohui Jia, Zheng Chen, Lingwei Zhu, Rikuto Kotoge, Jathurshan Pradeepkumar, Yasuko Matsubara, Jimeng Sun, Yasushi Sakurai and 1 more
ODEBRAIN introduces a continuous-time EEG graph model that leverages Neural Ordinary Differential Equations (NODEs) to represent dynamic brain networks. This framework achieves enhanced seizure detection performance, with an AUROC of 0.881 on the TUSZ dataset, and offers interpretable insights into brain state transitions through its learned dynamic field.
Key insights
- Modeling EEG dynamics in continuous time with NODEs provides a more accurate and granular representation of neural activity, capable of capturing transient changes and distinct brain state transitions.
- A robust initial state, derived from both spectral graph features and stochastic temporal characteristics, is critical for stable continuous trajectory inference in dynamic EEG network modeling.
- The gradient field of the learned ODE serves as an interpretable metric, revealing distinct dynamic patterns for different brain states (e.g., convergence centers during seizures).
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.