PHINN-EEG uses topological time-series analysis to improve dream-state EEG classification

The authors introduce PHINN-EEG, the first topological time-series framework for dream mentation analysis, which outperforms existing power spectral density (PSD) and catch22 benchmarks. By extracting Dynamic Betti Curves from multichannel pre-awakening EEG epochs via sliding-window Takens delay embeddings and Vietoris-Rips filtrations, the method characterizes the geometric architecture of neural activity rather than just its energy.

• PHINN-EEG targets an AUC of 0.82-0.90 on the 1,462-awakening open-access subset of the DREAM database.
• The framework combines topological invariants with topology-conditioned flow matching for dream-state EEG synthesis.
• A spectral-conditioned flow model serves as an ablation baseline to isolate the value of topological conditioning.
• The work proposes candidate Betti transition archetypes linking topology to phenomenological dream report categories.

This approach represents a paradigm shift from spectral energy to phase-space geometry in neural rare-event detection, with potential future implications for wearable BCI dream monitoring.