PHINN-EEG: Topological Time-Series Analysis of Dream-State EEG -- Dynamic Betti Curves for Dream Content Classification and Topology-Conditioned Neural Signal Synthesis
arXiv (Cornell University) July 10, 2026 Ren Takahashi, Emre Yusuf, Jayabrata Bhaduri
A new framework called PHINN-EEG uses topological data analysis to detect dreaming from EEG signals, moving beyond traditional power-spectral methods. By extracting dynamic Betti curves from multichannel EEG epochs via sliding-window embeddings and Vietoris-Rips filtrations, the approach targets an area under the receiver operating characteristic curve (AUC) of 0.82–0.90 on a subset of the DREAM database, compared to the current state-of-the-art AUC of about 0.70. The work also introduces a topology-conditioned flow model for synthesizing dream-state EEG and proposes candidate Betti transition archetypes linking brain topology to dream report categories, pending empirical validation. If confirmed, this could shift neural rare-event detection from spectral energy to phase-space geometry, with implications for wearable brain-computer interface dream monitoring.