Topological Entropy and Homology Reveal Interpretable and Real-Time Neural Signatures in Pediatric EEG

ObjectiveClassifying cognitive state from pediatric mobile electroencephalography (EEG) recorded during naturalistic, high-movement behavior is challenging due to motion artifacts, low channel count, and severe imbalance between rest and task conditions. Rather than optimizing accuracy under favorable or imbalance-agnostic evaluation settings, this study proposes an imbalance-aware and computationally feasible framework to distinguish Minecraft gameplay from eyes-open rest using low-density pediatric EEG. MethodsWe introduce Enriched Topological Features, which combine Takens phase-space embeddings with persistent homology, persistence landscapes, and persistence entropy. Evaluation was performed on four-channel Muse EEG from 43 children using subject-disjoint five-fold GroupKFold cross-validation. Class imbalance was handled exclusively within training folds using synthetic minority oversampling or within-subject random undersampling, while test folds preserved natural class proportions. Robustness to persistence landscape resolution and truncation depth was assessed using linear mixed-effects modeling. ResultsUnder oversampling-based training, the proposed framework achieved test balanced accuracy up to 65.39% and macro-averaged F1 score up to 64.48%, with minority-class rest recall reaching 49.51% using temporal electrodes. Feature attribution analysis associated gameplay predictions with frontal gamma-band topological structure, whereas rest predictions were associated with alpha-band organization at temporal sites. End-to-end processing required 0.676 s per 4.5 s epoch (0.15x real time), supporting near-real-time feasibility. ConclusionThe proposed approach enables imbalance-aware gameplay-rest classification from low-density pediatric mobile EEG and is compatible with near-real-time analysis. SignificanceThis work demonstrates that topological representations support robust cognitive-state monitoring in pediatric mobile neuroengineering under realistic motion and hardware constraints.