LERD: Latent Event-Relational Dynamics for Neurodegenerative Classification

Alzheimer’s disease (AD) alters brain electrophysiology and disrupts multichannel EEG dynamics, making accurate and clinically useful EEG-based diagnosis increasingly important for screening and disease monitoring. However, many existing approaches rely on black-box classifiers and do not explicitly model the latent event timing and cross-channel coordination behind their decisions. To address these limitations, we propose LERD, an end-to-end Bayesian latent event--relational dynamical system that infers latent neural events and their relational structure directly from multichannel EEG without event or interaction annotations. LERD combines a continuous-time event inference module with a stochastic event-generation process to capture flexible temporal patterns, while incorporating an electrophysiology-inspired dynamical prior to guide learning in a principled way. We further provide theoretical analysis that yields a tractable IVP-based KL regularizer and stability guarantees for the inferred relational dynamics. Extensive experiments on synthetic benchmarks and two real-world AD EEG cohorts demonstrate that LERD consistently outperforms strong baselines and yields physiology-aligned rate, timing, and graph summaries that help characterize group-level dynamical differences.

Alzheimer’s disease can change the brain’s electrical rhythms and the way different brain regions coordinate with each other. EEG, or electroencephalography, records these rhythms using scalp sensors and is relatively low-cost and non-invasive, but many AI systems for EEG-based diagnosis only output a label, making their decisions hard to interpret. We developed LERD, a machine-learning model that looks for hidden “events” in EEG signals: moments when brain activity changes in meaningful ways. LERD learns when these events happen and how events across different EEG channels line up over time, without requiring experts to mark the events by hand. The model is also guided by simple principles about plausible brain activity, such as timing constraints and rhythmic behavior. In tests on synthetic data and two Alzheimer’s-related EEG datasets, LERD improved classification performance compared with existing methods and produced interpretable summaries of timing, rhythm, and channel interactions. These summaries reflected known disease patterns, including slower brain rhythms in Alzheimer’s disease. This work could help researchers build more transparent tools for studying neurodegenerative disease from EEG, while supporting—not replacing—clinical judgment.