Spatio-Temporal Attention Network for Epileptic Seizure Prediction

TL;DR

This paper introduces a deep learning spatio-temporal attention network that accurately predicts epileptic seizures from EEG signals, outperforming existing methods and enabling early detection for clinical intervention.

Contribution

The study proposes a novel Spatio-Temporal Attention Network with an adversarial discriminator for patient-specific seizure prediction, improving accuracy and early detection capabilities.

Findings

Achieved over 96% sensitivity on CHB-MIT dataset.

Detected preictal states at least 15 minutes before seizures.

Significantly outperformed existing seizure prediction methods.

Abstract

In this study, we present a deep learning framework that learns complex spatio-temporal correlation structures of EEG signals through a Spatio-Temporal Attention Network (STAN) for accurate predictions of onset of seizures for Epilepsy patients. Unlike existing methods, which rely on feature engineering and/or assume fixed preictal durations, our approach simultaneously models spatio-temporal correlations through STAN and employs an adversarial discriminator to distinguish preictal from interictal attention patterns, enabling patient-specific learning. Evaluation on CHB-MIT and MSSM datasets demonstrates 96.6\% sensitivity with 0.011/h false detection rate on CHB-MIT, and 94.2% sensitivity with 0.063/h FDR on MSSM, significantly outperforming state-of-the-art methods. The framework reliably detects preictal states at least 15 minutes before an onset, with patient-specific windows extending to 45 minutes, providing sufficient intervention time for clinical applications.