Spatio-Temporal-Spectral Hierarchical Graph Convolutional Network With Semisupervised Active Learning for Patient-Specific Seizure Prediction

Abstract:

Seizure prediction using graph theory and EEG signals is advanced. Recent deep learning approaches, which fail to fully explore EEG characterizations and electrode correlations simultaneously, neglect the spatial or temporal dependencies in an epileptic brain and produce suboptimal seizure prediction performance. This article proposes a novel spatio-temporal-spectral hierarchical graph convolutional network with an active preictal interval learning scheme (STS-HGCN-AL) to predict patient-specific EEG seizures. Since different brain regions may have different epileptic frequencies, the STS-HGCN-AL framework first infers a hierarchical graph to concurrently characterize an epileptic cortex under different rhythms, whose temporal dependencies and spatial couplings are extracted by a spectral-temporal convolutional neural network and a variant self-gating mechanism, respectively. A hierarchical graph convolutional network captures and integrates critical intrarhythm spatiotemporal properties and maps them to recognition results. Since the preictal transition may vary from seconds to hours before seizure onset, our STS-HGCN-AL scheme estimates an optimal preictal interval patient-dependently using a semisupervised active learning strategy, which strengthens the proposed patient-specific EEG seizure predictor. Competitive experimental results show that the proposed method can extract critical preictal biomarkers, indicating its potential for automatic seizure prediction.

Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.