PROJECT TITLE :

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

ABSTRACT:

At the moment, one of the most cutting-edge approaches for seizure prediction is to perform graph theory analysis with electroencephalogram (EEG) signals. Recent deep learning approaches, which fail to fully explore both the characterizations in EEGs themselves and correlations among different electrodes simultaneously, generally neglect the spatial or temporal dependencies in an epileptic brain, and as a result, produce suboptimal seizure prediction performance as a consequence of this neglect. In this article, a patient-specific EEG seizure predictor is proposed by employing a novel spatio-temporal-spectral hierarchical graph convolutional network with an active preictal interval learning scheme. The purpose of this is to address the problem that has been brought up (STS-HGCN-AL). The proposed STS-HGCN-AL framework first infers a hierarchical graph to concurrently characterize an epileptic cortex under different rhythms. The temporal dependencies and spatial couplings of the epileptic cortex are then extracted by a spectral-temporal convolutional neural network and a variant self-gating mechanism, respectively. Specifically, because the epileptic activities in different brain regions may be of different frequencies, the proposed framework takes this into account. Utilizing a hierarchical graph convolutional network, critical intrarhythm spatiotemporal properties are then captured and integrated jointly before being mapped to the final recognition results. Our STS-HGCN-AL scheme estimates an optimal preictal interval patient dependently via a semisupervised active learning strategy, which further enhances the robustness of the proposed patient-specific EEG seizure predictor. In particular, because the preictal transition may be different from seconds to hours prior to the onset of a seizure in different patients, our STS-HGCN-AL scheme estimates an optimal preictal interval patient dependently. The proposed method has been shown to be effective in the extraction of critical preictal biomarkers, and competitive experimental results validate this efficacy, indicating that it may have promising applications in automatic seizure prediction.