loading page

Triple-Attention-based Spatio-Temporal-Spectral Convolutional Network for Epileptic Seizure Prediction
  • lianghui guo ,
  • Tao Yu
lianghui guo
Beihang University, Beihang University

Corresponding Author:[email protected]

Author Profile

Abstract

Seizure prediction of epileptic preictal period through electroencephalogram (EEG) signals is important for clinical epilepsy diagnosis and closed-loop treatment. Recent deep learning-based methods made great efforts on seizure prediction. However, multi-domain characterizations, including spatio-temporal-spectral dependencies in an epileptic brain are generally neglected or not considered simultaneously in current approaches, and this insufficiency commonly leads to suboptimal seizure prediction performance. Besides, the lack of efficient feature fusion leads to redundant information and fails to achieve the most discriminative features. To tackle the above issues, in this paper, an end-to-end patient-specific seizure predicting method is proposed by using a novel triple-attention-based spatio-temporal-spectral convolutional network (TA-STS-ConvNet). Specifically, the TA-STS-ConvNet firstly applies pyramid convolution net to extract multi-scale temporal and spectral representations under different rhythms from raw EEG signals. Then, a novel triple attention mechanism is employed to construct inter-dimensional interaction among multi-domain features and fused them into comprehensive feature maps. Moreover, we propose a spatio dynamic graph convolution network (sdGCN) to dynamically model the spatial relationships between electrodes and aggregate spatial information. The proposed TA-STS-ConvNet achieved a sensitivity of 96.7% and a false prediction rate of 0.072/h on the CHB-MIT scalp EEG database. We also validate the proposed method on clinical intracranial EEG (iEEG) database from the Xuanwu Hospital of Capital Medical University, and the predicting system yielded a sensitivity of 95%, a false prediction rate of 0.087/h. The experimental results outperform the state-of-the-art seizure prediction methods, which validate the efficacy of our proposed method. Our code is available at https://github.com/LianghuiGuo/TA-STS-ConvNet.
2023Published in IEEE Transactions on Neural Systems and Rehabilitation Engineering volume 31 on pages 3915-3926. 10.1109/TNSRE.2023.3322275