A multi-frame network model based on time-frequency analysis for epilepsy prediction

• Published in: Chinese Control Conference pp. 8262 - 8267
• Main Authors: Zhang, Feng, Wei, Xile, Lu, Liangfu
• Format: Conference Proceeding
• Language: English
• Published: Technical Committee on Control Theory, Chinese Association of Automation 28.07.2024
• Subjects: Analytical models; Brain modeling; deep learning; EEG; Epilepsy; epilepsy prediction; feature fusion; Fourier transforms; Predictive models; seizure; Sensitivity; Time-frequency analysis
• ISSN: 1934-1768
• DOI: 10.23919/CCC63176.2024.10662334

In this paper, a multi-frame network model based on time-frequency analysis is proposed for epilepsy prediction, which is used to electroencephalogram (EEG) data classification. Firstly, short-term Fourier transform and wavelet packet transform is used to extract time-frequency information respectively, which can obtain two data representations. Secondly, convolutional neural network and long and short-term memory network are constructed to extract high-level features from above two data representations. An innovative module named MFB++ is then used for feature fusion. Finally, the fused feature is input into classifier to obtain classification result. The experimental results show that the proposed model achieves sensitivity of 98.25% and 94.79%, and false prediction alarm rate of 0.07/h and 0.00/h on CHB-MIT and TJU-HH datasets, respectively. Moreover, the significance test shows that our model performs significantly better than unspecific random predictor for all subjects. The proposed model achieves expected prediction performance, which has important practical significance for improving the quality of life of epilepsy patients.