Seizure Prediction Based on Transformer Using Scalp Electroencephalogram

Abstract

Epilepsy is a chronic and recurrent brain dysfunction disease. An acute epileptic attack will interfere with a patient’s normal behavior and consciousness, having a great impact on their life. The purpose of this study was to design a seizure prediction model to improve the quality of patients’ lives and assist doctors in making diagnostic decisions. This paper presents a transformer-based seizure prediction model. Firstly, the time-frequency characteristics of electroencephalogram (EEG) signals were extracted by short-time Fourier transform (STFT). Secondly, a three transformer tower model was used to fuse and classify the features of the EEG signals. Finally, when combined with the attention mechanism of transformer networks, the EEG signal was processed as a whole, which solves the problem of length limitations in deep learning models. Experiments were conducted with a Children’s Hospital Boston and the Massachusetts Institute of Technology database to evaluate the performance of the model. The experimental results show that, compared with previous EEG classification models, our model can enhance the ability to use time, frequency, and channel information from EEG signals to improve the accuracy of seizure prediction.

Graph-generative neural network for EEG-based epileptic seizure detection via discovery of dynamic brain functional connectivity

“…For each training and testing round, we randomly shuffled 2033 cases for model training and the remaining 1014 cases for testing. We have compared the GGN testing results with other baseline ML/DL detection methods, including SVM 59 , 60 , CNNs based on Asif, et al 38 , GNNs based on Tang, et al 50 , and the transformer model by Yan, et al 53 .…”

“…Compared with the CNN-based model 38 and the transformer-based model 53 in our experiments, our GGN model uses much shallower hidden layers. This implies that we use fewer learnable parameters and thus has lower computational complexity.…”

“…Clinically recorded seizure dataset. 59,60 , CNNs based on Asif, et al 38 , GNNs based on Tang, et al 50 , and the transformer model by Yan, et al 53 .…”

“…These GNN models are based on fixed neural connectivity topologies such as distance-based, which failed to capture the spatial dynamics associated with seizure attacks. The latest works utilizing transformer models 53 have achieved some success in detecting connectivity thanks to the attention mechanism. However, it is still limited to discover new connectivity at different times.…”

A multi-frame network model for predicting seizure based on sEEG and iEEG data

“…To the influence of different kinds of subjects on the experimental results, we only selected canine subjects as the research objects. In addition, like most recent studies (Cheng et al, 2021 ; Gao et al, 2022 ; Yan et al, 2022 ), a period of 30 min before each seizure was defined as the prei-ctal period.…”