🤖 AI Summary
This study addresses the challenges of high inter-patient variability and complex temporal dynamics in electroencephalogram (EEG) signals for seizure prediction by proposing an adaptive Transformer framework tailored to individual patients. The approach employs a two-stage training strategy: first, a general-purpose EEG representation is learned via self-supervised pretraining; subsequently, patient-specific fine-tuning is performed by integrating noise-aware preprocessing, multi-channel signal tokenization, and autoregressive modeling. Evaluated on the TUH EEG dataset, the method achieves over 90% accuracy in predicting seizures within a 30-second horizon and attains an F1 score above 0.80, significantly outperforming existing approaches and demonstrating enhanced performance in personalized seizure forecasting.
📝 Abstract
Epileptic seizure prediction from electroencephalographic (EEG) recordings remains challenging due to strong inter-patient variability and the complex temporal structure of neural signals. This paper presents a patient-adaptive transformer framework for short-horizon seizure forecasting. The proposed approach employs a two-stage training strategy: self-supervised pretraining is first used to learn general EEG temporal representations through autoregressive sequence modeling, followed by patient-specific fine-tuning for binary prediction of seizure onset within a 30-second horizon. To enable transformer-based sequence learning, multichannel EEG signals are processed using noise-aware preprocessing and discretized into tokenized temporal sequences. Experiments conducted on subjects from the TUH EEG dataset demonstrate that the proposed method achieves validation accuracies above 90% and F1 scores exceeding 0.80 across evaluated patients, supporting the effectiveness of combining self-supervised representation learning with patient-specific adaptation for individualized seizure prediction.