An Effective Dual Self-Attention Residual Network for Seizure Prediction

As one of the most challenging data analysis tasks in chronic brain diseases, epileptic seizure prediction has attracted extensive attention from many researchers. Seizure prediction, can greatly improve patients' quality life ways, such as preventing accidents and reducing harm that may occur during seizures. This work aims to develop a general method for predicting seizures specific patients through exploring time-frequency correlation features obtained multi-channel EEG signals. We convert original signals into spectrograms represent characteristics by applying short-time Fourier transform (STFT) For first time, we propose dual self-attention residual network (RDANet) combines spectrum module integrating local with global features, channel mining interdependence between mappings achieve better forecasting performance. Our proposed approach achieved sensitivity 89.33%, specificity 93.02%, an AUC 91.26% accuracy 92.07% on 13 public CHB-MIT scalp dataset. experiments show different signal segment lengths are important factor affecting is competitive achieves good robustness without patient-specific engineering.