Early Prediction of Epileptic Seizure Based on the BNLSTM-CASA Model

Epilepsy is one of the world's most common neurological diseases. Reliable early prediction and warning seizures can provide timely treatment for patients with epilepsy, improve their quality life. Compared hand-designed methods, an automatic model that process original electroencephalogram (EEG) signals directly take into account leads optimization problem needed. In this paper, we proposed end-to-end seizure based on Batch Normalization Long Short Term Memory networks (BNLSTM) Channel Spatial attention (CASA). Firstly, raw EEG without any preprocessing are used as input to system, which reduce computation amount. Secondly, BNLSTM CASA retained time spatial information data respectively. (CA) achieved full-lead improved accuracy. (SA) adaptive learning feature parameters. Finally, a fully connected layer applied predict seizures. The performance evaluated 14 Area Under Curve (AUC) 0.986, accuracy (Acc) 0.956, specificity (Spe) 0.968, sensitivity (Sen) 0.942. addition, method provided accurate all 50 other 5 in generalization dataset. Experimental results show has certain performance, reliable basis epileptic