Objective To explore an epileptic seizure prediction method for patients with refractory epilepsy to improve the classification and prediction efficiency of epileptic electroencephalogram(EEG)signals.Methods The study used the long-term EEG database of patients with intractable epilepsy from Children's Hospital Boston(CHB-MIT).The EEG features of epileptic seizures and preictal periods were extracted from multiple dimensions such as EEG synchronization,complexity,and energy distribution,and then these features were input into the artificial neural network model for classification and identification,thereby achieving accurate prediction of epilepsy.The performance were optimized by adjusting the model parameters,and a comparative evaluation was conducted with existing deep learning models.Results The model proposed in this study showed an accuracy rate of 99.29％,a precision of 91.44％,a sensitivity of 96.46％,and a specificity of 99.46％.Compared with current epilepsy seizure prediction studies based on machine learning or deep learning frameworks,the model in this study improved its classification prediction capabilities and demonstrated higher prediction accuracy.Conclusion An effective prediction of epileptic seizures was achieved by manually extracting epileptic EEG features and constructing an artificial neural network model.The model demonstrated high accuracy and stability,providing reliable technique to support clinical treatment and prevention of epilepsy.

Objective To explore an epileptic seizure prediction method for patients with refractory epilepsy to improve the classification and prediction efficiency of epileptic electroencephalogram(EEG)signals.Methods The study used the long-term EEG database of patients with intractable epilepsy from Children's Hospital Boston(CHB-MIT).The EEG features of epileptic seizures and preictal periods were extracted from multiple dimensions such as EEG synchronization,complexity,and energy distribution,and then these features were input into the artificial neural network model for classification and identification,thereby achieving accurate prediction of epilepsy.The performance were optimized by adjusting the model parameters,and a comparative evaluation was conducted with existing deep learning models.Results The model proposed in this study showed an accuracy rate of 99.29％,a precision of 91.44％,a sensitivity of 96.46％,and a specificity of 99.46％.Compared with current epilepsy seizure prediction studies based on machine learning or deep learning frameworks,the model in this study improved its classification prediction capabilities and demonstrated higher prediction accuracy.Conclusion An effective prediction of epileptic seizures was achieved by manually extracting epileptic EEG features and constructing an artificial neural network model.The model demonstrated high accuracy and stability,providing reliable technique to support clinical treatment and prevention of epilepsy.