查看文章

Seizures are caused by abnormal and excessive electrical activity in the neurons (nerve cells) of the brain. Every year, roughly 125,000 fresh cases of epilepsy are found, with 30% of these cases occurring in those under the age of 18. Seizure prediction is an important subject that can enhance the lives of epilepsy sufferers, and it has gotten a significant amount of attention in recent years. In this venture, we suggested an approach employing deep-learning strategies that evaluate the outcomes of CNN and RNN models. Using EEG data, Dense Net and LSTM are used for identifying seizures caused by epilepsy. The datasets utilized in the proposal were obtained from UCI and CHB-MIT. The suggested method entails extracting relevant information for categorization. To comprehend an EEG signal, the built-in libraries viz. pyeeg and pyedflib are applied. The characteristics are then incorporated into classification algorithms, which determine the occurrence or dearth of seizure activity within the signal. On the UCI dataset the efficiency for the dense net model reached 96% while LSTM showed about 98%. On the CHB-MIT dataset, the metric-accuracy obtained for Dense Net was 86% and LSTM was 92% respectively.