Improving the Accuracy of Epileptic Seizure Detection through EEG Analysis: A Comprehensive Classification Strategy

Abstract

Epilepsy is a neurological disorder which impacts millions globally and continues to be a major public health challenge. The prompt identification of epileptic seizures is essential for effective treatment. In this study, we present an innovative methodology designed to enhance the accuracy of seizure detection through EEG data analysis. Our strategy involves creating a comprehensive EEG database that includes both healthy individuals and those experiencing seizures (ictal). We utilize a diverse range of classification models, including random forests, decision trees, XGBoost and k-nearest neighbors algorithm. For feature extraction, we have selected Linear Discriminant Analysis (LDA) as our preferred technique. The experimental results indicate that the random forest model is the most effective, achieving a perfect accuracy rate of 100% in detecting epileptic seizures. The decision tree model follows closely with an accuracy of 90.00%. Although the kNN algorithm has a slightly lower accuracy of 82.50%, it still plays a significant role in differentiating between normal and ictal EEG signals. Our results clearly demonstrate the effectiveness of our proposed method in reliably extracting spatial and temporal information from multi-channel EEG data, enabling accurate classification of epileptic seizures. This research highlights the robustness of our feature extraction approach and its potential to improve early diagnosis and treatment of epilepsy.