Enhanced Epileptic Seizure Detection Based on Information Fusion Techniques

Electroencephalogram (EEG) is the most effective tool for diagnosing a wide range of brain disorders. This work demonstrates how an EEG-based seizure detection system can differentiate between a normal, pre-ictal, and ictal state. From EEG data, several properties in the temporal, spectral, and temporal-spectral domains are first retrieved. Then, the wrapper technique with four classifiers is utilized for feature selection in order to select an optimal set of characteristics. This results in a new feature vector that has fewer elements. Then, in the classification step, Decision Tree (DT), Random Forest (RF), K Nearest Neighbor (KNN), Support Vector Machine (SVM) methods are used to classify chosen features from the EEG dataset as normal individuals, patients with epilepsy background, and those with epileptic seizures. Lastly, four classifiers are combined using Yager’s rule as a decision fusion approach. The proposed decision fusion framework achieves an accuracy of 99.2%, higher than previous studies on the same dataset and the individual classifiers utilized in the current work.