Abstract
Temporary disturbances in brain function are caused by epilepsy, a chronic disorder resulting from sudden abnormal firing of brain neurons. This research introduces an innovative real-time methodology representing detecting epileptic spasms from electroencephalogram (EEG) data. It employs a support vector machine (SVM) alongside embedded zero tree wavelet (EZW) transform. To facilitate precise multiresolution analysis of epileptic convulsions, the EZW method is selected for its capacity to efficiently compress multichannel EEG data while preserving crucial diagnostic features. EZW effectively captures and encodes key patterns in EEG signals, enabling detailed analysis of the subtle variations associated with seizures. This study extracts statistical features such as entropy, kurtosis, skewness, and mean from the compressed EEG segments. These features are then classified using the SVM to distinguish between normal and epileptic states. With a remarkable 99.02% classification accuracy and a false positive rate of only 1.1%, the proposed algorithm demonstrates excellent performance. The novelty lies in integrating SVM with EZW-based feature extraction and advanced preprocessing, enabling efficient real-time EEG analysis. Unlike previous works, this approach preserves critical information, enhances classification accuracy, and supports multichannel signals, offering a robust and practical solution for real-time epilepsy detection. Based on these findings, the method is considered highly suitable for real-time implementation in clinical environments.