Optimizing EEG Artifact Removal Using Fuzzy Logic and Elliptical Filtering for Accurate Seizure Detection

Abstract

Epilepsy diagnosis relies heavily on electroencephalography (EEG) analysis, but artifacts such as muscle activity, eye blinks, and environmental noise degrade the accuracy of seizure detection. This study introduces a fuzzy elliptical filter (FEF), an adaptive EEG denoising framework that integrates fuzzy logic with elliptical filtering to preserve clinically significant epileptic signatures. The system dynamically adjusts filter parameters using energy, entropy, and variance features through a Mamdani-type inference engine with a complete fuzzy rule base. After denoising, the EEG signals are classified using a support vector machine (SVM) to enhance seizure detection performance. Experiments conducted on the Kaggle epileptic EEG dataset show a 33.6% improvement in signal-to-noise ratio (SNR), a 33.3% reduction in mean squared error (MSE), and a 10% increase in SVM-based classification accuracy, along with a 28% reduction in computational time. The proposed method demonstrates superior artifact suppression while preserving epileptic spikes. These results indicate that the FEF–SVM pipeline is a promising signal-processing approach with strong potential for real-time electrotherapeutic systems.