Epileptic Seizure Classification of EEGs Using Time-Frequency Analysis Based Multiscale Radial Basis Functions

Area of Science:

• * Neuroscience
• * Biomedical Engineering
• * Signal Processing

Background:

• * Epileptic seizure detection from electroencephalography (EEG) is vital for diagnosis and treatment.
• * Dynamic and nonstationary nature of seizure activity presents challenges for accurate signal analysis.
• * Distinguishing rhythmic epileptic discharges from nonstationary background EEG is a key difficulty.

Purpose of the Study:

• * To propose a novel framework for adaptive and localized time-frequency representation in EEG signals.
• * To enhance the simultaneous time and frequency resolution for improved epileptic seizure detection.
• * To develop an effective classification method for distinguishing seizure from seizure-free EEG epochs.

Main Methods:

• * Developed a Multiscale Radial Basis Functions (MRBF) and Modified Particle Swarm Optimization (MPSO) framework for time-frequency feature extraction.
• * Employed Principle Component Analysis (PCA) for dimensionality reduction of extracted features.
• * Utilized a Support Vector Machine (SVM) classifier with a Radial Basis Function (RBF) kernel for classification.

Main Results:

• * The proposed MRBF-MPSO-SVM method demonstrated superior classification accuracy compared to state-of-the-art feature extraction algorithms and other classifiers.
• * Achieved effective separation between seizure and seizure-free EEG epochs.
• * Validated the effectiveness of the novel time-frequency feature extraction approach.

Conclusions:

• * The MRBF-MPSO-SVM framework offers a significant advancement in the automatic detection and classification of epileptic seizures from EEG data.
• * The proposed method effectively addresses the challenges posed by the nonstationary nature of EEG signals during seizures.
• * This approach shows high potential for clinical application in epilepsy diagnosis and management.