Automatic detection method of epileptic seizures based on IRCMDE and PSO-SVM

Area of Science:

• Biomedical Engineering
• Signal Processing
• Computational Neuroscience

Background:

• Multi-scale dispersion entropy (MDE) is utilized for analyzing electroencephalography (EEG) signals to detect epileptic seizures.
• Existing MDE methods suffer from information loss and poor robustness in measuring time-series complexity.

Purpose of the Study:

• To propose an improved method for automatic epilepsy detection using enhanced nonlinear feature extraction.
• To address the limitations of MDE in capturing the complexity of EEG signals for epilepsy diagnosis.

Main Methods:

• Introduced refined composite multi-scale dispersion entropy (RCMDE) and developed improved RCMDE (IRCMDE) by replacing segmented average with local maximum calculation.
• Normalized entropy values in IRCMDE to enhance the robustness of feature parameters.
• Employed particle swarm optimization support vector machine (PSO-SVM) for classifying epileptic EEG signals using IRCMDE features.

Main Results:

• IRCMDE effectively eliminates information loss present in MDE and RCMDE when analyzing signal complexity.
• IRCMDE demonstrates reduced sensitivity to parameter selection variations compared to MDE and RCMDE.
• The IRCMDE-PSO-SVM approach achieved superior recognition accuracy for epileptic EEG signals compared to MDE-PSO-SVM and RCMDE-PSO-SVM.

Conclusions:

• IRCMDE offers a more robust and accurate method for nonlinear feature extraction from EEG signals.
• The proposed IRCMDE-PSO-SVM method significantly improves the accuracy of automatic epilepsy detection.
• This advancement holds promise for more reliable clinical diagnosis of epilepsy.