Automatic identification of epilepsy by HOS and power spectrum parameters using EEG signals: a comparative study

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Epilepsy is a neurological disorder characterized by unpredictable seizures.
• Automatic seizure onset detection is crucial for patient safety and research.
• Higher-order spectra (HOS) offer a promising avenue for analyzing complex EEG signals.

Purpose of the Study:

• To develop and evaluate an automatic system for detecting epilepsy seizure onsets.
• To compare the efficacy of HOS-based features against traditional power spectrum density (PSD) features for seizure detection.

Main Methods:

• EEG signals were analyzed to extract features from the power spectrum and bispectrum.
• Non-linear features derived from higher-order spectra (HOS) were utilized.
• A Gaussian mixture model (GMM) classifier was employed to differentiate between normal, pre-ictal, and epileptic EEG signals.

Main Results:

• HOS-based features achieved a classification accuracy of 93.11%.
• Features derived from PSD resulted in a classification accuracy of 88.78%.
• Selected HOS features demonstrated superior performance in seizure detection.

Conclusions:

• HOS-based feature extraction is a highly effective method for improving the accuracy of automatic epilepsy seizure detection.
• This approach offers a significant advancement over traditional PSD methods.
• The developed system holds potential for real-world clinical applications and epilepsy research.