Classification of Normal, Ictal and Inter-ictal EEG via Direct Quadrature and Random Forest Tree

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neurology

Background:

• Epilepsy diagnosis relies on analyzing electroencephalographic (EEG) signals.
• EEG signals exhibit complex temporal and spectral disturbances during seizures.
• Accurate classification of EEG signals is crucial for timely diagnosis and treatment.

Purpose of the Study:

• To develop an accurate nonlinear classification method for diagnosing seizures from EEG signals.
• To improve the distinction between healthy, ictal, and inter-ictal EEG activities.
• To provide a robust and interpretable tool for clinical use.

Main Methods:

• EEG signals (healthy, ictal, inter-ictal) were decomposed using Empirical Mode Decomposition (EMD).
• Instantaneous amplitudes and frequencies were extracted using the Direct Quadrature (DQ) method, decoupling amplitude modulation effects.
• Shannon entropy of instantaneous amplitude and frequency values was calculated for each Intrinsic Mode Function (IMF).
• Classified entropy values using a Random Forest (RF) classifier.

Main Results:

• Achieved 100% accuracy in classifying healthy versus ictal EEG signals.
• Attained 98.3-99.7% accuracy for classifying healthy, ictal, and inter-ictal EEG signals.
• Demonstrated the method's robustness, speed, and user-friendliness.

Conclusions:

• The proposed nonlinear classification method offers high accuracy for seizure detection in EEG signals.
• The technique effectively separates amplitude and frequency information, preventing feature confusion.
• This data-driven approach with open-access interpretability is suitable for clinical applications.