EEG Channel-Selection Method for Epileptic-Seizure Classification Based on Multi-Objective Optimization.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Intelligence

Background:

• Epileptic seizure classification relies heavily on electroencephalographic (EEG) data.
• Selecting optimal EEG channels is crucial for efficient and accurate seizure detection.
• Existing methods may not effectively balance classification accuracy with channel reduction.

Purpose of the Study:

• To develop and evaluate a multi-objective optimization method for EEG channel selection in epileptic seizure classification.
• To utilize the Non-Dominated Sorting Genetic Algorithm (NSGA) for optimizing channel selection.
• To assess the trade-off between classification accuracy and the number of EEG channels used.

Main Methods:

• Applied Non-Dominated Sorting Genetic Algorithm II (NSGA-II) and NSGA-III for multi-objective optimization.
• Decomposed EEG data into frequency bands using Empirical Mode Decomposition (EMD) or Discrete Wavelet Transform (DWT).
• Extracted four features per sub-band: two energy values and two fractal dimension values.

Main Results:

• Achieved classification accuracies up to 1.00 using only a single EEG channel.
• Demonstrated superior performance compared to using all available EEG channels in certain cases.
• Observed higher accuracy with NSGA-selected channels (e.g., 0.975 with two channels) than with all channels (0.95) for patient 19.

Conclusions:

• The proposed NSGA-based method effectively selects optimal EEG channels for epileptic seizure classification.
• Accurate seizure classification is achievable with a reduced number of EEG channels.
• Findings support the development of portable, efficient EEG-based seizure detection devices.