Seizure detection from EEG signals using Multivariate Empirical Mode Decomposition

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Electroencephalogram (EEG) signals are crucial for diagnosing neurological disorders like epilepsy.
• Analyzing non-stationary EEG data for seizure detection presents significant challenges.
• Existing methods may struggle with the complexity and multi-channel nature of EEG signals.

Purpose of the Study:

• To develop and validate a data-driven approach for classifying ictal and non-ictal EEG signals.
• To utilize the multivariate empirical mode decomposition (MEMD) algorithm for enhanced EEG signal analysis.
• To improve the accuracy and reliability of automated epileptic seizure detection.

Main Methods:

• Application of the multivariate empirical mode decomposition (MEMD) algorithm for signal decomposition.
• Extraction of multiscale time-frequency (T-F) features from EEG data using MEMD.
• Classification of EEG signals into ictal and non-ictal categories using artificial neural networks (ANNs).
• Validation of the proposed method on extensive, publicly available EEG datasets.

Main Results:

• The MEMD-based feature extraction effectively captures relevant information from multi-channel EEG signals.
• Artificial neural networks achieved high accuracy in classifying EEG signals based on MEMD-derived features.
• The proposed method demonstrates robust performance across diverse and publicly available EEG datasets.

Conclusions:

• The proposed data-driven approach using MEMD and ANNs is effective for classifying ictal and non-ictal EEG signals.
• MEMD provides a powerful tool for analyzing the complex time-frequency characteristics of EEG data.
• This method offers a promising advancement for automated epileptic seizure detection systems.