Evolutionary transfer optimization-based approach for automated ictal pattern recognition using brain signals.

Area of Science:

• Neuroscience
• Computational Intelligence
• Biomedical Engineering

Background:

• Manual detection of epileptic seizures (ictal patterns) from brain signals is time-consuming and error-prone.
• Existing automated methods struggle with feature engineering due to inter-subject variability.
• Single-objective optimization methods provide suboptimal results for ictal pattern detection.

Purpose of the Study:

• To develop a novel expert system for automated detection of ictal patterns in brain signals.
• To employ evolutionary multi-objective optimization (EMO) for simultaneous minimization of features and error rates.
• To enhance the reliability and efficiency of seizure detection in clinical applications.

Main Methods:

• Utilized the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) for evolutionary multi-objective optimization (EMO).
• Extracted input features from phase space transformations, singular values, and wavelet packet transform coefficients.
• Implemented Evolutionary Transfer Optimization (ETO) to determine optimal feature sets and a Generalized Regression Neural Network (GRNN) for classification.

Main Results:

• The proposed EMO approach significantly reduced the feature set size by over 50%.
• Achieved high accuracy in ictal pattern detection with minimal computation time (<0.5 seconds).
• Demonstrated the reliability of the optimized feature set for clinical applications.

Conclusions:

• The novel expert system effectively automates ictal pattern recognition in EEG data.
• The EMO approach offers a reliable and efficient method for epilepsy diagnosis and treatment.
• Further validation across diverse datasets is recommended to explore limitations and potential.