A novel multi-class imbalanced EEG signals classification based on the adaptive synthetic sampling (ADASYN) approach

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Electroencephalography (EEG) is a key tool for epilepsy prediction.
• Early epilepsy diagnosis is vital for effective treatment and patient health.
• Community-wide epilepsy prevalence necessitates reliable prediction methods.

Purpose of the Study:

• To develop and evaluate novel methods for classifying five distinct EEG signal types.
• To improve the accuracy of epilepsy detection using machine learning techniques.
• To identify the most effective classification approach for EEG-based epilepsy prediction.

Main Methods:

• EEG signals were categorized into five classes: normal, tumor-related, epileptic, eyes closed, and eyes open.
• Four distinct classification approaches were proposed, involving feature extraction and Random Forest classifiers.
• The fourth approach integrated the adaptive synthetic (ADASYN) sampling method with Random Forest for imbalanced data handling.

Main Results:

• The first approach yielded 71.90% classification accuracy.
• The second and third approaches achieved 91.08% and 89% accuracy, respectively.
• The fourth approach, utilizing ADASYN and Random Forest, demonstrated the highest accuracy at 91.72%.

Conclusions:

• The combination of ADASYN sampling and Random Forest classifier is highly effective for classifying five-class EEG signals.
• This advanced method shows significant potential for real-time epilepsy event detection.
• Improved EEG signal classification can lead to earlier and more accurate epilepsy diagnoses.