A MACHINE LEARNING-BASED APPROACH TO EPILEPTIC SEIZURE PREDICTION USING ELECTRO-ENCEPHALOGRAPHIC SIGNALS

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Epilepsy affects approximately 1% of the global population, characterized by altered brain signal patterns leading to seizures.
• Accurate diagnosis and treatment of epilepsy are crucial due to its widespread impact.
• Non-invasive electroencephalography (EEG) offers a viable method for monitoring brain activity.

Purpose of the Study:

• To develop a patient-independent machine learning approach for predicting epileptic seizures.
• To classify interictal (between seizures) and preictal (before seizures) states using EEG signals.
• To evaluate the effectiveness of different signal characteristics for seizure prediction.

Main Methods:

• Utilized the CHB-MIT EEG database for analysis.
• Applied Discrete Wavelet Transform for EEG signal decomposition into 5 levels.
• Extracted features including Spectral Power, Mean, and Standard Deviation.
• Employed a Support Vector Machine (SVM) as the classification algorithm.

Main Results:

• The machine learning model achieved an accuracy of 92.30% using Spectral Power as the characteristic.
• Standard Deviation yielded an accuracy of 84.60%, while Mean resulted in 76.92% accuracy.
• The patient-independent approach demonstrated the model's generalizability across subjects.

Conclusions:

• Spectral Power is the most effective characteristic for predicting epileptic seizures using this machine learning model.
• The developed approach shows significant potential for improving the diagnosis and management of epilepsy.
• Further research can refine this method for real-time clinical applications.