Epileptic Seizure Prediction Using Machine Learning Techniques on Real-Time EEG Signals

This paper is designed to use Machine Learning Techniques on real-time EEG signals for predicting epileptic seizures. Effective and accurate seizure prediction systems can help epileptic patients to get proper and timely medical care and attention, thereby, leading the patients to have an improved life quality. The ability to accurately identify the pre-ictal state enables the patient to take sufficient and adequate preventive measures before-hand. Epileptic Seizure Prediction provides an accurate classification of the brain signals between the pre-ictal state and inter-ictal state. The Dual Tree-Complex Wavelet Transform (DT-CWT) is used as an EEG Signal processing and analysis unit. Frequency domain features and temporal domain features are extracted from the signals. Support Vector Machine with K-fold cross-validation is used as a classifier. This model gives 85.9% accuracy on real-time EEG signals with a prediction time of 60 minutes from the onset of an epileptic seizure.