EPILEPTIC SEIZURE PREDICTION FROM EEG SIGNAL RECORDINGS ON PEDIATRIC SUBJECTS USING MACHINE LEARNING TECHNIQUE

• Main Author: Nabila, Yumna
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/79799
• Institution: Institut Teknologi Bandung

Epilepsy is one of the most common cronic disease in the world. Nearly 50 million individuals worldwide are afflicted by this condition. The seizures are caused by excessive synchronization of neurons electrical excitation which spreads to the whole area of the brain. This diseas can affect anyone, regardless of their gender, age or racial group. Hence, individuals with epilepsy must adapt their lives around their illness carefully. When a seizure occur, it may bring some injury or even make life risky to the patient or others, mainly dealing with heavy machinery industry or driving vehicles. The purpose of this research is to design an epileptic seizure prediction system to warn patient before an attack occurs. The datasets used in this research are from CHB-MIT EEG Scalp database. The Seizure Prediction Horizon (SPH) used in this research is 10 minutes. The features chosen are energy and Dispersion Entropy (DispEN) acquired from Discrete Wavelet Transform (DWT) signal decomposition. At last, the classification and seizure prediction technique we deemed best is Support Vector Machine (SVM). The highest accuracy achieved by this model is 86,2% on a binary classification scenario Ictal and Non-Ictal, which is the Non-Ictal state is a combination of Normal and Pre-ictal state. While the highest sensitivity performance of our prediction system is 91%, with the lowest FPR average of 0,45.