Epilepsy seizure classification based on EEG signals using machine learning

Abstract

This thesis employs advanced machine learning approaches within a Design Science Research (DSR) framework to present comprehensive research into the classification of epilepsy seizures using electroencephalogram (EEG) signals. The need for precise and effective diagnostic tools is vital because epilepsy affects about 50 million people worldwide and has a significant negative impact on healthcare systems. The present research uses iterative development and evaluation to look at the efficacy of various machine learning algorithms, including Gradient Boosting, k-Nearest Neighbours (KNN), Support Vector Machines (SVM), Random Forest (RF), and k-Nearest Neighbours (RF).

The core objective is to enhance the precision and reliability of seizure classification beyond the capabilities of traditional EEG analysis methods, which are often time-consuming and subjective. By integrating multiple machine learning models, this study leverages their collective strengths, significantly improving diagnostic accuracy. The iterative DSR process culminates in the creation of a novel ensemble model using a stacking methodology, which demonstrates superior performance over individual algorithms through rigorous testing, achieving notable accuracy in distinguishing between different types of seizures.

In addition to offering an innovative computational point of view on epilepsy diagnosis, this work makes the way for future studies that will improve upon these techniques inside a standardised DSR framework. The implications of this study are profound, offering potential enhancements in personalized medicine and treatment planning for epilepsy, ultimately aiming to improve patient outcomes.