Loading...
Seizure Detection using Chaotic Analysis of EEG
Hallak, Alaa Matar, Amal Alamro, Hala
Hallak, Alaa
Matar, Amal
Alamro, Hala
Citations
Altmetric:
Type
Supervisor
Subject
Date
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
Epilepsy is a neurological disorder characterized by recurrent seizures caused by abnormal
electrical activity in the brain. Epileptic seizure detection is a critical challenge in neurological
care, often relying on electroencephalography (EEG) for monitoring abnormal brain activity.
While automatic EEG-based systems offer promise, traditional linear analysis methods are
limited in capturing the complex nonlinear dynamics underlying seizure events, resulting in
reduced reliability and missed events in clinical applications. This study addresses this research
gap by proposing an advanced seizure detection system that integrates time-domain, frequency
domain, and nonlinear chaotic features including the Largest Lyapunov Exponent. The system
utilizes the CHB-MIT public EEG dataset and employs a rigorous four-stage preprocessing
pipeline: band-pass and notch filtering, wavelet denoising, artifact suppression and
dimensionality reduction using PCA, followed by class balancing using SMOTE. Compact sets
of features are engineered for multiple EEG channels and are evaluated using three machine
learning classifiers, Random Forest, XGBoost, and Support Vector Machine, to assess
performance gains from nonlinear feature integration. Experiments demonstrate that including
chaotic metrics substantially improves detection metrics such as recall, F1-score, and ROC
AUC across all models. XGBoost achieves the strongest overall performance while SVM-RBF
provides the highest Recall and the lowest number of missed seizure events, demonstrating
strong clinical safety potential. These findings confirm that chaos-based EEG features can
significantly enhance automated seizure detection performance.