Document Type : Research Article
Authors
1
Health Technology Research Center, Imam Reza International University, Mashhad, Iran
2
Department of Engineering and Technology, University of Mazandaran, Iran,
3
Complex Systems and Cybernetic Control Laboratory, Biomedical Engineering Faculty, AmirKabir University of Technology, Tehran, Iran
4
5 Epilepsy Research Center, Department of Neurosurgery, Westfälische Wilhelms-Universitat Münster, Münster, Germany
10.22080/frai.2025.29345.1019
Abstract
Epilepsy is a neurological disorder characterized by recurrent seizures resulting from abnormal brain activity. To understand its underlying dynamics, this study applies nonlinear analysis to EEG data recorded during seizures in patients with temporal lobe epilepsy. Specifically, the study explores the evolution of nonlinear features, such as fractal dimension, correlation dimension, entropy, and Recurrence Quantification Analysis (RQA), to classify seizure patterns. EEG recordings from the Fz-Cz channel of 209 seizures across 24 patients were analyzed. These features were used to perform unsupervised clustering using k-means and hierarchical methods to identify recurring patterns and classify seizure quality. The fractal dimension, which captures signal complexity and self-similarity, proved especially informative by showing clear trends throughout seizure progression. The correlation dimension quantified the spatial structure of the signal’s reconstructed phase space, while entropy measured signal unpredictability typically decreasing during seizures, reflecting a transition from chaotic to more structured brain activity. The clustering analysis grouped seizures into 3 to 5 categories, with most seizures from each patient clustering together, suggesting consistent intra-patient seizure dynamics. Among the features, RQA-based measures were the most effective, clustering 87.7% of seizures into a single group, followed by phase space features (71.54%) and fractal dimension (56%). These findings suggest that seizure dynamics, while complex, exhibit repetitive and deterministic behavior across episodes for the same individual. The study supports the hypothesis that epileptic seizures reduce brain complexity, creating more structured and rhythmic patterns compared to normal function. In conclusion, nonlinear EEG analysis effectively characterizes and classifies seizure events, highlighting the deterministic nature of temporal lobe epilepsy. These insights could improve seizure prediction and aid in developing personalized treatment strategies. Future work should expand to other epilepsy types and
Keywords
)