Page 118 - AN-4-2

Advanced Neurology

ORIGINAL RESEARCH ARTICLE
Machine learning-based recognition of epileptic

and non-epileptic EEG signals

Daniel Nasef , Viola Sawiris, Demarcus Nasef, and Milan Toma*
Department of Osteopathic Manipulative Medicine, College of Osteopathic Medicine, New York
Institute of Technology, Old Westbury, NY, United States of America
(This article belongs to the Special Issue: Advances in the pathogenesis, diagnosis and treatment
of epilepsy)

Abstract
Epilepsy is a chronic neurological disorder affecting approximately 50 million
people worldwide. Accurate and efficient detection of epileptic seizures is crucial
for effective treatment and management. Electroencephalogram (EEG) signals,
being non-invasive and rich in temporal information, are widely used for epilepsy
diagnosis. However, manual inspection of EEG data is time-consuming and relies
heavily on the expertise of clinicians. Machine learning techniques offer promising
solutions for automating the classification of epileptic and non-epileptic EEG
signals. In this study, we investigate the performance of various machine learning
models – including Light Gradient Boosting Machine, deep learning architectures,
and convolutional neural networks (CNN)—in classifying EEG signals for epilepsy
detection. Our experiments demonstrate that CNN outperform other models due to
their ability to capture complex spatial and temporal patterns inherent in EEG data.
*Corresponding author: The CNN model achieved higher accuracy and better convergence, as evidenced by
Milan Toma the confusion matrix and learning curves. In contrast, Deep Neural Networks without
(tomamil@tomamil.com)
convolutional layers showed lower performance, likely due to their limitations in
Citation: Nasef D, Sawiris V, capturing the intricate features of EEG signals. Similarly, the Light Gradient Boosting
Nasef D, Toma M. Machine
learning-based recognition of Machine model exhibited good initial results but failed to generalize well to unseen
epileptic and non-epileptic EEG data, possibly due to overfitting and lack of convergence. These findings highlight
signals. Adv Neurol. 2025;4(2):112-122. the potential of CNN-based approaches in the automated recognition of epileptic
doi: 10.36922/an.7941
seizures using EEG signals, paving the way for more efficient and accurate diagnostic
Received: December 18, 2024 tools.
Revised: March 7, 2025
Accepted: March 11, 2025 Keywords: Epilepsy; Machine learning; Electroencephalogram; Dense Neural Network;
Convolution neural network
Published online: March 24, 2025
Copyright: © 2025 Author(s).
This is an Open-Access article
distributed under the terms of the
Creative Commons Attribution 1. Introduction
License, permitting distribution,
and reproduction in any medium, Epilepsy is a chronic neurological condition characterized by having at least two
provided the original work is unprovoked seizure episodes at least 2 h apart. A seizure is a sudden burst of activity in
1
properly cited. the brain. Seizures can have various causes, including strokes, head injuries, neoplasm,
2
Publisher’s Note: AccScience and infections. However, in many cases, the exact cause remains unknown. There are
Publishing remains neutral with numerous types of seizures, each affecting different parts of the body and presenting
regard to jurisdictional claims in 3
published maps and institutional unique symptoms. Most seizures can be separated into either generalized or focal seizures.
affiliations. Focal seizures are caused by abnormal brain activity in a specific area of the brain and

Volume 4 Issue 2 (2025) 112 doi: 10.36922/an.7941