Institutional Repository of Vilnius University
Epileptic seizure detection in electroencephalography (eeg) using machine learning methods /
Abstract
dc:descriptionEpileptic seizure detection is a critical application of machine learning in the medical field, aimed at assisting doctors in diagnosing, predicting, and monitoring epilepsy. This work uses Convolutional Neural Networks (CNNs) to detect seizures from EEG data—a complete process covering data preparation, choosing a model, and evaluating its performance. Raw EEG data from the CHB-MIT Scalp EEG Database was processed and augmented using a sliding window approach, balancing the dataset and improving the representation of the seizure class. Various pre-processing techniques were employed to ensure compatibility with CNN models, including filtering, normalization, and image transformation. This study compared two advanced CNN architectures, InceptionV3 and InceptionResNetV2. The InceptionV3 model demonstrated superior performance, achieving a sensitivity of 91.99%, specificity of 86.13%, and overall accuracy of 89.05%. The model was trained on a balanced dataset, addressing class imbalance and ensuring robust seizure detection. The proposed approach was validated by comparing the metrics against existing studies, demonstrating competitive results in sensitivity and F-score metrics and seizure classification. A web application was developed to visualize EEG data and automate seizure detection. It provides an accessible interface for users and includes features for uploading EEG files and visualizing seizure epochs and entire EEG data. This study highlights the effectiveness of CNN-based architectures in detecting seizures and offers a practical tool for clinical decision-making.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Zykutė, Gerda,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD229556379&prefLang=en_US