Back to search

Institutional Repository of Vilnius University

Epilepsijos priepuolių paieškos algoritmai EEG signaluose /

Abstract

dc:description

Epilepsy is a neurological disease that causes seizures which can occur as convulsions. It is dangerous because losing consciousness during seizures can lead to serious injuries, so machine learning algorithm implementation can help doctors automatically diagnose epilepsy, which will allow them to handle more patients at the same time than usual and focus on treatment, rather than analyzing EEG signals for hours. The goal of this bachelor’s thesis is to implement a support vector machine algorithm for automatic detection of epileptic seizures. The study begins with a literature analysis of EEG and their systems, EDF files, a selected dataset for the current research, the theory of support vector machines, and its performance evaluation metrics. The most important part of this research is the methodology of SVM algorithm implementation and its results. Research begins with the collection of technical information from patient EDF files, which were selected from Guttag, J. (2010) CHB-MIT scalp EEG dataset (version 1.0.0). After getting the necessary information, it is useful to perform channel unification, EEG signals preprocessing, segmentation, seizure labeling, EEG data balancing, discrete wavelet transformation (DWT), feature extraction, SVM training, and postprocessing. Feature extraction includes 4 main features such as EEG signal relative energy, amplitude, coefficient of variation, and fluctuation index. The experiment compared two SVM types: linear and non-linear (kernel). The kernel selected for this research was RBF. Results were compared before postprocessing, and after that, in both ways better accuracy and specificity results were reached by linear SVM, however, in sensitivity, RBF SVM appeared better. Three SVM models were developed during the study: the most sensitive, the most specific, and the compromise model. The most sensitive model detected 100 % of all seizures, the most specific model correctly identified 93 % of the seizures as absent, and the compromise model, after tuning the SVM function parameters, achieved extremely good results: accuracy of 94 %, sensitivity of 90 %, and specificity of 94 %. The latter model was used in a desktop application designed to automatically detect epileptic seizures. This research could be improved by testing the SVM model on other databases. It could also include removal of irrelevant data that do not become support vectors and perform misclassification analysis in order to find points that mislead the classification model.

Degree

thesis:*
Grantor dc:publisher
Institutional Repository of Vilnius University
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Grybauskaitė, Justina,

Rights

dc:rights
Statement dc:rights
  • info:eu-repo/semantics/openAccess
Language dc:language
lit

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
Vilnius University
Base URL
epublications.vu.lt/oai
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Grybauskaitė, Justina,. Epilepsijos priepuolių paieškos algoritmai EEG signaluose /. Institutional Repository of Vilnius University, 2026. https://repository.vu.lt/VU:ELABAETD277689596&prefLang=en_US