Back to search

Institutional Repository of Vilnius University

Giluminio mokymosi metodų taikymai EEG klasifikavimui pagal miego stadiją /

Abstract

dc:description

This thesis investigates the problem of automatic sleep stage classification based on singlechannel EEG signals using modern deep learning techniques. The main objective of the work is to analyze and compare different neural network architectures in terms of their ability to accurately classify sleep stages and generalize to previously unseen subjects. The study is conducted using the publicly available Sleep-EDF dataset, with a particular focus on subject-wise k-fold cross-validation to ensure reliable and unbiased performance evaluation. Three deep learning models are implemented and evaluated: a convolutional neural network (CNN) as a baseline approach, a bidirectional long short-term memory (BiLSTM) network for sequence modeling, and a Transformer-based model utilizing self-attention mechanisms. Extensive experiments are performed to assess the impact of temporal context modeling on classification performance. The results demonstrate that sequence-based models significantly outperform the epoch-based CNN model, achieving higher Macro-F1 scores and Cohen’s kappa coefficients. Furthermore, the Transformer model achieves the highest overall performance (96.9% accuracy, 89.2% macro F-1), while the BiLSTM model provides a favorable balance between classification accuracy and computational efficiency. Based on the experimental findings, it is concluded that incorporating temporal dependencies between EEG epochs is crucial for accurate sleep stage classification. The results of this work may contribute to the development of more reliable and efficient automated sleep analysis systems in both research and clinical settings. An online application based on Streamlit has been developed, integrating EEG data processing, model inference, and result visualization.

Degree

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

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Juknevičius, Vilius,

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

Juknevičius, Vilius,. Giluminio mokymosi metodų taikymai EEG klasifikavimui pagal miego stadiją /. Institutional Repository of Vilnius University, 2026. https://repository.vu.lt/VU:ELABAETD277689720&prefLang=en_US