Institutional Repository of Vilnius University
Transformeriais pagrįstų mašininio mokymosi metodų palyginimas epilepsijos priepuolių aptikimui /
Abstract
dc:descriptionThe aim of this work is to investigate the capabilities of transformer-based artificial neural networks in detecting epileptic seizures in electroencephalography data, taking into account the model’s resource consumption and accuracy. Experiments were conducted using the „CHB-MIT“ dataset by training 3 models under 3 different scenarios. The benefit of additional input data preprocessing was also evaluated. A set of resource consumption metrics was selected to assess each architecture. The results show that additional data preprocessing does not provide extra benefit for the tested transformer neural network architectures. The „EEGConformer“ architecture demonstrated the highest accuracy and specificity. The least resource-intensive model was „ATCNet“, which also showed comparable accuracy metrics to the „EEGConformer“. Additional filtering of model outputs was also investigated. With appropriate filtering, it is possible to significantly improve the model’s ability to accurately detect epileptic seizures.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Bakutis, Vilius,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD239029826&prefLang=en_US