Institutional Repository of Vilnius University
Elektroencefalografijos (EEG) analizė epilepsijos priepuoliams nustatyti naudojant mašininio mokymosi metodus /
Abstract
dc:descriptionThe aim of the research is to train a classifier that can classify images of EEG tests into pathological and non-pathological and create a web application that analyzes EEG tests and displays information about the epileptic seizures found in the tests. To create the training and testing data for the classifier, the „CHB-MIT Scalp EEG Database“ was chosen. The database consists of EEG tests from 22 patients of different ages. After conducting analysis of existing scientific articles about CNN usage for seizure detection, it was decided to split the EEG test into 5 second segments, where images were created from the signals. Three of the most popular CNN architectures for image classification – “ResNet50”, “InceptionV3”, “InceptionResNetV2” were compared to find out, which one will produce the best results. Since medical literature states, that epileptic seizures last at least 30 seconds, additional logic was implemented in order to clean up false positive classifications of seizures that are too short. With the additional logic added, the best results were produced by model with “InceptionResNetV2” architecture, which achieved 99.42% accuracy with testing data and found all the seizures that appeared in the test files. Additionally, an information system was created, where a user can upload EEG test files and see the sections, where the trained CNN model found seizures.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2022
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Aliulis, Martynas,
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:ELABAETD146240756&prefLang=en_US