Institutional Repository of Vilnius University
Giliojo mokymosi metodai sintetinių asmens medicininių duomenų bazių generavimui /
Abstract
dc:descriptionDeep learning is having a profound impact on all industries and medicine is no exception. However, progress in this field is not so fast and easy, especially when it comes to tabular medical data, which is linked to a patient's personal data. The lack of such accessible data is due to the increasingly stringent privacy requirements of legal regulation. The use of synthetic data could help to address this problem. Generated realistic, high-quality datasets can be used to accelerate progress. Therefore, the aim of this study is to build deep learning models for the generation of tabular medical data, exclusively with generative adversarial networks, and to assess the similarity of the generated data to real data, its quality and its usability in 2 ways, directly and indirectly. The experiments of the study succeeded in generating synthetic databases, but it was concluded that synthetic data cannot fully replace real data, at least in the current time. Synthetic data may be fully usable in the initial stages of work or research, but real data is still needed for further progress to maximise the performance of the models and subsequently the tools.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Orliukas, Lukas,
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:ELABAETD210647735&prefLang=en_US