Institutional Repository of Vilnius University
Internetinė vėžio diagnostikos programa: InceptionResNetV2 modelio pritaikymas krūties vėžio prognozavimui /
Abstract
dc:descriptionThis paper explores the application of one type of artificial neural networks in breast cancer classification using mammograms. The paper starts with an overview of neural networks, their structure and types. The work focuses more on convolutional neural networks, which are particularly effective in image analysis. The main components of a convolutional neural network are discussed, such as convolutional, pooling and fully connecte layers. Then, different neural network learning techniques and deep learning in artificial intelligence are discussed. In the paper, we also examine model overfitting and regularization techniques to address this problem. Loss and activation functions and optimisation algorithms are considered. The metrics used in classification are discussed. The dataset used in this work is a set of breast cancer mammograms in DICOM format with an unbalanced distribution of classes, cancer and non-cancer. The work highlights the challenges posed by this unbalanced set in model training. Finally, the process of data preparation for neural network training is reviewed. The process involves converting DICOM images to PNG format, fitting the sizes to the model, partitioning the data into training and validation sets and augmentation.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Augustauskas, Karolis,
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:ELABAETD192845032&prefLang=en_US