Institutional Repository of Vilnius University
Convolutional neural networks for neurodegenerative disease diagnosis /
Abstract
dc:descriptionThis research describes an experiment to train a convolutional neural network based on the ResNet3D architecture to classify magnetic resonance images into three categories: Alzheimer’s disease (AD), mild cognitive impairment (MCI), and no cognitive impairment (CN). The model was trained on a large amount of data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). The targets of this study were achieved by utilising Vilnius university high performance computing (HPC) resources and setting up Python and Clinica development environments. The model demonstrated high accuracy in the training phase (above 95%), and satisfactory results in the validation phase (above 75%). In the second test, the parameters of the training task were ad‐ justed by increasing the number of epochs from 10 to 100 and changing the batch size from 4 to 8. The results showed a slight improvement in accuracy during both training and validation, which is consistent with the findings of the analysed scientific literature. The findings of the study show that higher values of hyperparameters such as the number of epochs or the size of the batches do not significantly affect the accuracy of the model. Thus, further research should be conducted on adjusting internal parameters of these networks, as this may lead to a more significant increase in model’s performance metrics.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Vabolis, Valdas,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192846413&prefLang=en_US