- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- Predicting primary diffusion directions in white matter...
Open Collections
UBC Theses and Dissertations
UBC Theses and Dissertations
Predicting primary diffusion directions in white matter from T₁-weighted magnetic resonance images Connolly, Sam
Abstract
Orientation dependence studies are an ongoing research area in Magnetic Resonance Imaging (MRI), often relying on diffusion MRI to gain nerve fiber direction, a time-consuming scan not always included in research and clinical protocols. Using paired 3D T₁-weighted and diffusion images from the Human Connectome Project (HCP), a U-Net was trained to find voxel-wise principal diffusion directions from 3D T₁-weighted images. Different loss functions and data augmentations were used in model training. The models trained with different data augmentations and loss functions were evaluated using cosine distance during training. The trained model with the best performance had a group average cosine distance of 0.349 ± 0.007 and a group average difference in absolute polar angle of 8.762° ± 0.293° on the independent test set.
Item Metadata
| Title |
Predicting primary diffusion directions in white matter from T₁-weighted magnetic resonance images
|
| Creator | |
| Supervisor | |
| Publisher |
University of British Columbia
|
| Date Issued |
2024
|
| Description |
Orientation dependence studies are an ongoing research area in Magnetic Resonance Imaging (MRI), often relying on diffusion MRI to gain nerve fiber direction, a time-consuming scan not always included in research and clinical protocols. Using paired 3D T₁-weighted and diffusion images from the Human Connectome Project (HCP), a U-Net was trained to find voxel-wise principal diffusion directions from 3D T₁-weighted images. Different loss functions and data augmentations were used in model training. The models trained with different data augmentations and loss functions were evaluated using cosine distance during training. The trained model with the best performance had a group average cosine distance of 0.349 ± 0.007 and a group average difference in absolute polar angle of 8.762° ± 0.293° on the independent test set.
|
| Genre | |
| Type | |
| Language |
eng
|
| Date Available |
2024-04-15
|
| Provider |
Vancouver : University of British Columbia Library
|
| Rights |
Attribution-NonCommercial-NoDerivatives 4.0 International
|
| DOI |
10.14288/1.0441341
|
| URI | |
| Degree | |
| Program | |
| Affiliation | |
| Degree Grantor |
University of British Columbia
|
| Graduation Date |
2024-05
|
| Campus | |
| Scholarly Level |
Graduate
|
| Rights URI | |
| Aggregated Source Repository |
DSpace
|
Item Media
Item Citations and Data
Rights
Attribution-NonCommercial-NoDerivatives 4.0 International