Parkinson’s disease classification through multi-view learning of structural and functional connectome
The connectome is measured via the structural connectome, comprised of white matter connections between brain regions, and the functional connectome, comprised of correlated activations across brain regions. This warrants the need for multi-view learning techniques to encode the brain. However, rece...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2021
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/148057 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | The connectome is measured via the structural connectome, comprised of white matter connections between brain regions, and the functional connectome, comprised of correlated activations across brain regions. This warrants the need for multi-view learning techniques to encode the brain. However, recent advances that capture non-linearities in multiple views have not made use of novel graph encoding techniques. In this project, we propose Graph-encoded Generalized Canonical Correlation Analysis (GGCCA) for multi-view learning of the functional and structural connectome. GGCCA obviates the problem of overfitting due to high dimensionality of connectome data by direct encoding of the graph structure. Furthermore, GGCCA can aggregate node vectors at the level of functional modules, allowing multiple levels of integration across views. GGCCA is also able to connect connectomes from atlases of varied sizes, making it very generalizable to other combinations of modalities or similarity graphs. |
|---|