Han-Yi Chiu
BioE M.S. Defense
May 8th,2023
3:00 PM
Location: Coda C1308 Cabbagetown
Committee:
Anqi Wu, Ph.D. (Advisor) (School of Computational Science and Engineering, Georgia Institute of Technology)
Zhao Liang, Ph.D. (Department of Computer Science, Emory University)
Eva Dyer, Ph.D. (Coulter Department of Biomedical Engineering, Georgia Institute of Technology)
Vince D. Calhoun, Ph.D. (School of Electrical and Computer Engineering, Georgia Institute of Technology)
Structure-based dynamical functional connectivity study for neuroimaging
Decoding of brain neural network has been an intriguing topic in neuroscience for well-rounded understanding of different types of brain disorders and cognitive stimuli. Integration of different types of connectivity, e.g., Functional Connectivity(FC) and Structural Connectivity(SC),from multimodal imaging techniques can take their complementary information into account and therefore have potential getting better decoding capability. However, traditional approaches for integrating FC and SC overlook the dynamical variations, which stand a great chance to over-generalize the brain neural network. In this paper, we propose JGAT model, a Graph Attention Network(GAT) framework with Joint kernel to combine data from functional Magnetic Resonance Images(fMRI) and Diffusion Weighted Imaging(DWI) and preserve the dynamical information at the same time. We build Joint kernel Graph Attention Network(JGAT) model performing classification tasks on four independent datasets: three of 7T fMRI datasets from Human Connectome Project(HCP) and one dataset from animal neural recordings. Furthermore, with the Attention Scores(AS) and Frame Scores(FS) computed and learned from the model, we can locate several informative segments and build meaningful dynamical pathways along temporal domain for HCP datasets.