Name
Abstract | ||
|---|---|---|
Human cognitive control involves how mental resources are allocated when the brain processes various information. The study of such complex brain functionality is essential in understanding different neurological disorders. To investigate cognition control, various cognitive tasks have been designed and functional MRI data have been collected. In this paper, we study uncertainty representation, an important problem in human cognition study, with task-evoked fMRI data. Our goals are to learn how brain region of interests (ROIs) are activated under tasks with different uncertainty levels and how they interact with each other. We propose a novel neural network architecture to achieve the two goals simultaneously. Our architecture uses a 3D convolutional neural network (CNN) to extract a high-level representation for each ROI, and uses a graph neural network module to capture the interactions between ROIs. Empirical evaluations reveal that our method significantly outperforms the existing methods, and the derived brain network is consistent with domain knowledge. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.1007/978-3-030-33226-6_11 | MULTIMODAL BRAIN IMAGE ANALYSIS AND MATHEMATICAL FOUNDATIONS OF COMPUTATIONAL ANATOMY |
Keywords | DocType | Volume |
Graph neural network, Brain network learning | Conference | 11846 |
ISSN | Citations | PageRank |
0302-9743 | 0 | 0.34 |
References | Authors | |
0 | 5 | |