TY - GEN
T1 - Learning Human Cognition via fMRI Analysis Using 3D CNN and Graph Neural Network
AU - Ni, Xiuyan
AU - Gao, Tian
AU - Wu, Tingting
AU - Fan, Jin
AU - Chen, Chao
N1 - Publisher Copyright:
© Springer Nature Switzerland AG 2019.
PY - 2019
Y1 - 2019
N2 - 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.
AB - 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.
KW - Brain network learning
KW - Graph neural network
UR - https://www.scopus.com/pages/publications/85075587434
U2 - 10.1007/978-3-030-33226-6_11
DO - 10.1007/978-3-030-33226-6_11
M3 - Conference contribution
AN - SCOPUS:85075587434
SN - 9783030332259
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 93
EP - 101
BT - Multimodal Brain Image Analysis and Mathematical Foundations of Computational Anatomy - 4th International Workshop, MBIA 2019, and 7th International Workshop, MFCA 2019, Held in Conjunction with MICCAI 2019, Proceedings
A2 - Zhu, Dajiang
A2 - Yan, Jingwen
A2 - Huang, Heng
A2 - Shen, Li
A2 - Thompson, Paul M.
A2 - Westin, Carl-Fredrik
A2 - Pennec, Xavier
A2 - Joshi, Sarang
A2 - Nielsen, Mads
A2 - Sommer, Stefan
A2 - Fletcher, Tom
A2 - Durrleman, Stanley
PB - Springer
T2 - 4th International Workshop on Multimodal Brain Image Analysis, MBAI 2019, and the 7th International Workshop on Mathematical Foundations of Computational Anatomy, MFCA 2019, held in conjunction with the 22nd International Conference on Medical Imaging and Computer Assisted Intervention, MICCAI 2019
Y2 - 17 October 2019 through 17 October 2019
ER -