TY - GEN
T1 - Graph-Based Deep Learning for Prediction of Longitudinal Infant Diffusion MRI Data
AU - Kim, Jaeil
AU - Hong, Yoonmi
AU - Chen, Geng
AU - Lin, Weili
AU - Yap, Pew Thian
AU - Shen, Dinggang
N1 - Funding Information:
Acknowledgements This work was supported in part by NIH grants (NS093842, EB022880, EB006733, EB009634, AG041721, MH100217, and AA012388), an NSFC grant (11671022, China), and Institute for Information & communications Technology Promotion (IITP) grant (MSIT, 2018-2-00861, Intelligent SW Technology Development for Medical Data Analysis, South Korea).
Publisher Copyright:
© 2019, Springer Nature Switzerland AG.
PY - 2019
Y1 - 2019
N2 - Diffusion MRI affords great value for studying brain development, owing to its capability in assessing brain microstructure in association with myelination. With longitudinally acquired pediatric diffusion MRI data, one can chart the temporal evolution of microstructure and white matter connectivity. However, due to subject dropouts and unsuccessful scans, longitudinal datasets are often incomplete. In this work, we introduce a graph-based deep learning approach to predict diffusion MRI data. The relationships between sampling points in spatial domain (x-space) and diffusion wave-vector domain (q-space) are harnessed jointly (x-q space) in the form of a graph. We then implement a residual learning architecture with graph convolution filtering to learn longitudinal changes of diffusion MRI data along time. We evaluate the effectiveness of the spatial and angular components in data prediction. We also investigate the longitudinal trajectories in terms of diffusion scalars computed based on the predicted datasets.
AB - Diffusion MRI affords great value for studying brain development, owing to its capability in assessing brain microstructure in association with myelination. With longitudinally acquired pediatric diffusion MRI data, one can chart the temporal evolution of microstructure and white matter connectivity. However, due to subject dropouts and unsuccessful scans, longitudinal datasets are often incomplete. In this work, we introduce a graph-based deep learning approach to predict diffusion MRI data. The relationships between sampling points in spatial domain (x-space) and diffusion wave-vector domain (q-space) are harnessed jointly (x-q space) in the form of a graph. We then implement a residual learning architecture with graph convolution filtering to learn longitudinal changes of diffusion MRI data along time. We evaluate the effectiveness of the spatial and angular components in data prediction. We also investigate the longitudinal trajectories in terms of diffusion scalars computed based on the predicted datasets.
KW - Brain development
KW - Diffusion MRI
KW - Graph convolution
KW - Graph representation
KW - Longitudinal prediction
KW - Residual graph neural network
UR - http://www.scopus.com/inward/record.url?scp=85066881735&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-05831-9_11
DO - 10.1007/978-3-030-05831-9_11
M3 - Conference contribution
AN - SCOPUS:85066881735
SN - 9783030058302
T3 - Mathematics and Visualization
SP - 133
EP - 141
BT - Mathematics and Visualization
A2 - Bonet-Carne, Elisenda
A2 - Grussu, Francesco
A2 - Ning, Lipeng
A2 - Sepehrband, Farshid
A2 - Tax, Chantal M.W.
PB - Springer Heidelberg
T2 - International Workshop on Computational Diffusion MRI, CDMRI 2018 held with International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018
Y2 - 20 September 2018 through 20 September 2018
ER -