TY - GEN
T1 - Discriminative-Region-Aware Residual Network for Adolescent Brain Structure and Cognitive Development Analysis
AU - Pan, Yongsheng
AU - Liu, Mingxia
AU - Wang, Li
AU - Xia, Yong
AU - Shen, Dinggang
PY - 2019/1/1
Y1 - 2019/1/1
N2 - The brains of adolescents undergo profound cognitive development, especially the development of fluid intelligence (FI) that is the ability to reason and think logically (independent of acquired knowledge). Such development may be influenced by many factors, such as changes in the brain structure caused by neurodevelopment. Unfortunately, the association between brain structure and fluid intelligence is not well understood. Cross-sectional structural MRI data released by the Adolescent Brain Cognitive Development (ABCD) study pave a way to investigate adolescents’ brain structure via MRIs, but each 3D volume may contain irrelevant or even noisy information, thus degrading the learning performance of computer-aided analysis systems. To this end, we propose a discriminative-region-aware residual network (DRNet) to jointly predict FI scores and identify discriminative regions in brain MRIs. Specifically, we first develop a feature extraction module (containing several convolutional layers and ResNet blocks) to learn MRI features in a data-driven manner. Based on the learned feature maps, we then propose a discriminative region identification module to explicitly determine the weights of different regions in the brain, followed by a regression module to predict FI scores. Experimental results on 4, 154 subjects with T1-weighted MRIs from ABCD suggest that our method can not only predict fluid intelligence scores based on structural MRIs but also explicitly specify those discriminative regions in the brain.
AB - The brains of adolescents undergo profound cognitive development, especially the development of fluid intelligence (FI) that is the ability to reason and think logically (independent of acquired knowledge). Such development may be influenced by many factors, such as changes in the brain structure caused by neurodevelopment. Unfortunately, the association between brain structure and fluid intelligence is not well understood. Cross-sectional structural MRI data released by the Adolescent Brain Cognitive Development (ABCD) study pave a way to investigate adolescents’ brain structure via MRIs, but each 3D volume may contain irrelevant or even noisy information, thus degrading the learning performance of computer-aided analysis systems. To this end, we propose a discriminative-region-aware residual network (DRNet) to jointly predict FI scores and identify discriminative regions in brain MRIs. Specifically, we first develop a feature extraction module (containing several convolutional layers and ResNet blocks) to learn MRI features in a data-driven manner. Based on the learned feature maps, we then propose a discriminative region identification module to explicitly determine the weights of different regions in the brain, followed by a regression module to predict FI scores. Experimental results on 4, 154 subjects with T1-weighted MRIs from ABCD suggest that our method can not only predict fluid intelligence scores based on structural MRIs but also explicitly specify those discriminative regions in the brain.
UR - http://www.scopus.com/inward/record.url?scp=85076300960&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85076300960&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-35817-4_17
DO - 10.1007/978-3-030-35817-4_17
M3 - Conference contribution
AN - SCOPUS:85076300960
SN - 9783030358167
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 138
EP - 146
BT - Graph Learning in Medical Imaging - 1st International Workshop, GLMI 2019, held in Conjunction with MICCAI 2019, Proceedings
A2 - Zhang, Daoqiang
A2 - Zhou, Luping
A2 - Jie, Biao
A2 - Liu, Mingxia
PB - Springer
T2 - 1st International Workshop on Graph Learning in Medical Imaging, GLMI 2019 held in conjunction with the 22nd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2019
Y2 - 17 October 2019 through 17 October 2019
ER -