TY - GEN
T1 - Interpretable Feature Learning Using Multi-output Takagi-Sugeno-Kang Fuzzy System for Multi-center ASD Diagnosis
AU - Wang, Jun
AU - Zhang, Ying
AU - Zhou, Tao
AU - Deng, Zhaohong
AU - Huang, Huifang
AU - Wang, Shitong
AU - Shi, Jun
AU - Shen, Dinggang
N1 - Funding Information:
Acknowledgements.. This work was supported in part by the Natural Science Foundation of Jiangsu Province (BK20151299, BK20151358, BK20161268 and BK20181339).
Publisher Copyright:
© 2019, Springer Nature Switzerland AG.
PY - 2019
Y1 - 2019
N2 - Autism spectrum disorder (ASD) is a neurodevelopmental disorder closely related to potential dysfunction of the brain. Although multiple functional connectivity (FC) features such as low-order functional connectivity (LOFC) and high-order functional connectivity (HOFC) provide complementary knowledge to each other, it is still challenging to find interpretable LOFC and HOFC features for multi-center ASD diagnosis. To this end, we develop a novel interpretable feature learning method based on multi-output TSK fuzzy system (MO-TSK-FS) for multi-center ASD diagnosis. Specifically, both the LOFC and HOFC features are first mapped to a high-dimensional space using the premise part of MO-TSK-FS, which shares the common knowledge across multiple centers. Then, the mapped features are transformed to a low-dimensional feature space using a transformation matrix. A novel unsupervised learning problem is formulated to find the optimal transformation matrix. Finally, a multi-modality support vector classifier (M2SVC) is constructed for classification. The experimental results show that the proposed interpretable feature learning method for multi-center ASD classification can effectively extract important features from the original LOFC and HOFC features, resulting in an efficient M2SVC for multi-center ASD classification.
AB - Autism spectrum disorder (ASD) is a neurodevelopmental disorder closely related to potential dysfunction of the brain. Although multiple functional connectivity (FC) features such as low-order functional connectivity (LOFC) and high-order functional connectivity (HOFC) provide complementary knowledge to each other, it is still challenging to find interpretable LOFC and HOFC features for multi-center ASD diagnosis. To this end, we develop a novel interpretable feature learning method based on multi-output TSK fuzzy system (MO-TSK-FS) for multi-center ASD diagnosis. Specifically, both the LOFC and HOFC features are first mapped to a high-dimensional space using the premise part of MO-TSK-FS, which shares the common knowledge across multiple centers. Then, the mapped features are transformed to a low-dimensional feature space using a transformation matrix. A novel unsupervised learning problem is formulated to find the optimal transformation matrix. Finally, a multi-modality support vector classifier (M2SVC) is constructed for classification. The experimental results show that the proposed interpretable feature learning method for multi-center ASD classification can effectively extract important features from the original LOFC and HOFC features, resulting in an efficient M2SVC for multi-center ASD classification.
KW - Autism
KW - Interpretable feature learning
KW - Manifold regularization
KW - Resting-state functional magnetic resonance imaging (rs-fMRI)
UR - http://www.scopus.com/inward/record.url?scp=85075641478&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-32248-9_88
DO - 10.1007/978-3-030-32248-9_88
M3 - Conference contribution
AN - SCOPUS:85075641478
SN - 9783030322472
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 790
EP - 798
BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2019 - 22nd International Conference, Proceedings
A2 - Shen, Dinggang
A2 - Yap, Pew-Thian
A2 - Liu, Tianming
A2 - Peters, Terry M.
A2 - Khan, Ali
A2 - Staib, Lawrence H.
A2 - Essert, Caroline
A2 - Zhou, Sean
PB - Springer Science and Business Media Deutschland GmbH
T2 - 22nd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2019
Y2 - 13 October 2019 through 17 October 2019
ER -