TY - GEN
T1 - A Generative Modeling Approach for Interpreting Population-Level Variability in Brain Structure
AU - Liu, Ran
AU - Subakan, Cem
AU - Balwani, Aishwarya H.
AU - Whitesell, Jennifer
AU - Harris, Julie
AU - Koyejo, Sanmi
AU - Dyer, Eva L.
N1 - Publisher Copyright:
© 2020, Springer Nature Switzerland AG.
PY - 2020
Y1 - 2020
N2 - Understanding how neural structure varies across individuals is critical for characterizing the effects of disease, learning, and aging on the brain. However, disentangling the different factors that give rise to individual variability is still an outstanding challenge. In this paper, we introduce a deep generative modeling approach to find different modes of variation across many individuals. Our approach starts with training a variational autoencoder on a collection of auto-fluorescence images from a little over 1,700 mouse brains at 25 μ m resolution. We then tap into the learned factors and validate the model’s expressiveness, via a novel bi-directional technique that makes structured perturbations to both, the high-dimensional inputs of the network, as well as the low-dimensional latent variables in its bottleneck. Our results demonstrate that through coupling generative modeling frameworks with structured perturbations, it is possible to probe the latent space of the generative model to provide insights into the representations of brain structure formed in deep networks.
AB - Understanding how neural structure varies across individuals is critical for characterizing the effects of disease, learning, and aging on the brain. However, disentangling the different factors that give rise to individual variability is still an outstanding challenge. In this paper, we introduce a deep generative modeling approach to find different modes of variation across many individuals. Our approach starts with training a variational autoencoder on a collection of auto-fluorescence images from a little over 1,700 mouse brains at 25 μ m resolution. We then tap into the learned factors and validate the model’s expressiveness, via a novel bi-directional technique that makes structured perturbations to both, the high-dimensional inputs of the network, as well as the low-dimensional latent variables in its bottleneck. Our results demonstrate that through coupling generative modeling frameworks with structured perturbations, it is possible to probe the latent space of the generative model to provide insights into the representations of brain structure formed in deep networks.
KW - Brain architecture and neuroanatomy
KW - Interpretable deep learning
KW - Variational autoencoder
UR - http://www.scopus.com/inward/record.url?scp=85092693169&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85092693169&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-59722-1_25
DO - 10.1007/978-3-030-59722-1_25
M3 - Conference contribution
AN - SCOPUS:85092693169
SN - 9783030597214
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 257
EP - 266
BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 - 23rd International Conference, Proceedings
A2 - Martel, Anne L.
A2 - Abolmaesumi, Purang
A2 - Stoyanov, Danail
A2 - Mateus, Diana
A2 - Zuluaga, Maria A.
A2 - Zhou, S. Kevin
A2 - Racoceanu, Daniel
A2 - Joskowicz, Leo
PB - Springer
T2 - 23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020
Y2 - 4 October 2020 through 8 October 2020
ER -