A Domain Guided CNN Architecture for Predicting Age from Structural Brain Images

Pascal Sturmfels, Saige Rutherford, Mike Angstadt, Mark Peterson, Chandra Sripada, Jenna Wiens
Proceedings of the 3rd Machine Learning for Healthcare Conference, PMLR 85:295-311, 2018.

Abstract

Given the wide success of convolutional neural networks (CNNs) applied to natural images, researchers have begun to apply them to neuroimaging data. To date, however, exploration of novel CNN architectures tailored to neuroimaging data has been limited. Several recent works fail to leverage the 3D structure of the brain, instead treating the brain as a set of independent 2D slices. Approaches that do utilize 3D convolutions rely on architectures developed for object recognition tasks in natural 2D images. Such architectures make assumptions about the input that may not hold for neuroimaging. For example, existing architectures assume that patterns in the brain exhibit translation invariance. However, a pattern in the brain may have different meaning depending on where in the brain it is located. There is a need to explore novel architectures that are tailored to brain images. We present two simple modifications to existing CNN architectures based on brain image structure. Applied to the task of brain age prediction, our network achieves a mean absolute error (MAE) of 1.4 years and trains 30% faster than a CNN baseline that achieves a MAE of 1.6 years. Our results suggest that lessons learned from developing models on natural images may not directly transfer to neuroimaging tasks. Instead, there remains a large space of unexplored questions regarding model development in this area, whose answers may differ from conventional wisdom.

Cite this Paper

BibTeX
@InProceedings{pmlr-v85-sturmfels18a, title = {A Domain Guided CNN Architecture for Predicting Age from Structural Brain Images}, author = {Sturmfels, Pascal and Rutherford, Saige and Angstadt, Mike and Peterson, Mark and Sripada, Chandra and Wiens, Jenna}, booktitle = {Proceedings of the 3rd Machine Learning for Healthcare Conference}, pages = {295--311}, year = {2018}, editor = {Doshi-Velez, Finale and Fackler, Jim and Jung, Ken and Kale, David and Ranganath, Rajesh and Wallace, Byron and Wiens, Jenna}, volume = {85}, series = {Proceedings of Machine Learning Research}, month = {17--18 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v85/sturmfels18a/sturmfels18a.pdf}, url = {https://proceedings.mlr.press/v85/sturmfels18a.html}, abstract = {Given the wide success of convolutional neural networks (CNNs) applied to natural images, researchers have begun to apply them to neuroimaging data. To date, however, exploration of novel CNN architectures tailored to neuroimaging data has been limited. Several recent works fail to leverage the 3D structure of the brain, instead treating the brain as a set of independent 2D slices. Approaches that do utilize 3D convolutions rely on architectures developed for object recognition tasks in natural 2D images. Such architectures make assumptions about the input that may not hold for neuroimaging. For example, existing architectures assume that patterns in the brain exhibit translation invariance. However, a pattern in the brain may have different meaning depending on where in the brain it is located. There is a need to explore novel architectures that are tailored to brain images. We present two simple modifications to existing CNN architectures based on brain image structure. Applied to the task of brain age prediction, our network achieves a mean absolute error (MAE) of 1.4 years and trains 30% faster than a CNN baseline that achieves a MAE of 1.6 years. Our results suggest that lessons learned from developing models on natural images may not directly transfer to neuroimaging tasks. Instead, there remains a large space of unexplored questions regarding model development in this area, whose answers may differ from conventional wisdom.} }
Endnote
%0 Conference Paper %T A Domain Guided CNN Architecture for Predicting Age from Structural Brain Images %A Pascal Sturmfels %A Saige Rutherford %A Mike Angstadt %A Mark Peterson %A Chandra Sripada %A Jenna Wiens %B Proceedings of the 3rd Machine Learning for Healthcare Conference %C Proceedings of Machine Learning Research %D 2018 %E Finale Doshi-Velez %E Jim Fackler %E Ken Jung %E David Kale %E Rajesh Ranganath %E Byron Wallace %E Jenna Wiens %F pmlr-v85-sturmfels18a %I PMLR %P 295--311 %U https://proceedings.mlr.press/v85/sturmfels18a.html %V 85 %X Given the wide success of convolutional neural networks (CNNs) applied to natural images, researchers have begun to apply them to neuroimaging data. To date, however, exploration of novel CNN architectures tailored to neuroimaging data has been limited. Several recent works fail to leverage the 3D structure of the brain, instead treating the brain as a set of independent 2D slices. Approaches that do utilize 3D convolutions rely on architectures developed for object recognition tasks in natural 2D images. Such architectures make assumptions about the input that may not hold for neuroimaging. For example, existing architectures assume that patterns in the brain exhibit translation invariance. However, a pattern in the brain may have different meaning depending on where in the brain it is located. There is a need to explore novel architectures that are tailored to brain images. We present two simple modifications to existing CNN architectures based on brain image structure. Applied to the task of brain age prediction, our network achieves a mean absolute error (MAE) of 1.4 years and trains 30% faster than a CNN baseline that achieves a MAE of 1.6 years. Our results suggest that lessons learned from developing models on natural images may not directly transfer to neuroimaging tasks. Instead, there remains a large space of unexplored questions regarding model development in this area, whose answers may differ from conventional wisdom.
APA
Sturmfels, P., Rutherford, S., Angstadt, M., Peterson, M., Sripada, C. & Wiens, J.. (2018). A Domain Guided CNN Architecture for Predicting Age from Structural Brain Images. Proceedings of the 3rd Machine Learning for Healthcare Conference, in Proceedings of Machine Learning Research 85:295-311 Available from https://proceedings.mlr.press/v85/sturmfels18a.html.

Related Material