GeoMorph: Geometric Deep Learning for Cortical Surface Registration

Mohamed A. Suliman, Logan Z. J. Williams, Abdullah Fawaz, Emma. C Robinson
Proceedings of the First International Workshop on Geometric Deep Learning in Medical Image Analysis, PMLR 194:118-129, 2022.

Abstract

We present GeoMorph, a geometric deep learning image registration framework that takes two cortical surfaces on the spherical space and learns a smooth displacement field that aligns the features on the moving surface to those on the target. GeoMorph starts with feature extraction: independently extracting low-dimensional feature representations for each input surface using graph convolutions. These learned features are then registered in a deep-discrete manner by learning the optimal displacement for a set of control points that optimizes the overlap between features across the two surfaces. To ensure a smooth deformation, we propose a regularization network that considers the input sphere structure based on a deep conditional random field (CRF), implemented using a recurrent neural network (RNN). Results show that GeoMorph improves over existing deep learning methods by improving alignment whilst generating smoother and more biologically plausible deformations. Performance is competitive with classical frameworks, generalizing well even for subjects with atypical folding patterns.

Cite this Paper

BibTeX
@InProceedings{pmlr-v194-suliman22a, title = {GeoMorph: Geometric Deep Learning for Cortical Surface Registration}, author = {Suliman, Mohamed A. and Williams, Logan Z. J. and Fawaz, Abdullah and Robinson, Emma. C}, booktitle = {Proceedings of the First International Workshop on Geometric Deep Learning in Medical Image Analysis}, pages = {118--129}, year = {2022}, editor = {Bekkers, Erik and Wolterink, Jelmer M. and Aviles-Rivero, Angelica}, volume = {194}, series = {Proceedings of Machine Learning Research}, month = {18 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v194/suliman22a/suliman22a.pdf}, url = {https://proceedings.mlr.press/v194/suliman22a.html}, abstract = {We present GeoMorph, a geometric deep learning image registration framework that takes two cortical surfaces on the spherical space and learns a smooth displacement field that aligns the features on the moving surface to those on the target. GeoMorph starts with feature extraction: independently extracting low-dimensional feature representations for each input surface using graph convolutions. These learned features are then registered in a deep-discrete manner by learning the optimal displacement for a set of control points that optimizes the overlap between features across the two surfaces. To ensure a smooth deformation, we propose a regularization network that considers the input sphere structure based on a deep conditional random field (CRF), implemented using a recurrent neural network (RNN). Results show that GeoMorph improves over existing deep learning methods by improving alignment whilst generating smoother and more biologically plausible deformations. Performance is competitive with classical frameworks, generalizing well even for subjects with atypical folding patterns.} }
Endnote
%0 Conference Paper %T GeoMorph: Geometric Deep Learning for Cortical Surface Registration %A Mohamed A. Suliman %A Logan Z. J. Williams %A Abdullah Fawaz %A Emma. C Robinson %B Proceedings of the First International Workshop on Geometric Deep Learning in Medical Image Analysis %C Proceedings of Machine Learning Research %D 2022 %E Erik Bekkers %E Jelmer M. Wolterink %E Angelica Aviles-Rivero %F pmlr-v194-suliman22a %I PMLR %P 118--129 %U https://proceedings.mlr.press/v194/suliman22a.html %V 194 %X We present GeoMorph, a geometric deep learning image registration framework that takes two cortical surfaces on the spherical space and learns a smooth displacement field that aligns the features on the moving surface to those on the target. GeoMorph starts with feature extraction: independently extracting low-dimensional feature representations for each input surface using graph convolutions. These learned features are then registered in a deep-discrete manner by learning the optimal displacement for a set of control points that optimizes the overlap between features across the two surfaces. To ensure a smooth deformation, we propose a regularization network that considers the input sphere structure based on a deep conditional random field (CRF), implemented using a recurrent neural network (RNN). Results show that GeoMorph improves over existing deep learning methods by improving alignment whilst generating smoother and more biologically plausible deformations. Performance is competitive with classical frameworks, generalizing well even for subjects with atypical folding patterns.
APA
Suliman, M.A., Williams, L.Z.J., Fawaz, A. & Robinson, E.C.. (2022). GeoMorph: Geometric Deep Learning for Cortical Surface Registration. Proceedings of the First International Workshop on Geometric Deep Learning in Medical Image Analysis, in Proceedings of Machine Learning Research 194:118-129 Available from https://proceedings.mlr.press/v194/suliman22a.html.

Related Material