Deep Active Latent Surfaces for Medical Geometries

Patrick Møller Jensen, Udaranga Wickramasinghe, Anders Dahl, Pascal Fua, Vedrana Andersen Dahl
Proceedings of the 6th Northern Lights Deep Learning Conference (NLDL), PMLR 265:120-132, 2025.

Abstract

Shape priors have long been known to be effective when reconstructing 3D shapes from noisy or incomplete data. When using a deep-learning based shape representation, this often involves learning a latent representation, which can be either in the form of a single global vector or of multiple local ones. The latter allows more flexibility but is prone to overfitting. In this paper, we advocate a hybrid approach representing shapes in terms of 3D meshes with a separate latent vector at each vertex. During training the latent vectors are constrained to have the same value, which avoids overfitting. For inference, the latent vectors are updated independently while imposing spatial regularization constraints. We show that this gives us both flexibility and generalization capabilities, which we demonstrate on several medical image processing tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v265-jensen25a, title = {Deep Active Latent Surfaces for Medical Geometries}, author = {Jensen, Patrick M{\o}ller and Wickramasinghe, Udaranga and Dahl, Anders and Fua, Pascal and Dahl, Vedrana Andersen}, booktitle = {Proceedings of the 6th Northern Lights Deep Learning Conference (NLDL)}, pages = {120--132}, year = {2025}, editor = {Lutchyn, Tetiana and Ramírez Rivera, Adín and Ricaud, Benjamin}, volume = {265}, series = {Proceedings of Machine Learning Research}, month = {07--09 Jan}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v265/main/assets/jensen25a/jensen25a.pdf}, url = {https://proceedings.mlr.press/v265/jensen25a.html}, abstract = {Shape priors have long been known to be effective when reconstructing 3D shapes from noisy or incomplete data. When using a deep-learning based shape representation, this often involves learning a latent representation, which can be either in the form of a single global vector or of multiple local ones. The latter allows more flexibility but is prone to overfitting. In this paper, we advocate a hybrid approach representing shapes in terms of 3D meshes with a separate latent vector at each vertex. During training the latent vectors are constrained to have the same value, which avoids overfitting. For inference, the latent vectors are updated independently while imposing spatial regularization constraints. We show that this gives us both flexibility and generalization capabilities, which we demonstrate on several medical image processing tasks.} }
Endnote
%0 Conference Paper %T Deep Active Latent Surfaces for Medical Geometries %A Patrick Møller Jensen %A Udaranga Wickramasinghe %A Anders Dahl %A Pascal Fua %A Vedrana Andersen Dahl %B Proceedings of the 6th Northern Lights Deep Learning Conference (NLDL) %C Proceedings of Machine Learning Research %D 2025 %E Tetiana Lutchyn %E Adín Ramírez Rivera %E Benjamin Ricaud %F pmlr-v265-jensen25a %I PMLR %P 120--132 %U https://proceedings.mlr.press/v265/jensen25a.html %V 265 %X Shape priors have long been known to be effective when reconstructing 3D shapes from noisy or incomplete data. When using a deep-learning based shape representation, this often involves learning a latent representation, which can be either in the form of a single global vector or of multiple local ones. The latter allows more flexibility but is prone to overfitting. In this paper, we advocate a hybrid approach representing shapes in terms of 3D meshes with a separate latent vector at each vertex. During training the latent vectors are constrained to have the same value, which avoids overfitting. For inference, the latent vectors are updated independently while imposing spatial regularization constraints. We show that this gives us both flexibility and generalization capabilities, which we demonstrate on several medical image processing tasks.
APA
Jensen, P.M., Wickramasinghe, U., Dahl, A., Fua, P. & Dahl, V.A.. (2025). Deep Active Latent Surfaces for Medical Geometries. Proceedings of the 6th Northern Lights Deep Learning Conference (NLDL), in Proceedings of Machine Learning Research 265:120-132 Available from https://proceedings.mlr.press/v265/jensen25a.html.

Related Material