Guided Filter Regularization for Improved Disentanglement of Shape and Appearance in Diffeomorphic Autoencoders

Hristina Uzunova, Heinz Handels, Jan Ehrhardt
Proceedings of the Fourth Conference on Medical Imaging with Deep Learning, PMLR 143:774-786, 2021.

Abstract

Diffeomorphic and deforming autoencoders have been recently explored in the field of medical imaging for appearance and shape disentanglement. Both models are based on the deformable template paradigm, however they show different weaknesses for the representation of medical images. Diffeomorphic autoencoders only consider spatial deformations, whereas deforming autoencoders also regard changes in the appearance, however no uniform template is generated for the whole training dataset, and the appearance is modeled depending on a very few parameters. In this work, we propose a method that represents images based on a global template, where next to the spatial displacement, the appearance is modeled as the pixel-wise intensity difference to the unified template. To however ensure that the generated appearance offsets adhere to the template shape, a guided filter smoothing of the appearance map is integrated into an end-to-end training process. This regularization significantly improves the disentanglement of shape and appearance and thus enables multi-modal image modeling. Furthermore, the generated templates are crisper and the registration accuracy improves. Our experiments also show applications of the proposed approach in the field of automatic population analysis.

Cite this Paper

BibTeX
@InProceedings{pmlr-v143-uzunova21a, title = {Guided Filter Regularization for Improved Disentanglement of Shape and Appearance in Diffeomorphic Autoencoders}, author = {Uzunova, Hristina and Handels, Heinz and Ehrhardt, Jan}, booktitle = {Proceedings of the Fourth Conference on Medical Imaging with Deep Learning}, pages = {774--786}, year = {2021}, editor = {Heinrich, Mattias and Dou, Qi and de Bruijne, Marleen and Lellmann, Jan and Schläfer, Alexander and Ernst, Floris}, volume = {143}, series = {Proceedings of Machine Learning Research}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v143/uzunova21a/uzunova21a.pdf}, url = {https://proceedings.mlr.press/v143/uzunova21a.html}, abstract = {Diffeomorphic and deforming autoencoders have been recently explored in the field of medical imaging for appearance and shape disentanglement. Both models are based on the deformable template paradigm, however they show different weaknesses for the representation of medical images. Diffeomorphic autoencoders only consider spatial deformations, whereas deforming autoencoders also regard changes in the appearance, however no uniform template is generated for the whole training dataset, and the appearance is modeled depending on a very few parameters. In this work, we propose a method that represents images based on a global template, where next to the spatial displacement, the appearance is modeled as the pixel-wise intensity difference to the unified template. To however ensure that the generated appearance offsets adhere to the template shape, a guided filter smoothing of the appearance map is integrated into an end-to-end training process. This regularization significantly improves the disentanglement of shape and appearance and thus enables multi-modal image modeling. Furthermore, the generated templates are crisper and the registration accuracy improves. Our experiments also show applications of the proposed approach in the field of automatic population analysis.} }
Endnote
%0 Conference Paper %T Guided Filter Regularization for Improved Disentanglement of Shape and Appearance in Diffeomorphic Autoencoders %A Hristina Uzunova %A Heinz Handels %A Jan Ehrhardt %B Proceedings of the Fourth Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2021 %E Mattias Heinrich %E Qi Dou %E Marleen de Bruijne %E Jan Lellmann %E Alexander Schläfer %E Floris Ernst %F pmlr-v143-uzunova21a %I PMLR %P 774--786 %U https://proceedings.mlr.press/v143/uzunova21a.html %V 143 %X Diffeomorphic and deforming autoencoders have been recently explored in the field of medical imaging for appearance and shape disentanglement. Both models are based on the deformable template paradigm, however they show different weaknesses for the representation of medical images. Diffeomorphic autoencoders only consider spatial deformations, whereas deforming autoencoders also regard changes in the appearance, however no uniform template is generated for the whole training dataset, and the appearance is modeled depending on a very few parameters. In this work, we propose a method that represents images based on a global template, where next to the spatial displacement, the appearance is modeled as the pixel-wise intensity difference to the unified template. To however ensure that the generated appearance offsets adhere to the template shape, a guided filter smoothing of the appearance map is integrated into an end-to-end training process. This regularization significantly improves the disentanglement of shape and appearance and thus enables multi-modal image modeling. Furthermore, the generated templates are crisper and the registration accuracy improves. Our experiments also show applications of the proposed approach in the field of automatic population analysis.
APA
Uzunova, H., Handels, H. & Ehrhardt, J.. (2021). Guided Filter Regularization for Improved Disentanglement of Shape and Appearance in Diffeomorphic Autoencoders. Proceedings of the Fourth Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 143:774-786 Available from https://proceedings.mlr.press/v143/uzunova21a.html.

Related Material