Exploring Deep Registration Latent Spaces - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2021

Exploring Deep Registration Latent Spaces

Marie-Pierre Revel

Abstract

Explainability of deep neural networks is one of the most challenging and interesting problems in the field. In this study, we investigate the topic focusing on the interpretability of deep learning-based registration methods. In particular, with the appropriate model architecture and using a simple linear projection, we decompose the encoding space, generating a new basis, and we empirically show that this basis captures various decomposed anatomically aware geometrical transformations. We perform experiments using two different datasets focusing on lungs and hippocampus MRI. We show that such an approach can decompose the highly convoluted latent spaces of registration pipelines in an orthogonal space with several interesting properties. We hope that this work could shed some light on a better understanding of deep learningbased registration methods.
Fichier principal
Vignette du fichier
arxiv_submission.pdf (3.69 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-03524105 , version 1 (13-01-2022)

Identifiers

Cite

Théo Estienne, Maria Vakalopoulou, Stergios Christodoulidis, Enzo Battistella, Théophraste Henry, et al.. Exploring Deep Registration Latent Spaces. DART in MICCAI 2021, the 24th International Conference on Medical Image Computing and Computer Assisted Intervention, Sep 2021, Strasbourg, France. pp.112-122, ⟨10.1007/978-3-030-87722-4_11⟩. ⟨hal-03524105⟩
81 View
130 Download

Altmetric

Share

Gmail Mastodon Facebook X LinkedIn More