Longitudinal Variational Autoencoders learn a Riemannian progression model for imaging data - Inserm - Institut national de la santé et de la recherche médicale Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Longitudinal Variational Autoencoders learn a Riemannian progression model for imaging data

Résumé

Interpretable progression models for longitudinal neuroimaging data are crucial to understanding neurodegenerative diseases. Well validated geometric progression models for biomarkers do not scale for such high dimensional data. In this work, we analyse a recent approach that combines a Variational Autoencoder with a latent linear mixed-effects model, and demonstrate that imposing a Euclidean prior on the latent space allows the network to learn the geometry of the observation manifold, and model non linear dynamics.
Fichier principal
Vignette du fichier
GeoMedia_2022___Geometric_insights (1).pdf (416.47 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03800499 , version 1 (06-10-2022)

Identifiants

  • HAL Id : hal-03800499 , version 1

Citer

Benoît Sauty, Stanley Durrleman​. Longitudinal Variational Autoencoders learn a Riemannian progression model for imaging data. GeoMedIA Workshop 2022 - Geometric Deep Learning in Medical Image Analysis, Nov 2022, Amsterdam, Netherlands. ⟨hal-03800499⟩
91 Consultations
135 Téléchargements

Partager

Gmail Facebook X LinkedIn More