Comparing algorithms for diffeomorphic registration: Stationary LDDMM and Diffeomorphic Demons

Abstract : The stationary parameterization of diffeomorphisms is be- ing increasingly used in computational anatomy. In certain applications it provides similar results to the non-stationary parameterization alle- viating the computational charge. With this characterization for diffeo- morphisms, two different registration algorithms have been recently pro- posed: stationary LDDMM and diffeomorphic Demons. To our knowl- edge, their theoretical and practical differences have not been analyzed yet. In this article we provide a comparison between both algorithms in a common framework. To this end, we have studied the differences in the elements of both registration scenarios. We have analyzed the sen- sitivity of the regularization parameters in the smoothness of the final transformations and compared the performance of the registration re- sults. Moreover, we have studied the potential of both algorithms for the computation of essential operations for further statistical analysis. We have found that both methods have comparable performance in terms of image matching although the transformations are qualitatively different in some cases. Diffeomorphic Demons shows a slight advantage in terms of computational time. However, it does not provide as stationary LD- DMM the vector field in the tangent space needed to compute statistics or exact inverse transformations.
Type de document :
Communication dans un congrès
Xavier Pennec. 2nd MICCAI Workshop on Mathematical Foundations of Computational Anatomy, Oct 2008, New-York, United States. pp.24-35, 2008
Liste complète des métadonnées

https://hal.inria.fr/inria-00629883
Contributeur : Alain Monteil <>
Soumis le : jeudi 6 octobre 2011 - 17:37:29
Dernière modification le : jeudi 11 janvier 2018 - 16:43:45
Document(s) archivé(s) le : mardi 13 novembre 2012 - 15:20:26

Fichier

mfca08_1_3.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : inria-00629883, version 1

Collections

Citation

Monica Hernandez, Salvador Olmos, Xavier Pennec. Comparing algorithms for diffeomorphic registration: Stationary LDDMM and Diffeomorphic Demons. Xavier Pennec. 2nd MICCAI Workshop on Mathematical Foundations of Computational Anatomy, Oct 2008, New-York, United States. pp.24-35, 2008. 〈inria-00629883〉

Partager

Métriques

Consultations de la notice

456

Téléchargements de fichiers

246