A modular multi-scale model to build atlases of shapes with large deformations

Barbara Gris (1, 2) , Stanley Durrleman (2) , Alain Trouvé (1)

1 CMLA - Centre de Mathématiques et de Leurs Applications
2 ARAMIS - Algorithms, models and methods for images and signals of the human brain = Algorithmes, modèles et méthodes pour les images et les signaux du cerveau humain [ICM Paris]

A central aspect of Computational Anatomy is the description of the variations among a population of shapes. A common approach to study them is the computation of an atlas, namely an average template shape and a family of template-to-subject deformations. Large deformations are often obtained through the integration of a vector field, deformation models vary therefore according to the model of vector fields. We present here a new model for large deformations in which vector fields are modular and multi-scale, which favours sparsity of parametrisation and its interpretability.

Domaines: Mathématiques [math]; Sciences cognitives

Barbara Gris, Stanley Durrleman, Alain Trouvé. A modular multi-scale model to build atlases of shapes with large deformations. Organization for Human Brain Mapping 2015, Jun 2015, Honolulu, United States. pp.1. ⟨hal-01191547⟩