Résumé
Whole-body anatomically correct high-resolution 3D medicalimages are instrumental for physical simulations. Unfortunately, onlya limited number of acquired datasets are available and the scope ofpossible applications is limited by the patient's posture. In this paper,we propose an extension of the interactive cage-based deformationpipeline VoxMorph [1], for labeled voxel grids allowing to eciently explorethe space of plausible poses while preserving the tissues' internalstructure. We propose 3 main contributions to overcome the limitationsof this pipeline: (i) we improve the robustness by proposing a deformationdiusion scheme, (ii) we improve the accuracy by proposing a newerror-metric for the renement process of the motion adaptive structure,(iii) we improve the scalability by proposing an out-of-core implementation.Our method is easy to use for novice users, robust and scales upto 3D images that do not t in memory, while oering limited distortionand mass loss. We evaluate our approach on postured whole-body segmentedimages and present an electro-magnetic wave exposure study forhuman-waves interaction simulations.