Hierarchical Multi-Scale Supervoxel Matching using Random Forests for Automatic Semi-Dense Abdominal Image Registration

Pierre-Henri Conze (1) , Florian Tilquin (1) , Vincent Noblet (1) , François Rousseau (2, 3) , Fabrice Heitz (1) , Patrick Pessaux (4)

1 ICube - Laboratoire des sciences de l'ingénieur, de l'informatique et de l'imagerie
2 IMT Atlantique - ITI - Département lmage et Traitement Information
3 LaTIM - Laboratoire de Traitement de l'Information Medicale
4 IRCAD/EITS - Institut de Recherche Contre les Cancers de l'Appareil Digestif-European Institute of Telesurgery

This paper addresses the estimation of pairwise supervoxel correspondences toward automatic semi-dense medical image registration. Supervoxel matching is performed through random forests (RF) with supervoxel indexes as label entities to predict matching areas in another target image. Ensuring accurate supervoxel boundary adherence requires a fine super-voxel decomposition which highly increases learning complexity. To alleviate this issue, we extend RF based super-voxel matching from single to multi-scale using a recursive hierarchical supervoxel representation. Output RF matching probabilities obtained for the last scale are gathered with ancestor matching probabilities which acts as a coarse-to-fine matching guidance. The effectiveness of our method is highlighted for semi-dense abdominal image registration relying on liver label propagation and consistency assessment.

Mots clés: Index Terms— semi-dense image registration; super- voxel matching; random forests; hierarchical multi-scale

Domaines: Traitement des images [eess.IV]

Pierre-Henri Conze, Florian Tilquin, Vincent Noblet, François Rousseau, Fabrice Heitz, et al.. Hierarchical Multi-Scale Supervoxel Matching using Random Forests for Automatic Semi-Dense Abdominal Image Registration. ISBI 2017 : IEEE 14th International Symphosium on Biomedical Imaging, Apr 2017, Melbourne, Australia. ⟨10.1109/ISBI.2017.7950567⟩. ⟨hal-01465421⟩