Markov Random Field Optimization for Intensity-based 2D-3D Registration

Darko Zikic (1) , Ben Glocker (1) , Oliver Kutter (1) , Martin Groher (2) , Nikos Komodakis (3, 4) , Ali Khamene (5) , Nikos Paragios (4, 6, 7, 8, 9, 10) , Nassir Navab (1)

1 CAMPAR - Computer Aided Medical Procedures & Augmented Reality
2 CAMP
3 CVN - Centre de vision numérique
4 GALEN - Organ Modeling through Extraction, Representation and Understanding of Medical Image Content
5 Siemens Corporate Research
6 MAS - Mathématiques Appliquées aux Systèmes - EA 4037
7 imagine [Marne-la-Vallée]
8 LIGM - Laboratoire d'Informatique Gaspard-Monge
9 UPEM - Université Paris-Est Marne-la-Vallée
10 ENPC - École des Ponts ParisTech

We propose a Markov Random Field (MRF) formulation for the intensity-based N-view 2D-3D registration problem. The transformation aligning the 3D volume to the 2D views is estimated by iterative updates obtained by discrete optimization of the proposed MRF model. We employ a pairwise MRF model with a fully connected graph in which the nodes represent the parameter updates and the edges encode the image similarity costs resulting from variations of the values of adjacent nodes. A label space refinement strategy is employed to achieve sub-millimeter accuracy. The evaluation on real and synthetic data and comparison to state-of-the-art method demonstrates the potential of our approach.

Domaines: Imagerie médicale

Darko Zikic, Ben Glocker, Oliver Kutter, Martin Groher, Nikos Komodakis, et al.. Markov Random Field Optimization for Intensity-based 2D-3D Registration. SPIE 2010, Medical Imaging 2010: Image Processing, 2010, San Diego, United States. ⟨10.1117/12.837232⟩. ⟨hal-00858402⟩