Registration of PET and MR Hand Volumes Using Bayesian Networks

Abstract. A method for the non-rigid, multi-modal, registration of volumetric scans of human hands is presented. PET and MR scans are aligned by optimising the configuration of a tube based model using a set of Bayesian networks. Efficient optimisation is performed by posing the problem as a multi-scale, local, discrete (quantised) search, and using dynamic programming. The method is to be used within a project to study the use of high-resolution HIDAC PET imagery in investigating bone growth and erosion in arthritis.

• Type: Conference
• Year: 2005
• Where: CVBIA
• Authors: Derek R. Magee, Steven Tanner, Michael Waller, Dennis McGonagle, Alan P. Jeavons
• Keywords: Computer Vision, CVBIA 2005, MR Scans, Tube Based Model, Volumetric Scans