Multi-scale Adaptive Mask 3D Rigid Registration of Ultrasound and CT Images

Abstract

3D Rigid intra-operative image registration is an important technique to provide guidance for pre-operative information from different image modalities. Due to the artefacts that cause correspondence ambiguity, accurate registration of US with other image modalities such as computed tomography (CT) is still a challenging problem. We propose a method which considers the registration problem of US and CT images as a multi-scale regional information saliency and local similarity selection process. We design our method as a multi-stage approach in which global and local rigid registrations alternate in each stage. During the local registration, the US image regions with high feature saliency and similarity with CT image will be selected and joined as region masks for the next global registration, other parts will not be considered in order to remove the correspondence ambiguity. The performances of our method are compared with a typical intensity based registration method on phantom and real patient images.