Multi-View Global 2D-3D Registration Based on Branch and Bound Algorithm

Abstract

In image-guided minimally invasive surgery, the clinician relies on image guidance to observe, plan and navigate. In order to show invisible vessels or planning annotations in live X-ray images, or update the live information for planning, the correspondence establishment between 2D X-ray images and pre-operatively acquired 3D CT images is a fundamental step. Accurate image alignment is needed and can be provided by 2D-3D registration by bringing a pre-operative 3D image and intra-operative 2D image into the same coordinate. In this work, we propose a novel method to register 3D volume with multi-view 2D images. Based on the Branch and Bound algorithm, the method can globally search the optimal pose and find the correspondence. Compared to the single-view setting, the multi-view 2D image information can speed up the searching. Extensive experiments are conducted to evaluate the effectiveness of the proposed method. The accuracy is improved and the iteration numbers are reduced with the introduction of new views.