Registration of CT-ultrasound images of the liver based on efficient vessel-filtering and automatic Initial transform prediction

Abstract

In this paper, we present a registration algorithm of CT and 3-D ultrasound liver images that are acquired from the same patient but at different times The algorithm automatically segments vessels from the ultrasound data, and predicts accurate initial registration parameters by examining the geometric correlation between CT and US vessel skeletons in a non-iterative manner. Experiments show that the proposed algorithm provides optimal affine parameters to compensate time-varying motion of the liver.