Registration of Lung CT Images Using B-spline Based Free-Form Deformation

Abstract

Image registration of lung CT images plays a crucial role in estimating lung motion that is of great significance in image guided radiation therapy. However, the respiratory-induced deformation makes registration of lung CT images rather difficult and challenging. In this study, we present a deformable image registration framework in which the B-spline method is taken for transformation of registration and the compressible flow theory with L2 norm of displacement vectors is used for similarity measure. Results show that our method obtains more accurate registration of lung CT images than the often used SSD, CC and some other state-of-the-art methods.