A pulmonary deformation registration framework for biplane X-ray and CT using Sparse Motion Composition

Lung cancer has the top fatality rate among all kinds of cancers. In clinical practice, CT guided nodule biopsy is an effective way to gain in vivo tissue samples for accurate diagnosis. However, the localization of lung lesions is very difficult due to severe lung respiration and tissues deformation. X-ray image benefits from less radiation and can be used to describe the deformation. In this paper, we propose a biplane X-ray images to pulmonary CT deformation registration framework, it uses Sparse Motion Composition (SMC) method to obtain an estimation of pulmonary motion which linearly combines the respiratory deformation vector field (DVF) of training samples, and then we use parametric control points on CT to refine the non-rigid pulmonary deformation. Finally, we get the accurate volumetric 3D registered image which is useful in intervention operations like image guided biopsy. The experimental results show that the proposed framework effectively uses the 2D X-ray images to accurately register the 3D pulmonary CT. The registration error is 2.83 (1.61) mm which is applicable for clinical practice.