A Fast CT and CT-Fluoroscopy Registration Algorithm With Respiratory Motion Compensation for Image-Guided Lung Intervention

Abstract

CT-fluoroscopy (CTF) is an efficient imaging technique for guiding percutaneous lung intervention such as biopsy and ablation. In CTF-guided procedures, four to ten axial images are captured in a very short time period during breath holding to provide near real-time feedback of patients' anatomy so that physicians can adjust the needle as it is advanced toward a target lesion. Although popularly used in clinics, this procedure requires frequent scans to guide the needle, which may cause increased procedure time, complication rates, and radiation exposure to both clinicians and patients. In addition, CTF only generates a limited number of 2-D axial images and does not provide sufficient 3-D anatomical information. Therefore, how to provide volumetric anatomical information using CTF while reducing intraoperative scan is an important and challenging problem. In this paper, we propose a fast CT-CTF deformable registration algorithm that warps the inhale preprocedural CT onto the intraprocedural CTF for guidance in 3-D. In the algorithm, the deformation in the transverse plane is modeled using 2-D B-Spline, and the deformation along z-direction is regularized by smoothness constraint. A respiratory motion compensation framework is also incorporated for accurate registration. A parallel implementation strategy is adopted to accomplish the registration in several seconds. With electromagnetic tracking, the needle position can be superimposed onto the deformed inhale CT image, thereby providing 3-D image guidance during breath holding. Experiments were conducted using both simulated CTF images with known deformation and real CTF images captured during lung cancer biopsy studies. The experiments demonstrated satisfactory registration results of our proposed fast CT-CTF registration algorithm.