Pericardium based model fusion of CT and non-contrasted C-arm CT for visual guidance in cardiac interventions.

Minimally invasive transcatheter cardiac interventions are being adopted rapidly to treat a range of cardiovascular diseases. Pre-operative imaging, e.g., computed tomography (CT), plays an important role in surgical planning and simulation of cardiac interventions. Overlaying a 3D cardiac model extracted from pre-operative images onto real-time fluoroscopic images provides valuable visual guidance during the intervention. However, direct 3D to 2D fusion is difficult and may require quite amounts of user interaction. Intra-operative non-contrasted C-arm CT can be used as an intermedium for model fusion. The cardiac model is first warped to C-arm CT and later overlaid onto fluoroscopy. The C-arm CT to fluoroscopy overlay is straightforward since both images are captured on the same machine and the C-arm projection geometry can be directly used for overlay. Though various image registration methods may be used to fuse pre-operative images and C-arm CT, cross-modality image registration is not robust due to the significant difference in image characteristics (contrasted vs. non-contrasted). In this work we propose a model based fusion method using the pericardium to align pre-operative CT to intra-operative C-arm CT. After automatic segmentation of the pericardium in both CT and C-arm CT, the deformation field is estimated and then applied to warp the cardiac model extracted from CT to C-arm CT. The proposed method can be applied to fuse different cardiac models (e.g., chambers, aorta, coronary arteries, and cardiac valves). A feasibility study on aortic root model fusion shows that a reasonable accuracy can be achieved using a generic model (from a different patient), while more accurate results come from a patient-specific model. Intelligently weighted fusion can further improve the accuracy by using all available cardiac models in a pre-collected training set.