Abstract. In the context of image-guided TIPS (Transjugular, Intrahepatic, Portosystemic Shunt formation) surgery, the pose of a 3-dimensional model of the liver's vasculature is updated in real-time by tracking an ovoid balloon catheter. Therefore, it is critical to know the exact physical relationship between these objects prior to any real-time tracking. We have developed a new model-to-image registration technique which aligns a 3-dimensional model of the vasculature with two orthogonal fluoroscopic projections. Our technique is driven by the gradient information from the projection pairs at sample points along the centerlines of the model. Our algorithm shows speed, accuracy and consistency given clinical data.
Julien Jomier, Elizabeth Bullitt, Mark H. Van Horn