A maximum likelihood approach for 2-D to 3-D registration

Abstract

In this paper, we examine the problem of point-based registration in which a pair of two-dimensional projective views of an object is used to reconstruct the three-dimensional position of given points within the object. The particular application we consider is that of integrating preoperative, three-dimensional information, for example that provided by a CT scan, with the two-dimensional information given by intraoperative fluoroscopy. To facilitate registration, before the preoperative scan we embed small, removable spheres in the patient’s anatomy proximal to the site of proposed surgery; the spheres remain in place until the image-guided phase of the operation is complete.