An improved CT to fluoroscopy registration algorithm for the kinematic analysis of knee joints

Understanding the kinematics of the knee joint as it undergoes normal physical activities is an important goal of the orthopaedic research community. The limitations of previous approaches to capturing knee kinematics include the requirement for tantalum beads to be implanted prior to imaging and the use of biplanar X-ray imaging which is not commonly available in hospital imaging departments. A recently proposed alternative approach is to register 3D CT data with individual 2D video frames of the knee captured using single-plane X-ray fluoroscopy. The main limitation of this approach has been the inaccuracy of the registration algorithm for out-of-plane translation. In this paper we propose an improved registration algorithm which uses the relatively new cross-cumulative residual entropy similarity measure. Experimental results show that the accuracy of our proposed technique is superior to previous approaches particularly for out-of-plane translation.