A class of novel point similarity measures based on MAP-MRF framework for 2D-3D registration of X-ray fluoroscopy to CT images

One of the main factors that affect the accuracy of 2D-3D registration of X-ray fluoroscopy to CT images is the similarity measure, which is a criterion function that is used in the registration procedure for measuring the quality of image match. Previously it was reported that pattern intensity was able to register accurately and robustly, even when soft tissues and interventional instruments were present in the fluoroscopy images. Unfortunately, pattern intensity was designed directly by using some heuristic rules. This paper presents a Markov random field (MRF) model for 2D-3D registration based on Bayesian framework. The optimal solution is defined as the maximum a posteriori (MAP) estimate of the MRF. By using this unified MAP-MRF framework, we derive a class of novel point similarity measures. We point out that pattern intensity is a member of this class. As examples, we present two novel point similarity measures of this class. Their behaviors are evaluated using a phantom and a human cadaveric spine specimen together with their respective ground truth and are compared to previously introduced similarity measures. We report their behavior comparison results and their registration accuracy