Evaluation of Similarity Measures and Optimization Methods for 2D/3D Rigid Registration in Image Guided Interventions

In image-guide intervention, 2D/3D image registration provides the surgeon with the precise positions of patient. Successful registration involves many aspects, such as similarity measures and optimization methods, and they had the different influence on the registration results. In order to investigate the respective influence of similarity measures and optimization methods in 2D/3D image registration, a comparison was performed by combining five similarity measures with four optimization methods and using the publicly available porcine skull phantom datasets from Medical University Vienna. The experiments were evaluated based on success rate and registration time. The results showed that the best registration was achieved with pattern intensity (PI) combined with Powell and the best search strategy for 2D/3D rigid registration was Powell.