Automated Global Optimization Surface-Matching Registration Method for Image-to-Patient Spatial Registration in an Image-Guided Neurosurgery System

摘要

Image-to-patient spatial registration is the most basic yet simultaneously the most critical technology in the image-guided neurosurgery system (IGNS), particularly as it directly impacts the system's accuracy. Because of the drawbacks associated with the marker-based paired-point registration as well as its impracticability in clinical application, image-to-patient spatial registration based on the surface-matching method garners a good deal of attention. Therefore, in this paper, we propose a novel surface registration approach for the image-to-patient registration in such challenging scenarios. We divide the registration process into a coarse registration and a fine registration. The coarse registration method is based on an improved 4PCS algorithm and is for improving the registration speed as well as dealing with the problem of local minimum in the iterative process. The fine registration method is based on the Iterative Closest Point (ICP) algorithm, which can achieve a high registration accuracy when a good initialization is provided. We then demonstrate the proposed method's effectiveness by performing several experiments on the cranium and on real patient CT images. From the experiment results, it is demonstrated that the proposed method is a highly-precise, fully automatic, and robust surface-matching registration method for the image-to-patient registration suitable for IGNS.