Global Optimization Surface-Based Registration for Image-to-Patient Registration Using Gaussian Mixture Model

Image-to-patient spatial registration plays a central role in image-guided neurosurgery systems (IGNS). Although the marker-based paired-point registration is widely used for image-to-patient registration, this method is impractical for clinical application. As an alternative, surface-matching registration based on the surface geometry of the face is being developed. In this paper, a global optimization surface-based method for image-to-patient registration using Gaussian Mixture Model (GMM) is presented. The key idea of this method is that each point set of the surface to be registered is considered as a whole, and point-to-point correlations in the overall space and subspace of the point set are-selected as the characteristics for the-registration process. The method was tested on 2D data sets, a 3D range scan face data set, and head phantom data for rigid point set registration. For the 2D rigid registration, the results obtained from the proposed method are compared to the results of two mainstream registration methods based on GMM from the impacts of the ini-tial position. The experimental results demonstrate that the proposed method has a good registration performance, is robust for the initial location on the registration results, and is easily implemented.