3D skull registration method based on contour curve feature / 中国法医学杂志

Objective Aiming at the 3D skull model with great difference in initial posture and resolution, a skull registration method based on feature contour line is proposed in the paper. Methods Firstly, contour lines which include eye, nose, temporal bone, maxilla and jawbone are extracted; Secondly, the types of contour are distinguished according to their length and shortest distance between them, and the corresponding relationship is established between the two skulls needed to register, thus the coarse registration of skulls is completed; Finally, an improved iterative closest point (ICP) algorithm is proposed by integrating weight into ICP, which is used to complete fine registration of skulls, thus the accurate registration of skulls is achieved at last. Results Registration between a unknown skull and 300 reference skulls is done, and the experimental results show that the proposed registration method could complete 3D skull registration, which could get high accuracy and speed in fine registration. Conclusion So the skull registration method based on local contour line is an accurate and fast 3D skull registration method.

Research on Registration Algorithm for Three-dimension Cardiac Electrical Maps / 航天医学与医学工程

Objective To align the mapping points from the locations of catheter electrodes with CT/MR image three-dimension data of the cardiac chamber. Methods Most of existing algorithms assumed that two spaces spanned respectively by mapping points and by CT/MR data were orthogonal, and thus modulated only 9 free-degrees of three-dimension translations, rotations and scaling for registration. However, our research showed that the above two spaces might be not always orthogonal. So we introduced affine transformation model based on 12 free-degrees and the corresponding iterative closest point algorithm to realize the registration and further incorporate the registration algorithm with the clinical operations. Combining with OpenGL, we realized the proposed algorithm via Visual C]+]+ 6.0 software. Results Simulation results showed that the residual registration errors on average were decreased from 40 pixels before registration to 1.5 pixels after registration. This algorithm was further validated in animal experiments. Conclusion This proposed three-dimension mapping registration algorithm well realizes the registration in three-dimension cardiac mapping and has practical application values.