A Two-Stage Point Cloud Registration Method for Knee Joint Replacement Navigation

To resolve the problem that the iterative closest point (ICP) algorithm commonly used in knee joint replacement surgery registration had low registration accuracy and was easy to fall into local optimal solutions. An improved registration method based on feature point clouds was proposed. First, the Harris-3D algorithm was used to extract the feature point cloud of the three-dimensional model of the knee joint. Optical tracking system (OTS) was used to collect the point cloud of the corresponding area of the sawbone. Then, the Sample Consensus Initial Alignment (SAC-IA) algorithm was used to perform coarse registration on two point clouds. The ICP algorithm was used to make the registration matrix converge to an optimal solution. The k-d tree was used to find neighboring points to accelerate the iterative process. Finally, the simulation of the registration method was taken to prove the feasibility of the method. The accuracy of the knee joint surgery navigation registration experiment was carried out with the knee joint femur model as the object. Result shows that the error of the proposed registration algorithm is 2.12mm, while the error of the ICP algorithm is 6.30mm, which verifies the effectiveness of this method.