A method of 3D-3D multi-stage non-rigid registration of the spine based on binocular structured light

Intraoperative deformation and radiation are common problems in spinal surgery. A three-dimensional multi-stage dynamic iterative non-rigid registration method of the spine based on binocular structured light is proposed in this paper to overcome these problems.

The problem of intraoperative radiation in traditional X-ray and CT is overcome by using binocular structured light. A three-dimensional spinal mask based on binary code is designed to reduce the influence of non-interested regions on the operation. Principal component analysis (PCA) algorithm is used to complete the rough registration between the preoperative CT model of the spine and the reconstructed surface of the intraoperative structured light. A new framework of multi-stage dynamic iterative non-rigid registration of the spine is proposed. The Iterative Closest Point (ICP) algorithm based on bidirectional selection is proposed to complete the single-stage registration of the spine. Then the multi-stage dynamic iterative registration of the spine is completed to solve the problem of large registration error caused by the deformation of the spine.

The method proposed in this paper is compared with traditional registration methods, and its application is verified experimentally. The results show that the registration accuracy and time of the proposed method are $$  mm and $$  s, respectively. The accuracy of the method is 81.5% and 78.2% higher than that of the contour method and the method of marker points, respectively.

The method can effectively avoid intraoperative radiation, reduce the registration error caused by the deformation of the spine, and has a high practicability.