Target Registration Error in Point-based Rigid-body Registration: a Worst-case Analysis

The model has to be evaluated in terms of accuracy, robustness and speed. In a first evaluation, sequences of anatomic structures were used. Accuracy and robustness were evaluated by registering reconstructed point clouds with a polygon model from a laser scanner. The first results indicated that the surface model is reconstructed with millimeter precision. An important criterion for the usability of the surface model is its real-time capability because it has to handle live intraoperative video streams. When we only reconstruct the point cloud without meshing it, we achieve 20–30 fps on our test system with a resolution of 320 9 240. When point cloud meshing is performed for every fourth point (about 10,000 reconstructed points) by the fast meshing method and the mesh is rendered with VTK, we achieve about 17–20 fps. With the standard triangulation from VTK we only achieve about 4 fps. When meshing is performed for every reconstructed point (about 40,000), the speed drops to 10–14 fps (compared to 0.8 fps from VTK). The mesh quality is only slightly worse at the edge of the image and in areas with erroneous points that are not reconstructed. Conclusion First results promise that the surface model is accurate, robust and fast enough so that it can be used intraoperatively to register it with a preoperative soft tissue model. These results have to be verified in a next evaluation step by registering models of silicone phantoms with the corresponding surface model and, later, in a real surgical environment. Additional optimization is necessary to ensure the practicability of this method. Further research activities focus on 3D registration with a volume model of the surgical site so that this model is updated and adjusted. The next step is the integration of a forcetorque sensor in a laparoscopic instrument to add more information in the registration process. Secondly, an appropriate volume model has to be developed. References 1. Nicole Atzpadin, Peter Kauff, and Oliver Schreer. Stereo analysis by hybrid recursive matching for real-time immersive video conferencing. IEEE Transactions on Circuits and Systems for Video Technology, 14(3), März 2004. 2. S. Röhl, S. Speidel, G. Sudra, B. Müller-Stich, C. Gutt, and R. Dillmann. Dynamic 3d-reconstruction of endoscopic image sequences. 2008. 3. W.J. Schroeder, K.W. Martin, and B. Lorensen. The visualization toolkit. Kitware, Inc.