Robust and efficient 3D motion tracking in robotic assisted beating heart surgery

Abstract

This paper, presents a region-based motion tracking scheme to avoid problems caused by the complexity of the heart physical motion in assisted robotic minimally invasive beating heart surgery. Since, performance of region based methods strongly depend upon control points, selecting them in a region with sufficient texture is very important. Unlike previous methods, which control points are selected manually or placed on centers of grid, the novelty of our approach, is automatically selecting control points. By using automatic selection of the control points, performance and precision of motion tracking is improved and it leads to a robust tracking. In addition to increase accuracy and facilitate selection of control points, in proposed method, surgical instruments and specular reflection are detected and eliminated from the image. Finally, a modified version of 3D tracking algorithm based on a Thin-Plate Spline (TPS) deformable model is developed to improve the motion tracking. To evaluate the proposed algorithm, recorded stereo image sequence of an in-vivo data collected with DaVinci surgical robotic platform is fed to the algorithm.