A Filtering Approach for Surgical Registration with Unknown Stiffness

Abstract

Snake-like robots have the potential to reach deeper within the anatomy while reducing patient trauma. To help guide these tools, model-based image-guidance can be used to fuse live tracking data with preoperative imaging in order to visualize a robot’s pose relative to the anatomy. In this paper, we introduce a new probabilistic filtering algorithm for registering live tracking data with a preoperative surface model while
simultaneously estimating the surface compliance. Registration for medical imaging is often performed using the iterative closest point method (ICP) [1] or with a number of deformable registration methods that account for organ deformation [2]. Our probabilistic approach overcomes the delay required for offline batch processing by estimating registration parameters online with an iterative extended Kalman filter (IEKF). Existing methods for in-vivo registration use line-of-sight measurements (e.g., 3D vision and range scanning). Our approach alternatively uses palpation to sense organ shape and compliance (e.g,. [3], [4]). In this paper, we extend our prior work [5] by relaxing the assumption that the organ stiffness is known. We instead simultaneously estimate organ compliance along with parameters defining the registration of the robot through the use of contact force measurements.