2377-3766 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/LRA.2018.2801481, IEEE Robotics and Automation Letters IEEE ROBOTICS AND AUTOMATION LETTERS. PREPRINT VERSION. ACCEPTED JANUARY, 2018 1 Preliminary Evaluation of an Online Estimation Method for Organ Geometry and Tissue Stiffness Preetham Chalasani 1 , Long Wang 2 , Rashid Yasin 2 Nabil Simaan 2 and Russell H. Taylor 1 Abstract—During open surgeries, surgeons use tactile palpa- tion to form an understanding of the anatomy, including any underlying anatomical structures such as arteries or tumors. This paper explores methods for restoring this capability in the context of minimally invasive robot-assisted surgery. Previous works have demonstrated the ability of robots to use discrete palpation to characterize organ shape and, when followed with offline data processing, to produce a stiffness map of the organ. In our earlier work, we presented an offline estimation technique, independent of palpation strategy, to estimate organ shape and stiffness using Gaussian Processes (GPs). This study extends our prior work by demonstrating a fast online technique for estimation of organ shape and stiffness. Our goal is to provide near video-frame-rate updates of the organ geometry and tis- sue stiffness during force controlled exploration. Two different palpation modes are experimentally explored: a) autonomous palpation, b) constrained semi-autonomous teleoperation. We report the experimental evaluation of our approach for stiffness estimation using autonomous palpation. And we demonstrate the feasibility of using our method during interactive teleoperation in a simulated surgical scenario. We believe that future use of the online stiffness and geometry information based on our proposed method can help surgeons in improving their understanding of the surgical scene and its correlation to pre-operative imaging - thereby increasing safety and improving surgical outcomes. Index Terms—Medical Robots and Systems; Surgical Robotics: Laparoscopy; Telerobotics and Teleoperation; Force and Tactile Sensing; Control Architectures and Programming I. INTRODUCTION R EALTIME access to mechanical properties of anatomy helps surgeons in identifying subsurface anatomy in- cluding tumors, nodules, nerve bundles and tumors. To better understand the anatomy, surgeons typically use their fingers to palpate the tissue, in a sweeping or circular motion [1]. Some studies ( [2], [3]) show that tactile feedback is greatly diminished in minimally invasive surgery (MIS), due to lack of direct contact. Surgical assistants like Intuitive Surgical’s da Vinci system [4] provide increased dexterity and control to the surgeons during precise MIS procedures ( [5], [6]). With the Manuscript received: September, 10, 2017; Revised December, 20, 2017; Accepted January, 17, 2018. This paper was recommended for publication by Editor Rocco, Paolo upon evaluation of the Associate Editor and Reviewers’ comments. *This work was supported in part by NRI grant IIS-1327657, IIS-1327566, and in part by Johns Hopkins University and Vanderbilt University internal funds 1 Preetham Chalasani and Russell H. Taylor are with the Department of Computer Science, Johns Hopkins University, Baltimore, MD 21218. rht@jhu.edu 2 Long Wang, Rashid Yasin and Nabil Simaan are with the Department of Mechanical Engineering, Vanderbilt University, Nashville, TN 37212. nabil.simaan@vanderbilt.edu Digital Object Identifier (DOI): see top of this page. increase in robot assisted minimally invasive surgery, there have been advances in providing haptic feedback capability to the surgeon [7]. A number of finger-like tactile and force sensors have been developed to provide tool-to-tissue inter- action forces [8]–[15]. Many groups have also sensorized the surgical robotic instruments for force sensing (e.g. [16]–[20]). Using these force-sensing instruments, tissue stiffness can be estimated by measuring force changes as tools are pressed into the tissue. Mavash et al. [21] reported a method for results of stiffness estimation of a tissue model based on discrete palpation. They developed a control system for the da Vinci surgical system that provided force feedback with a position-position controller with friction and inertia compensation. In more recent work, methods to estimate forces acting at the tip of continuum robots were developed using joint-level actuation forces [22], [23]. These methods have been used to enable force-controlled exploration of flexible anatomy [24]. Surface information is useful in registration of preoperative data and in providing “augmented reality” information support to the surgeon. Many authors have reported computer vision methods for estimating surface geometry e.g. [25]–[28]. Dis- crete probing can be used to obtain additional information at large number of points, which can be used to correct for camera-to-robot misalignment. While discrete probing can generate a model of tissue stiffness across an organ, the process can be very time consuming.. In an effort to improve efficiency of probing for detecting tumors, Ayvali et al. [29] explored Bayesian optimization strategies to guide the probe towards unexplored regions that would result in maximum information gain in predicting stiff regions. Caccamo et al. [30] developed an online probabilistic frame- work for autonomous estimation of a deformability distribution map of heterogeneous elastic surfaces from few physical interactions. Caccamo’s framework combines both visual and haptic measurements with active exploration and builds de- formability maps. There are other vision based algorithms available for online surface reconstruction [31]. In more recent work, Garg et al. [32] have presented an autonomous tumor localization technique using GP adaptive sampling. They use a palpation probe to estimate surface stiffness using discrete probing and an implicit level-set upper confidence bound (ILS-UCB) algorithm for an offline estima- tion of tumor boundary. All the aforementioned works on stiffness and surface esti- mation depend on discrete probing strategies. In our previous work [33], we demonstrated an offline estimation technique