1083-4435 (c) 2016 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/TMECH.2016.2583259, IEEE/ASME Transactions on Mechatronics Calibration and Integration of B-mode Optical Coherence Tomography for Assistive Control in Robotic Micro-Surgery Haoran Yu, Jin–Hui Shen, Karen M. Joos, Nabil Simaan † Abstract—During retinal microsurgery, surgeons cannot ad- equately visualize sub-surface anatomical structures. In our previous work, a customized B-mode optical coherence tomog- raphy (OCT) probe was integrated into an ophthalmic robotic system to provide depth perception. This paper presents new approaches for implementing and achieving real-time feedback and assistive robotic control based on B-mode OCT imaging. The robotic system was comprised of a parallel robot, a micro- injection tool, and a telemanipulation master interface. A method for calibrating the B-mode OCT image scaling and distortion was presented using thin plate splines. Determining the OCT scanning plane relative to the robot base frame is presented through experiments and analyzed for sensitivity. A dual-rate controller using low frequency OCT feedback and high frequency position servoing was presented and tested for accuracy and latency. 3D assistive telemanipulation virtual fixtures based on microscope and OCT feedback is presented. The experimental evaluation demonstrated following target anatomy and semi- automated micro injection. These results present the key steps towards achieving an integrated system for OCT feedback control using a miniature intraocular B-mode probe. I. I NTRODUCTION Ophthalmic micro-surgery requires surgeons to operate at extreme precision with hand-held instruments using only vi- sual feedback with a lack of information about sub-surface anatomy. Robotic-assistance has been proposed to overcome the limitations of precision and physiological tremor. Image- feedback to surgeons has also been proposed using optical coherence tomography (OCT). The integration of OCT image feedback to guide robots during telemanipulation control as- sistance is a new approach to combine the benefits of robotic precision and OCT perception of anatomical structures. While there have been studies dealing separately with robotics (e.g. [1]–[7]) or OCT guidance (e.g. [8], [9]), there is a sparsity of works presenting solutions enabling the combination of both technologies (e.g. [10], [11]). This paper aims to address a gap in knowledge on robotic image guidance and assistive control using intraocular B-mode OCT imaging. Despite a tremendous number of publications using OCT for imaging, there is a limited number using B-mode OCT for guiding surgical intervention. Simpler A-scan probes are being developed by several groups for ophthalmic surgery (e.g. [10], [12], [13]). However, A-scan OCT probes only provide This work was supported by Vanderbilt Discovery Grant 4229990095, by Unrestricted Grant to Vanderbilt Eye Institute from Research to Prevent Blindness, Inc., Ellis Family Research Fund, and The Jackson Family Donor Advised Fund of Waccamaw Community Foundation. Haoran Yu was partly supported by Vanderbilt Institute for Surgery and Engineering Fellowship. J. H. Shen and K. M. Joos are with Vanderbilt Eye Institute, Vanderbilt University, Nashville, TN 37232, USA. H. Yu and N. Simaan † (Corresponding author) are with the Department of Mechanical Engineering, Vanderbilt University, Nashville, TN 37212, USA. one dimensional depth information when the probe is fixed. By scanning the entire surface, these A-scan probes could be used to provide 3D scans of the retinal structures [14]. This movement constraint precludes the ability to simultaneously image attached tools and underlying tissue anatomy for the purpose of real-time image feed-back. Joos et al. presented in [15] a miniature intraocular B-mode OCT probe that provided a real-time cross-sectional images of the retina and the tool. This probe offers a unique opportunity for robotic assistance since it provides high-resolution cross-sectional images of the retina for a fixed probe location. This paper extends prior results by demonstrating robot- control algorithms for OCT guided surgery in assistive robot telemanipulation. In [11] preliminary results using OCT for guidance during membrane peeling, ocular and intra-ocular manipulations and stent deployment in phantoms were pre- sented. In [16] user performance was compared with or without robotic and/or OCT assistance. The following are the three key hurdles hindering the use of OCT probes for surgical guidance. First, a strategy for calibrating and registering the real-time B-mode OCT image to a known robot frame is needed. Second, a method for OCT feed-back and OCT-based assistive robot control needs to be developed. Finally, the robustness, accuracy and latency of OCT-guided robotic control needs to be validated. Works related to addressing these challenges include prior art on OCT image calibration algorithms (e.g. [17]–[22]) dealing with calibrating probe intrinsic parameters such as optical and fan distortion. However, the complex actuation to drive the imaging fiber inside miniature OCT probes produce nonlinear scanning artifacts. Thus; there is a need for a more general calibration method for miniature B-mode OCT scanning probes that are not limited to only calibrating the probe intrinsics. Thin plate splines (TPS) was used for ultrasound image dewarping [23]–[26]. To the best of authors knowledge, there have been no prior work dealing with OCT image dewarping using TPS. This paper presents an adaptation of calibration setup and an algorithm for B-mode OCT probe image calibration. Our approach produced interpolation map images to scale artifacts in the horizontal and vertical direc- tions. The purpose of OCT image dewarping is twofold: to enable size measurement from the OCT image and to improve visual servoing robustness. Virtual fixtures (VF) [27] have been used by several groups for vision-guided micro-manipulation and tremor filtering. Previous works [28]–[30] have demonstrated computer vision VF to improve the accuracy of ophthalmic interventions. However, the VF of these works was limited to assistive