Telemanipulation for Micromanipulation and Microsurgery

Retinal Microsurgery

Retinal microsurgery procedures are at the limits of human ability. An error of few micrometers can cause the instrument to exert damaging force on the retina, causing loss of vision at the spot due to hemorrhage. Tactile feedback is known to improve manipulation effciency, but the forces experienced during retinal surgeries are far below the sensable limit for humans. Hence, the surgeon has to rely on visual feedback only. To further complicate the problem, patient movement has to be accounted for by the surgeon as movement due to patient breathing is a common problem.

Retinal surgery begins with a vitrectomy (the removal of the vitreous), which relieves the traction on the retina and enables surgical instruments to gain access to the retina. Three trocars through the sclera are prepared, one each for a fiber optic light source, a suction needle, and for the different instruments that will be used during the surgery. Using instruments inserted through the trocar, tasks such as epiretinal-membrane peeling are performed using a sequence of abrasive, cutting, and pulling manipulations.

Membrane peeling surgery being performed by instruments inserted in the eye.

⚠ (:youtube oRQ4zeLFVaE loop=1 fs=1 scale=1:)
This video is a short oral talk of our paper presented at MICCAI 2015 (see Selected Publications). This paper provides a description of our retinal microsurgery system, and presents results from preliminary telemanipulation experiments.

⚠ (:youtube baTfdqTIGlI loop=1 fs=1 scale=1:)
This video shows the retinal microsurgery manipulator being telemanipulated using a Geomagic Touch. An Alcon microforceps with a soft actuator is attached to manipulator. Surgery is being performed on the phantom eye.

⚠ (:youtube v5jgBc5gqbA loop=1 fs=1 scale=1:)
This video shows a magnified view of an Alcon microforceps with a soft actuator telemanipulated using our microforceps stylus on the Geomagic Touch.

⚠ (:youtube FPV3SB9759A loop=1 fs=1 scale=1:)
Microscope view of a probe attached to the retinal manipulator. This video demonstrates the high resolution at the end-effector that can be achieved with our manipulator.

⚠ (:youtube BGBdD-tEHtQ loop=1 fs=1 scale=1:)
This video shows a comparison between manual and robotic membrane peeling surgery. Two different instruments are used to perform surgery on an artificial paper membrane attached to the retina in our phantom eye.

Selected Publications

M. Nambi, P. S. Bernstein, and J. J. Abbott, "Effect of Haptic-interface Virtual Kinematics on the Performance and Preference of Novice Users in Telemanipulated Retinal Surgery ," IEEE Robotics and Automation Letters, 2(1):64-71, 2017.

M. Nambi, P. S. Bernstein, and J. J. Abbott, "A Compact Telemanipulated Retinal-surgery System that uses Commercially Available Instruments with a Quick-change Adapter," J. Medical Robotics Research, 1(2):1630001, 2016.

M. Nambi, P. S. Bernstein, and J. J. Abbott, "A Compact Retinal-surgery Telemanipulator that uses Disposable Instruments," In N. Navab et al. (Eds.): MICCAI 2015, Part I, LNCS 9349, pp. 258265, Springer, 2015.


Dr. Paul Bernstein


We would like to thank Intuitive Surgical. The development of the retinal telemanipulator would not have been possible without their support.

Intuitive Tele-operation of Micromanipulators with Piezoelectric Stick-slip Actuators

Micro/Nano-manipulation deals with handling of extremely small objects on the order of 10-3 to 10-9m. Development of scanning and transmission electron microscopes (SEMs/TEMs) has enabled researchers to image and manipulate objects as small as 1\,nm in size. However, due to difficulties in simultaneous imaging and manipulating, micro/nano-manipulation is still a laborious task requiring great patience.

The goal of our research is to enable tele-operated rate control of micro/nano-manipulators without relying on any feedback from the vision system, other than the human user looking at the microscope image. This involves obtaining empirical models for the micro/nano-manipulator joints so that reliable position estimates can be obtained in the absence of feedback and development of control methods to move the end-effector using position estimates obtained from the open-loop models. Through intelligent algorithm development and drift compensation techniques we can greatly improve the efficiency with which such piezoelectric stick-slip actuators are operated.

Functional description of a piezoelectric stick-slip actuator.

Piezoelectric stick-slip actuators have become the foundation of modern micro/nano-manipulation due to their simple structure, high positional accuracy, unlimited movable distance, and high stability as they are supported by guiding surfaces. These actuators consist of a piezoelectric element and a sliding mass that moves relative to the piezoelectric element. A saw-tooth voltage is applied to the piezoelectric element. As the voltage slowly increases from 1 to 2, the piezoelectric element stretches by a distance D, and due to stick-slip friction between the piezoelectric element and the sliding mass, the sliding mass also advances (stick phase). When the voltage is quickly reduced from 2 to 3, the piezoelectric element quickly shrinks, but the inertia of the sliding mass prohibits it from moving backward as quickly, resulting in a net forward displacement of the sliding mass of d < D (slip phase). Typically, these actuators have no sensor feedback, and hence, the individual joints of the manipulators are controlled open-loop, using one knob per joint. It is not always clear what combination of joint commands will lead to a desired end-effector movement from only a microscope image. In addition, manipulators are often mounted on moving stages and on different surfaces at varying angles. The user observing the end-effector of such a manipulator under a microscope has to perform the difficult task of mapping the image frame to his/her egocentric frame of reference. Use of multiple manipulators further complicates the situation. Thus, micro/nano-manipulation is currently unintuitive and time consuming.

⚠ (:youtube LrVCAp1I40I loop=1 fs=1 scale=1.8:)
This video shows a Kleindiek MM3A micromanipulator being controlled using one knob per joint (on the left) vs. using coordinated rate control of the end-effector using a joystick (on the right). Direct joint control is unintuitive and time consuming, but is very common with this type of micromanipulator. More detail can be found in the paper "Toward Intuitive Teleoperation of Micro/Nano-Manipulators with Piezoelectric Stick-Slip Actuators", which appeared at IROS 2011.

⚠ (:youtube LyGktcuVxd4 loop=1 fs=1 scale=1.8:)
This video shows a Kleindiek MM3A micromanipulator being controlled using coordinated rate control with a simulated human user that simply attempts to move the end-effector toward the goal. The developed algorithm chooses the correct number of steps commands to send to each joint to accomplish the goal. Both videos use an open-loop model of joint step size. In the video on the left, the estimate of the end-effector position is updated, although it can be seen that error in the estimate grows with each step. In the video on the right, the estimate of the end-effector's position is never updated from the starting point. It can been seen in both videos that control of the end-effector is robust to these errors. More detail can be found in the paper "Toward Intuitive Teleoperation of Micro/Nano-Manipulators with Piezoelectric Stick-Slip Actuators", which appeared at IROS 2011.

Selected Publications

A. Damani, M. Nambi, and J. J. Abbott, "An Empirical Study of Static Loading on Piezoelectric Stick-Slip Actuators of Micromanipulators," Int. Symp. Experimental Robotics, 2012.

M. Nambi, A. Damani, and J. J. Abbott, "Toward Intuitive Teleoperation of Micro/Nano-Manipulators with Piezoelectric Stick-Slip Actuators," IEEE/RSJ Int. Conf. Intelligent Robots and Systems, pp. 445-450, 2011.

Page last modified on September 30, 2016, at 07:46 PM