1.  Background

Physicians increasingly seek to use MRI machines for not only diagnosis and preoperative imaging, but also for real-time image guided procedures. Limited patient access within the MR bore presents a challenge and in current procedures, patients are repeatedly extracted from the bore to let physicians iteratively position tools between scans in an approach that does not take advantage of real-time imaging and is susceptible to anatomy shifts. Robotic solutions can improve access and tool positioning accuracy, but MR-compatibility requirements constrain their design and limit transparency, preventing physicians from feeling tool/tissue interaction forces. These devices isolate patients from physicians, forcing them to rely on visual cues. Physicians want haptic feedback and find, after years of practice, that they rely on feel as much as vision for some procedures, such as membrane puncture or identification of dense tissue.

2.  Vision

We are working in collaboration with Dr. Bruce Daniels at Stanford Radiology and Dr. Peter Whitney at Northeastern University on a 3-DOF teleoperated needle manipulator for MR-guided biopsies. The device utilizes a rolling-diaphragm hydrostatic transmission to achieve haptic transparency in the needle insertion direction. Prostate and liver biopsies are the initial use cases we are targeting. Liver biopsy is a particularly convincing application as the liver is subject to relatively large displacements during procedures making real-time image guidance desirable.

3.  1-DOF Hydrostatic Manipulator

A 1-DOF device was created to characterize the force transparency and position tracking accuracy of the rolling-diaphragm hydrostatic transmission. A single water-filled line maintains stiffness and a second, thinner and more flexible line with pressurized air maintains a positive pressure under all conditions. Opposed rolling diaphragm actuators are located at each end and connected via a timing belt to a rotary joint. Although the prototype has a few ferrous components, it is simple to substitute them with aluminum, brass, ceramics, and/or polymers for tests under MR. At the needle side, we convert rotary to linear motion using a capstan drive, as commonly used in haptic devices for its combination of low friction, smoothness and high stiffness. In an initial version of the device, a lever at the input and uses a single stroke to drive the needle. After conducting user tests and obtaining user feedback, a second version of the apparatus was created. It was motivated by the observation that physicians often drive the biopsy needle with multiple motions, regrasping between strokes. An illustration of the two version is presented below.

A) Master and slave sides of the transmission use paired rolling diaphragm actuators. One line is water-filled for stiffness; the other contains pressurized air. The slave side uses a capstan/cable drive to propel the needle on a linear track.

(B) Modified version uses knob for enhanced tactile sensitivity and smaller capstan wheel. A roller bearing clutch allows multiple input rotations over full travel of needle.

We conducted experiments against springs to characterize the system performance. The experimental setup and results are shown.

1-DoF transmission system with associated sensors for characterizing force and position transparency. Optical encoders track input and output rotation as well as linear displacement of the needle. Force sensors measure the input torque and force experienced at the needle tip.


Comparison of system transparency when interacting with soft (A) and stiff (B) springs. Upper row: Changes in force are tracked; magnitudes are expressed with a small lag and higher relative accuracy in the stiff case. Middle row: Displacement errors are approximately equal in the soft and stiff cases. Bottom row: Stiffness as experienced by the input (master) and output (needle). For the soft spring, input stiffness matches endpoint stiffness; for the stiff spring, elasticity in the apparatus contributes noticeably to the apparent stiffness at the master.


User tests with phantom tissue were also conducted. The tests measured users’ ability to detect membrane puncture events and to distinguish between different stiffness levels using the teleoperator system versus manipulating the needle directly. Membrane punctures had an average of 77% success (standard deviation was 9%, range = 63-88%). Puncture forces were realistically sized on the lighter end of membranes one would encounter in transperineal prostate biopsy, ranging from 0.31-0.66 N (average = 0.48N, standard deviation 0.10). Users had 100% accuracy in distinguishing all pairs of spring combinations except two: 88% accuracy in distinguishing 2.86 N/mm and 1.54 N/mm, and 63% accuracy for 0.82 N/mm and 0.57 N/mm.

4.  2-DOF "Drive by Wire" Manipulator

(TODO: Embed video of 2-DOF manipulator)

5.  Custom Long Stroke Rolling Diaphragms


6.  Research Team and Collaborations

  • PI: Mark Cutkosky, Stanford University
  • Graduate Students: Samuel Frishman, Natalie Burkhard, Alex Gruebele
  • Collaborators:
    • Genliang Chen, Shanghai Jiao Tong University
    • Bruce Daniels, Stanford Radiology
    • Peter Whitney, Northeastern University
  • Funding
    • This work is funded in part under NSF CHS:Small:Collaborative: #1617122 and #1615891 "Teleoperation with passive, transparent force feedback for MR-guided interventions"

7.  Publications

Burkhard, N., Frishman, S., Gruebele, A., Whitney, J.P., Goldman, R., Daniel, B.L., Cutkosky, M.R., "A rolling-diaphragm hydrostatic transmission for remote MR-guided needle insertion," IEEE ICRA 2017 (pdf). Best Robotics Paper Finalist

Page last modified on May 02, 2017, at 04:36 PM