0
TECHNICAL PAPERS

# A Prototype Manipulator for Magnetic Resonance-Guided Interventions Inside Standard Cylindrical Magnetic Resonance Imaging Scanners

[+] Author and Article Information
Nikolaos V. Tsekos1

Cardiovascular Imaging Laboratory, Mallinckrodt Institute of Radiology and Department of Biomedical Engineering,  Washington University, Room 1300, CB 8225, 4525 Scott Avenue, St. Louis, MO 63110tsekosn@mir.wustl.edu

Alpay Özcan

Biomedical MR Laboratory,  Washington University, Room 2313, CB 8227, 4525 Scott Avenue, St. Louis, MO 63110alpay@wuchem.wustl.edu

Eftychios Christoforou

Cardiovascular Imaging Laboratory, Mallinckrodt Institute of Radiology and Department of Electrical and Systems Engineering,  Washington University, Room 1300, CB 8225, 4525 Scott Avenue, St. Louis, MO 63110christoforoue@mir.wustl.edu

1

To whom correspondence should be addressed.

J Biomech Eng 127(6), 972-980 (Aug 01, 2005) (9 pages) doi:10.1115/1.2049339 History: Received March 21, 2005; Revised August 01, 2005

## Abstract

The aim of this work is to develop a remotely controlled manipulator to perform minimally invasive diagnostic and therapeutic interventions in the abdominal and thoracic cavities, with real-time magnetic resonance imaging (MRI) guidance inside clinical cylindrical MR scanners. The manipulator is composed of a three degree of freedom Cartesian motion system, which resides outside the gantry of the scanner, and serves as the holder and global positioner of a three degree of freedom arm which extends inside the gantry of the scanner. At its distal end, the arm’s end-effector can carry an interventional tool such as a biopsy needle, which can be advanced to a desired depth by means of a seventh degree of freedom. These seven degrees of freedom, provided by the entire assembly, offer extended manipulability to the device and a wide envelope of operation to the user, who can select a trajectory suitable for the procedure. The device is constructed of nonmagnetic and nonconductive fiberglass, and carbon fiber composite materials, to minimize artifacts and distortion on the MR images as well as eliminate effects on its operation from the high magnetic field and the fast switching magnetic field gradients used in MR imaging. A user interface was developed for man-in-the-loop control of the device using real-time MR images. The user interface fuses all sensor signals (MR and manipulator information) in a visualization, planning, and control command environment. Path planning is performed with graphical tools for setting the trajectory of insertion of the interventional tool using multislice and∕or three dimensional MR images which are refreshed in real time. The device control is performed with an embedded computer which runs real-time control software. The manipulator compatibility with the MR environment and image-guided operation was tested on a $1.5T$ MR scanner.

<>

## Figures

Figure 1

Topology of the MR scanner, with a human and the manipulator in scale. The gray lines represent the scanner gantry (diameter=60cm and length=160cm) and the diagonally stripped areas the patient couch. The embedded MR images of a volunteer (6ft.1in. tall and weighting 220lbs) were collected with a 50×50cm2 field-of-view and correspond to a (a) transverse, (b) coronal, and (c) and (d) sagittal planes. The arm of the device is 1in.(2.54cm) thick. The white cross identifies the isocenter of the scanner.

Figure 2

(a) Solid model of the seven DOF remotely actuated and controlled device (six computer controlled DOF to set x, y, z, θ1, θ2, and θ3 and one manually actuated to set the depth Δ). The DOF are depicted with the gray arrows together with their envelope of motion, means of transmission and construction materials. (b) Close up of the distal end of the device with the three rotational (θ1, θ2, and θ3) and the translational DOF to set the depth. (c) Illustration of the synergy of the two elbow joints (θ1 and θ2) in setting the Euler angle on the plane of the manipulator (ϕ) in a counterclockwise direction. The second elbow joint rotates the end-effector in the desired direction, while the first one brings it downward toward the patient and avoids collision with the gantry. (d) This maneuver is not necessary in setting ϕ on the clockwise direction.

Figure 3

Photographs of the system. (a) The major hardware components: the manipulator, the IOE host PC, the real-time embedded PC, the motor control electronics and the wire interface boxes (which reside inside the Faraday cage). (b) Side view photograph of the manipulator with the arm and the XYZ positioner anchored on the base. (c) Close-up of the end-effector with a biopsy needle attached on it. (d) Close-up of the first elbow showing the “through-joint transmission train,” based on double universal joints, and the bevel gears for the actuation of the joint.

Figure 4

Block diagram of the system hardware architecture depicting the individual components and their connections. To eliminate noise on the MR images, all the control electronics and power supplies are inside a Faraday cage and all wires are shielded with a multilayer aluminum sheath (delineated by the dotted boxes). The optical encoder wires (gray line) are also shielded and directed inside the Faraday cage to one of the wire interfacing boxes. The components inside the scanner room (which reside in the Faraday cage) are connected to those outside (i.e., PCI cards) via a shielded ribbon cable which passes under the scanner room shielded door. Raw MR data (black dotted lines) are transferred from the PC of the MRI to the IOE host workstation for processing and fusion with the manipulator position sensors and the planning∕control software. Control commands (black lines), generated by the control software, are send to the real-time embedded PC which processes them and directs them to the corresponding PCI cards and then to the motor control electronics which reside inside the scanner room.

Figure 5

(a) Screen caption of the visualization, planning and control software GUI from a MRI study on a phantom. The three windows are refreshed in real-time as raw MR data are received in the MR scanner and then transferred to the IOE host PC. In this example, two parallel images are used to set the entrance (left upper window) and the targeted (right upper window) points. The 3D window includes the MR images, displayed in the upper windows, and a virtual manipulator with dimensions, position and orientations scaled relative to the real device and to the MR images (i.e., relative to their field-of-view). The GUI includes buttons for manual control of the device and continuously provides the current position of the end-effector. (b) screen caption of the augmented reality window from another study where two orthogonal imaging planes were used. (c) Example output of the safety control routine depicting the motion-forbidden zone defined by the scanner’s gantry and the border of the subject (as calculated from MR scout images) and the motion paths of the joints of the device. The routine ensures that the device remains outside the forbidden zones.

Figure 6

Example of a three-slice set of orthogonal MR images used for calibration of the position (x0,y0,z0) of the end-effector. The MR marker (notched gray arrows) is a cross (3.1mm diameter) filled with 15mM Gd-DTPA attached on the end-effector. A custom-made fiberglass needle (white arrow) also filled with the same Gd solution is depicted in the Sagittal and Coronal slices. Zoomed-in areas of the marker are embedded with a gray border for better appreciation of the quality of the marker images. Calibration images are collected with the rotation DOF angles zeroed. From the sagittal, coronal and transverse planes are extracted the {y0,z0}, {x0,y0}, and {x0,y0}, respectively, which are then averaged per axis and entered in the control software. The calibration procedure is 1–2min long and can be repeated whenever required during a study.

Figure 7

MR compatibility studies of the device. (a) Phantom studies to evaluate the effect of electronic and motor operation on the SNR. When the motor is not actuated, the SNR of a GRE image is 180 and drops substantially to 32 when the unshielded motor is powered on. However, with all electronics shielded (the motor controllers and power supplies placed inside the Faraday cage and all wiring wrapped with an aluminum sheath) the SNR recovers to 120. As demonstrated the achieved SNR is sufficient for real-time in vivo MR studies. Example images of a pig collected with HASTE (b) and TrueFISP (c) sequences without the device in place, with the device in place and the arm over the abdomen, and the device continuously actuated. Panels (b) and (c) and graphs (d) and (e) demonstrate that the SNR and CNR (fat vs muscle) are reduced during the operation of the device, but they are not substantially compromised. Graphs (d) and (e) review the SNR and CNR data from TSE, HASTE, and TrueFISP highlighting that the device actuation is compatible with real time MR imaging.

Figure 8

Frames from an MR-guided maneuvering of the device to align a custom-made Gd-filled fiberglass needle (white notched arrow) along a predetermined trajectory (dotted white lines). The panels show activation (gray arrows) of the (a) X DOF and (b) θ3 DOF on the same transverse slice. Each one of the panels show five frames of only one slice out of the four-slice VINAOS multislice acquisitions (3D reconstructions are shown in Fig. 9).

Figure 9

(a) and (b) Selected frames from the activation of the X DOF showing two of the four slices collected with VINAOS acquisition to visualize the motion of the device (same study as this in Fig. 8). The transverse slice in (a) shows the entire motion of the needle which appears on the sagittal image (centered at the final position of the needle) at the conclusion of the motion. (c) and (d) Two different perspectives of the 3D reconstruction of all the slices collected with the VINAOS acquisitions. These reconstructions allow inspection of the path and trajectory of the needle along predetermined planes to facilitate assessment of its alignment relative to structures inside the target.

## Discussions

Some tools below are only available to our subscribers or users with an online account.

### Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related Proceedings Articles
Related eBook Content
Topic Collections

• TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
• EMAIL: asmedigitalcollection@asme.org