Neuromaster: a robot system for neurosurgery

Abstract

This work introduces a robot system for minimally invasive frameless stereotactic neurosurgery. The system consists of a robot arm for precise positioning of surgical tools, a vision system for intro-operative registration, and a preoperative planning system. The robot is a custom designed accurate arm with five degree-of-freedom. The vision system uses two cameras to automatically generate the target position of the robot The robot can be controlled autonomously as well as interactively through an intuitive way. Experiments and clinical trials approve the robot system is effective and powerful.

Full text

Page 1

NeuroMaster: A Robot System for Neurosurgery*

Zengmin Tian
Navy General Hospital
No.6 Fucheng Road, Haidian District,
Beijing, China, 100037
Junchuan Liu, Yuru Zhang,
Tianmiao Wang and Hongguang Xing
Robotics Institute
Beihang University
No.37 Xueyuan Road, Haidian District,
Beijing, China, 100083
buaaljc@sina.com




minimally invasive frameless stereotactic neurosurgery. The
system consists of a robot arm for precise positioning of surgical
tools, a vision system for intro-operative registration, and a
preoperative planning system. The robot is a custom designed
accurate arm with five degree-of-freedom. The vision system uses
two cameras to automatically generate the target position of the
robot. The robot can be controlled autonomously as well as
interactively through an intuitive way. Experiments and clinical
trials approve the robot system is effective and powerful.


Index Terms – Robot; Frameless Stereotactic Neurosurgery;
Registration; Planning; Cameras

I. INTRODUCTION

* This work is supported by the National High Technology Program of China under the contract 2001AA422110.
Abstract - This paper introduces a robot system for
The stereotactic neurosurgery is widely adopted as a
minimally invasive method in the clinic, which operates in a
three-dimensional anatomic space with the help of a reference
system. At present, a stereotactic frame is often used as the
reference system to be fixed on the patient’s head to localize
the part for therapy. The fixation causes much damage and
pain to the patient. Furthermore, it may limit the path of a
surgical tool in some orientations and force the surgeon to
choose a worse path during the operation.
To overcome the disadvantages of the stereotactic frame,
Kwoh et al. made an early attempt to use an industrial PUMA
200 robot for the CT-guided brain tumor biopsies [1]. After
that a number of neurosurgical robot systems have been
developed and marketed [2-6].
Our group has been devoted to the development of robot
systems for neurosurgery since 1995. The first system
integrating a PUMA 260 industrial robot was successfully
applied to the clinic on May 6, 1997 in Navy General Hospital
in China [7]. The second system integrates a special designed
robot arm with five passive joints to achieve cost-effective
solution and improve the safety [8]. We also developed a
simulation and training system of robot assisted surgery based
on virtual reality [9].
This paper describes NeuroMaster, a robot system for
stereotactic neurosurgery developed recently at the Robotics
Institute of Beihang University. The goal of the new
development is to increase accuracy, dexterity, and especially
the level of automation for the application of remote operation
[10]. The contents of this paper are arranged as follows.
Section 2 describes the procedure of the neurosurgery using
our robot system. Section 3 introduces the major parts of the
robot system. Experimental and clinical results are given in
section 4. Section 5 is the conclusions.
II. SURGICAL PROCEDURE
The surgical procedure using NeuroMaster is as follows.
1) Surgical planning: CT or MRI images of patients are
used for pre-operative planning. The surgeon specifies the
tumor and the target position and orientation of surgical
instruments. The planning software reconstructs the tumor and
the skull to provide an intuitive 3D display.
2) Registration: Four markers fixed with patient head are
used for registration of the image space to the robot space. A
binocular stereovision system is calibrated in the operation
space in which the patient head is to be placed. The robot
locates the four markers in its own coordinate frame through
the vision system.
3) Insertion: the robot receives the target location from
the planning software and moves automatically to the location,
placing the tool holder at the entry point for operations. The
surgeon prepares the hole and inserts the surgical instrument
along the specified straight-line path.
III. ROBOT SYSTEM

include the robot arm with its controller, the 3D vision system,
the surgical planning software, and the remote control system.
The details of the robot and the vision systems are presented
in the following. Details on other components can be found in
references [7], [8] and [10].

A. Kinematics
In the stereotactic neurosurgery, typical manipulation is
the insertion of surgical instruments through a small hole on
the patient head. During the operation, it is important to
accurately position the tip of the instrument inside the patient
head. Meanwhile, the path of insertion is carefully planed to
minimize the damage to the normal tissue of the patient brain.
As shown in Fig. 1, the major components of the system

Page 2

Since the end effector is needle-like, the freedom of its
rotation can be ignored. This requires that the robot has at
least five degree-of-freedom (DOF).


Fig. 1. The NeuroMaster system

The robot we developed in this system is a 5 DOF
manipulator mounted on a mobile platform. Its first axis from
the base frame is translational being able to move up and
down vertically. The other four are rotational. The end
effector is a tool holder. The assembly of the robot arm and its
kinematic diagram are shown in Fig. 2 and Fig. 3 respectively.


Fig. 2. The NeuroMaster robot

d1
a2
d4
d5
d6
root
tip

Fig. 3. Kinematic structure of the robot

In the world frame of the robot, let
),,(
zyx
tttT
and
),,(
zyx
aaaa
be the tip position and the orientation of the
surgical instrument, where
x
a ,
y
a ,
z
a are the third-column
elements of the homogenous transformation matrix. The
forward kinematics of the robot is as follow
+(=
sdscdst
y
54
5235423
5 235423
145
s
5
c
4
c
6
224562345
+
54623
22456234554623
=
+=
=
+=
+)+(+)+(=
+)+()
ssa
ccccsa
cca
dcdssdt
sadcdc
cadcdssdscdct
z
y
x
z
x

where
i
θ
i
θs
s
sin
sin(
=
=
,
ii
θ
θc
+
cos
,
c
=
,
cos()
jiij
)+=
ji ij
θθ
,
and
iθ is the ith joint angle of the robot.
There are no analytical solutions to the kinematic
equation sets above. We developed a novel iterative method
for efficient numerical solutions. Using the iterative method
we can transform the nonlinear problem of multiple variables
to that of only one variable, and get rapid and reliable
solutions. There are four possible solutions for each given
position of the end effector. The feasible solutions are found
by considering the constraints for the range limit and the
consistency of the joint motion. Each feasible solution
corresponds to a configuration of the robot. The surgeon can
select the best one for the ease of operation.

B. Dexterity
When planning for the insertion, the surgeon must select
the path carefully to avoid some important function area or
blood vessels inside the brain. It may cause serious damage to
patient health if the insertion path is not properly selected.
This means that for a given position, it is desired that any
orientation selected by the surgeon can be achievable by the
robot. This requires that the robot can provide necessary
dexterity.
During the kinematic design of the robot, the dexterity
needs to be analyzed. For a tip position of the robot end
effector, T, the orientation of the end effector can be defined
using two parameters of α and β , as shown in Fig. 4.
X
Y
Z
α
β
T
R

Fig. 4 Diagram of the end effector orientation

Page 3

We define the dexterity of the robot as
∑∑
=
m
1
∑∑
=
m
1
=
=
a
=
b
n
ab
n
D
D
1
1
1
),(
βα

where a and b are the numbers of sampling of α and β
respectively,
0=),(βαD
when the orientation can’t be
achieved and
1=),(βαD
when the orientation can be
achieved. For the NeuroMaster robot, the dexterity is
calculated in its workspace (Fig. 5). The value of dexterity
changes with the configuration of the robot. It is important to
find the most dexterous area and to know if it satisfies the
dexterity requirement. Fig. 6 shows the dexterous property of
the robot where “^”, “.” and “*” indicate the dexterity less
than 0.4, between 0.4 and 0.8, and larger than 0.8,
respectively. From the simulation result we can see that the
“*” area is the optimal workspace we should use in the clinic.
In this area the robot can reach a point from many different
orientations.


Fig. 5. Horizontal section of the workspace


Fig. 6. Dexterity analysis
C. Registration Through Vision
Coordinate relationship between the medical images
space and the robot space are established using markers fixed
on the patient head [8]. This process is called registration. In
order to make the registration, the markers must be measured
in the robot space. There are different types of measurement,
such as mechanical, ultrasound, infrared, etc. We developed a
binocular stereovision system to perform the measurement.
The images of patient head are captured by two CCD cameras
with
576× 768
resolution and processed in a personal
computer.
There are two steps for the registration. The first step is
to correlate the camera image space with the robot space. The
second step is to correlate the camera image space and the
medical image space. During the first step, the cameras are
calibrated using 12 points carefully selected in the task space
(Fig. 7). Once the calibration is finished, the relation between
the camera space and the robot spaces is determined. In the
second step, a transformation matrix between the two image
spaces is calculated using the marker coordinates in the two
spaces. Since the patient is fixed on the bed during the
registration and operation, there is no need of a vision based
control loop. After the registration, the desired position of the
surgical instrument specified by the surgeon in the medical
image can be transformed into the target position for the robot
end effector.


Fig. 7. Registration through vision

D. Safety Measures
Safety is the most important factor we considered in the
design of the system. To ensure the safety, redundant safety
functions are designed into the system from aspects of
hardware and software.
In the hardware of the system, mechanical and electrical
sensors are used for collision protection of the robot itself.
Each joint of the robot has an electromagnetic brake that can
stop the joint when the power is off. A PMAC controller,
widely used in industry, is used for robot motion control. The
watchdog on the PMAC can stop the whole system if there is
something wrong with the controller. The software for motion
and surgical planning runs on an industrial computer. The
surgeon can stop the robot immediately using an emergency
button.

Page 4

In the software, the motion of the robot can be simulated
and displayed graphically to predict the robot configuration
for the task [10]. The software allows the surgeon to select the
optimal configuration of the robot to avoid the collision
between the robot and the patient. The software also defines
the limit positions of each joint to protect the robot itself.
Safety functions are also designed into the process of
operation. The sequence of operation process is implemented
in the software. Warning is provided to ensure correct
sequence for the operation. Before going to the target, the
robot performance will be verified by moving it to one of the
marker. This step verifies the correctness and accuracy of the
positioning.

E. Joystick Control
The robot can move automatically as programmed. We
also provide a manual mode for interactive control. The
operator can send his command by moving a joystick and the
robot moves up and down, left and right, forward and
backward accordingly. Using the joystick, the surgeon can
control the robot intuitively. It also provides a convenient way
to move the robot around in the workspace during the
calibration of the vision system as well as the measurement for
the system accuracy.
IV. EXPERIMENTAL RESULT
Before this system is used in the clinic, we have done a
large amount of experiments in laboratory and hospital. These
experiments gave us confidence to apply this system in the
clinic.

A. Accuracy
We use a skull to test the positioning accuracy of the
system. In the experiments, the same procedure as the one
used in the real clinical application is performed. Four
markers are pasted on the skull as shown in Fig. 8. A steel ball
of 2 millimeter in diameter is mounted inside the skull to serve
as the target point for the surgical instrument.
We use an accurate mechanical measuring device (Fig. 9)
to measure the desired target position defied by the steel ball
and the real target position defied by the tip of the robot end
effector. The distance between the two points indicates the
positioning accuracy of the system.


Fig. 8. Skull for experiment


(a)

(b)
Fig. 9. Experiments for accuracy

Fig. 10 shows the data obtained from the experiments. In
the figure, the horizontal axe is the number of positions of the
skull, and the vertical axe is the error corresponding to each
position. For different positions of the skull, the maximum
positioning error is less than 2.5 millimeter, and the average
error is less than 1.5 millimeter.

0
0.5
1
1.5
2
2.5
(mm)
the Positioning Error
13
5
7
9
11
13
the Number Instead of
the Position of the Skull

Fig. 10. Positioning error

B. Clinical Trial
The first clinical trial on August 4, 2003 proved the
success of the NeuroMaster (Fig.11). The operation was
performed according to the process mentioned in section 2.
There have been 15 operations performed in the Navy General
Hospital in Beijing since the first one, including cerebral
hemorrhage therapy, tumor biopsy, and tissue stimulation with

Page 5

an electrode. On September 10, 2003, the first remote surgery
was carried out successfully between Beijing and Shenyang
with 600 kilometers separating surgeon and patient [10].


Fig. 11. Clinical application
V. CONCLUSIONS
We have presented a new robotic system for neurosurgery.
The features of the system include:
? The newly designed robot which provides necessary
accuracy and dexterity for neurosurgical applications.
? The binocular stereovision system which proves to be
a good way to make accurate and efficient registration
during the surgery.
? The safety functions designed into the system which
proves to be important to make successful clinical
applications and bring the possibility of remote
operations.
Our future work will focus on increasing the efficiency
and accuracy of the visual registration system.
ACKNOWLEDGEMENT
This research is supported by the National High-tech

Program of China under the contract 2001AA422110.
The authors also present thanks to Da Liu, Dangxiao
Wang, Bo Liu, Wei Li, Shouhong Zhang who have also
participated in the research.
REFERENCES
[1] Y.S. Kwoh, J. Hou, E.A. Jonckheere, S. Hayati, “A robot with improved
absolute positioning accuracy for CT guided stereotactic brain surgery,”
Biomedical Engineering, IEEE Transactions on, Vol. 35 Issue: 2, pp.
153 – 160, Feb. 1988.
[2] S. Lavallee, J. Troccaz, L. Gaborit, P. Cinquin, et al., “Image guided
operating robot: a clinical application in stereotactic neurosurgery,”
Robotics and Automation, 1992. Proceedings, 1992 IEEE International
Conference on, vol.1, pp. 618 – 624, May 1992.
[3] C.W. Burckhardt, P. Flury, D. Glauser; “stereotactic brain surgery,”
Engineering in Medicine and Biology Magazine, IEEE, Vol. 14 Issue: 3,
pp. 314 – 317, May-June 1995.
[4] K. Masamune, E. Kobayashi, Y. Masutani, M. Suzuki, et al.,
“Development of an MRI-compatible needle insertion manipulator for
stereotactic neurosurgery,” J Image Guid Surg, vol. 1, pp. 242-8, 1995.
[5] B. Davies, S. Starkie, S.J. Harris, E. Agterhuis, et al., “Neurobot: a
special-purpose robot for neurosurgery,” Robotics and Automation, 2000.
Proceedings. ICRA '00. IEEE International Conference on, vol. 4, pp.
4103 – 4108, April 2000.
[6] Qing Hang Li, Lucia Zamorano, et al., “The Application Accuracy of the
NeuroMate Robot-A Quantitative Comparison with Frameless and Frame-
Based Surgical Localization Systems,” Computer Aided Surgery, 7: 90 –
98, 2002.
[7] M.D. Chen, T. Wang, Q.X. Zhang, Y. Zhang, Z.M. Tian, “A robotics
system for stereotactic neurosurgery and its clinical application,”
Robotics and Automation, 1998. Proceedings. 1998 IEEE International
Conference on, Vol. 2, pp. 995 – 1000, May 1998.
[8] D. Liu, T. Wang, Z.M. Tian, J.X. Du, et al., “Study on robot-assisted
minimally invasive neurosurgery and its clinical application,” Robotics
and Automation, 2001. Proceedings 2001 ICRA. IEEE International
Conference on, vol. 2, pp. 2008 – 2013, 2001.
[9] T. Wang, D. Liu, Z.G. Wang, et al., “A simulation and training system of
robot assisted surgery based on virtual reality,” Medical Imaging and
Augmented Reality, 2001. Proceedings. International Workshop on, pp.
103 – 107, June 2001.
[10] Cai Meng, Tianmiao Wang, Yuru Zhang, Zengmin Tian, et al., “ Remote
Surgery Case: Robot-Assisted Tele-neurosurgery, ” accepted by ICRA
2004.