Content uploaded by San Yoon
Author content
All content in this area was uploaded by San Yoon
Content may be subject to copyright.
ARTHROBOT : A New Surgical Robot System for Total Hip Arthroplasty
Dong-Soo Kwon*, Yong-San Yoon*, Jung-Ju Lee*, Seong-Young Ko*,Kwan-Hoe Huh*,
Jong-Ha Chung*, Young-Bae Park*, Chung-Hee Won**
* KAIST Mechanical Engineering Department (e-mail: kwonds@me.kaist.ac.kr)
** Chungbuk University Hospital (e-mail: chwon@med.chungbuk.ac.kr)
Abstract
This paper presents mechanisms and control methods
of a new surgery robot for total hip arthroplasty (THA). To
minimize the disadvantages of the conventional
registration method, a new gauge-based registration
method has been proposed, and a 3-DOF robot has been
developed that can be mounted on a femur. The proposed
surgical robot can operate along a pre-programmed path
autonomously, in addition to allowing a surgeon to
directly control the motion of the surgical robot with their
experience and judgment during an operation. For this
purpose, a master is attached to the surgical robot and
admittance display is used in control. ARTHROBOT, this
new arthroplastic surgical robot system, is expected to be
adaptable to surgical needs and practice in the operating
room.
1. Introduction
Either by trauma or disease, a hip joint can be damaged,
and this induces pain and reduces the range-of-motion in
the hip joint. In this case, Total Hip Arthroplasty (THA),
which is the name of an operation for replacing the
damaged hip joint with an artificial hip-joint, is performed.
The artifical hip joint is composed of an acetabular
component and a femoral stem, the stem being either one
of two types, cemented or cementless. As its name
suggests, a cementless stem does not need cement because
the bone grows into the porous part on the stem and the
stem is fixed to the femur. When using a cementless stem,
the conformity between the bone and the implant greatly
affects the success of the surgery and the recovery of the
patient [1]. Thus it is very important to carve a hole in the
femur that precisely fits the shape of the artificial hip
implant in order to increase conformity and minimize gaps.
In a conventional cementless THA, the surface
conformity between the bone and the implant is less than
30%. This causes slow recovery and shortens the life of
the implant [2]. To improve this situation, robotic surgical
systems that can make a precise cavity in the femur were
developed [3]. In particular, Integrated Surgical System
Co. developed a commercial THA system, the
ROBODOC® Surgical Assistant System [4].
By using these surgical robots, the surface conformity
can be improved and patients can recover more rapidly [5].
In such robotic systems, to register the surgical robot to
the femur, the fiducial markers are implanted onto the
femur of a patient before surgery, and a CT scan of the
femur is performed. In real surgery, the robot system is
registered through comparing the measured positions of
the markers and the positions of the markers in CT scans.
This paper introduces a new arthroplastic surgical
robot system which we call ARTHROBOT. Since a small
surgical robot is mounted onto the patient’s femur by a
bone clamp, the registration procedure becomes simple,
and the cost of the operation can be reduced. Since the
robot has a master/slave-combined structure, a surgeon
can directly control the motion of the surgical robot like
an advanced surgical tool. Through an admittance display,
the surgeon can feel the comfortably pre-defined virtual
environment. Also a virtual hard wall is displayed at the
surgical boundary to ensure surgical accuracy.
2. Gauge-based Registration Method
/
We have proposed a greatly simplified cavity
machining method for robot-assisted total hip athroplasty
surgery that requires neither CT scanning nor the insertion
of fiducial markers before surgery [6].
In this technique, a surgeon prepares the distal portion
of the femoral cavity using a conventional manual
reaming process which centers the distal end of the cavity
relative to the cortical bone. Next, the surgeon inserts a
reamer-shaped registration gauge into the prepared distal
cavity aligning the front of the gauge with the direction of
the femoral neck; this defines the orientation for the final
implant (Fig. 1).
The surgeon then attaches the base for a small surgical
robot to the gauge part of an external femoral clamp using
an adjustable linkage consisting of two sets of ball and
socket joints and a slider. Both the base frame and the
registration gauge have matching mating surfaces so that,
with the linkage unlocked, the surgeon can maneuver the
base frame into solid contact with the registration gauge
(Fig. 2).
Proceedings of the 2001 IEEE/RSJ
International Conference on Intelligent Robots and Systems
Maui, Hawaii, USA, Oct. 29 - Nov. 03, 2001
0-7803-6612-3/01/$10.00 2001 IEEE 1123
Fig. 1. Proposed hip arthroplasty methods
Fig. 2. Alignment of gauge and base frame
(a) A frame on the sawbone (b) A clamp on the femur
Fig. 3. The frame attached on the femur
Once the linkage is locked, the base frame is fixed
relative to the femur and the registration gauge can be
removed. At this point, the robot can be mounted on the
base frame to machine the femoral hole (Fig 3(a)). It was
verified in the operating room that a bone clamp used in
the frame could be attached to the femur and that there
was enough space for the robot to be mounted (Fig 3(b)).
3. A Surgical Robot Design
In order to perform the proposed surgical method, a
small surgical robot needs to be mounted on the femur.
The minimum degree-of-freedom (DOF) of the robot is
three for most stem shapes of artificial hip joints. This
section presents the mechanism of a surgical robot and its
optimal design.
3.1 The mechanism of the surgical robot
Considering the dexterity of the THA, five kinds of
candidates for the mechanism of the surgical robot are
chosen as shown in Fig. 4.
Fig. 4. Candidates for the surgical robot
Since the robot must be attached onto the patient’s
femur to perform a proposed operation, not only precision
and rigidity, but also weight and size are important. For
decision criteria, the following factors of precision,
rigidity, weight, size, number of joint, disinfection, and
workspace are chosen. Five kinds of candidates are rated
on three levels about each criterion, and these grades are
multiplied by proper weighting factors. Each candidate is
evaluated by the sum of the weighted grades. From these
evaluations, the parallel-3RPS type (Fig. 4 (e)) is selected
as the suitable mechanism for the surgical robot.
3.2. Performance index
A performance index has been formulated to optimize
dexterity, force requirement, and uniformity. To represent
a global dexterity index D
g
, a global external force index
F
g
, and the gradient value of these (GD
g
, GF
g
), the overall
performance index is defined as Eq.(1).[7][8]
(a) Femur is exposed.
(b) Femoral head is removed, and frame is attached to femur, the
hole is made using a starter.
(c) A reamer-shaped gauge is inserted into femur
(d) Frame is maneuvered into place against reamer-shaped gauge
and locked into position
(e) Gauge is removed and robot installed on frame to machine
cavity in femur
1124
gggg
GFGDFDPI +++=
(1)
Here, the bar symbol indicates a normalized value of
each index value.
3.3. Optimization
The surgical robot is optimized over the possible range
of all design variables. The workspace of initial/optimized
design and the performance indices are shown in Fig. 5
and Table 1. The workspace of optimized design is more
similar to the required workspace than that of the initial
design. The precision at the tool tip is much improved by
the concentrated workspace. The maximum force index
shows a triple value compared to the initial design after
optimization. This makes it possible to use a smaller
actuator, so the total weight and size of the robot can be
decreased. Moreover, the overall performance index PI is
doubled. The optimized prototype is shown in Fig. 6 (a)
and a real surgical robot is manufactured as shown in Fig.
6 (b) based upon this optimization result.
(a) Initial design (b) Optimized design
Fig. 5. Comparison of workspace
a global
dexterity
index D
g
a global
force
index F
g
a gradient
dexterity
index GD
g
a gradient
force
index GF
g
overall
performance
index PI
Initial
design
0.929 0.248 4.26 0.18 1.71
Optimal
design
0.964 0.705 3.46 0.174 3.43
Table 1. Comparison of performance index
(a) An optimized prototype (b) A manufactured robot
Fig. 6. The parallel-type surgical robot
4. Control of a Master/Slave-Combined
Surgical Robot
Most previous robots used in THA grind the bone
along a pre-programmed path. In surgery, the surgeon’s
experience and judgment are very important factors
affecting the success of its outcome. By combining a
master and a slave robot, a surgeon can control the motion
of the surgical robot directly, thus allowing the surgeon to
exercise his or her discretion during surgery, in addition,
the robot also retains the capability to operate
autonomously along a pre-programed path.
In the master/slave-combined surgical robot, the force
exerted by a surgeon to the master is used as command
signals for desired motion. For easy operation of a
surgical robot in variable environmental characteristics,
the admittance between human guide forces and surgical
robot velocity must be shaped properly. And the virtual
hard wall display has been adopted to make a surgeon feel
the cavity boundary.
4.1. 1-DOF system modeling
The characteristics of a master/slave-combined surgical
robot can be modeled as a 1-DOF mass, damper linear
system as shown in Fig. 7.
Fig. 7. 1-DOF master/slave-combined robot system
In Fig. 7, f
h
and f
e
denote an interaction force between
human arm and robot, and between robot and environment,
respectively. The dynamics of a human arm, a robot and
an environment are given by the following equations:
roproprophop
xkxbxmf ++=−
DDD
τ
(2)
rrrrehr
xbxmff
DDD
+=−+
τ
(3)
rerereee
xkxbxmf ++=−
DDD
τ
(4)
where
τ
op
and
τ
e
denote muscle force of the human arm
and force/torque of environment, respectively.
m
op
,b
op
and k
op
denote mass, damping coefficient, and stiffness of
the human arm, while
m
r
, b
r
and m
e
,b
e
,k
e
denote those of
the robot and environment.
1125
4.2. Admittance display
In order to carve a hole in the femur, it is desired that
the master/slave-combined surgical robot has high
stiffness, and low friction. In this kind of a robot system,
measured interaction forces or torques are used as the
inputs to calculate the desired trajectory. This control
scheme is generally called admittance display mode [9].
In this study, the controller is constructed to maintain
desired admittance between human guide force at the
master and the robot velocity. Measured force
f
h
is used as
input for admittance model Y
D
(s) to generate a trajectory
of the surgical robot based on desired admittance. The
master/slave-combined surgical robot is position-
controlled. If the position controller perfectly follows the
trajectory based on desired admittance, it is possible to
have easy maneuvering of the surgical robot regardless of
environmental conditions.
Local position-controller of the surgical robot is as
follows:
)()(
rdvrdpr
xxKxxK
DD
−+−=
τ
(5)
The trajectory generation based on desired admittance
is as follows:
ddd
D
ksbsm
sY
++
=
2
1
)(
(6)
)(sYfx
Dhd
=
(7)
)(ssYfx
Dhd
=
D
(8)
where,
m
d
, b
d
and k
d
denote mass, damping coefficient,
and stiffness of desired admittance model. From equations
(5)~(8), the controller of the master/slave-combined
surgical robot is constructed as shown in Fig. 8,
Fig. 8. Controller for admittance display
where H and E denote impedance of human arm and
environment. G
r
is the robot system.
4.3. Virtual hard wall display
Since the geometrical information of the artificial
implant is known, the surgical region can be divided as
shown in Fig. 9. In previous research, a virtual hard wall
is displayed using a virtual wall model composed of a
spring and damper [10,11]. However, we have adopted a
new strategy for the master/slave-combined surgical robot
having high stiffness and friction to ensure surgical
accuracy.
Fig. 9. Surgical region division
In Region I, far from the boundary, relatively high
admittance is displayed for easiness of operation. And in
Region II, near the boundary, relatively low admittance is
displayed and the surgeon feels difficulty in maneuvering
the surgical robot. This leads to low robot velocity with
respect to guide force and increases safety. This also gives
the surgeon a feeling of being near the boundary of the
surgical region. Near the surgical boundary that is defined
based on the 3-D CAD model of a femoral stem, the
desired position of surgical robot is generated so as not to
go over the boundary. This makes it impossible for the
robot to exceed the surgical boundary regardless of the
guide force of a surgeon within the actuator capacity.
5. Experiments
Before applying the control method to the proposed
parallel-type surgical robot, to evaluate the machining
performance of a manufactured surgical robot, we carry
out an experiment carving a hole using a parallel-type
surgical robot. The robot is controlled autonomously
along a pre-programmed tool-path by the PID position
controller. The tool-path is generated based on the 3-D
CAD model of a femoral stem. Because the robot is fixed
by the bone clamp in real surgery, the experiment should
be carried out with actual frame to know the accuracy of
the whole system. However, at first, since we want to
know just the accuracy of the robot alone, the experiment
is performed using the jig in Fig.10. Wood is chosen as the
machining material, because its material property lies
between that of cortical and cancellous bone.
After the machining process, the specimen is cut into
halves using a milling machine. Then the surface of
1126
processed specimen is measured by a Coordinate
Measuring Machine (DUKIN Co., ASTRO 543C).
Measuring points are distributed at 2mm intervals. 35~51
valid data points are acquired from each specimen and the
results are compared with the CAD model. The relative
error has been calculated without considering the offset
(Fig. 11).
Fig. 10. Parallel-type surgical robot and experimental set
It can be known that most of the errors are in the range
of ±0.2mm. This is a much improved result compared
with that of manually operated surgery, and below the
allowable error of total hip replacement surgery[13].
Therefore, it is expected that this surgical robot can be
used in real hip surgery.
Fig. 11. Machining error
After the experiment on the machining performance,
we apply the control method to the master/slave-combined
surgical robot. In the experiment, the surgical region is
simplified as a concentric circle. To easily evaluate the
control method, we control the robot in only 2DOF
motion. The cutting materials are polyethylene,
MDF(Medium Density Fiber) board and the femur of a
cow. The Young’s modulus of polyethylene is 1.1GPa ,
and that of MDF is 0.3 Gpa. The spongy bone of a human
is 0.1 ~ 1.0GPa [12], so these materials are suitable to
simulate human spongy bone characteristics.
The first experiment is to view how the operator feels
different force and velocity w.r.t. an admittance model. In
Region I,
b
d
= 0.2 [Nsec/mm] and its radius is 13mm. In
Region II,
b
d
= 0.8 [Nsec/mm], k
d
= 0.05 [N/mm], and its
radius is 20mm. When an operator pushes the handle of
the master/slave-combined manipulator in the y-direction,
position and forces are measured as shown in Fig. 12. In
this case, the operator has high-speed motion and
relatively low force for maneuvering in Region I. The
operating force is about 1.0N in Region I, and 2.5N in
Region II. Desired admittance of Region I is
Y
D
= 1/b
d
=
5.0 and real admittance in Region I is about 4.983, which
close to the desired admittance value. This result shows
that admittance of human force to robot velocity can be
shaped to desired value, and the operator can feel the
master/slave-combined system as desired admittance.
2 3 4 5 6 7 8 9 10 11
-10
0
10
20
30
y (mm)
2 3 4 5 6 7 8 9 10 11
-2
0
2
4
6
time (sec)
Fy(N) [solid]
region [dash]
Fig. 12. Admittance display with different model
The second experiment shows the performance of the
virtual hard wall display. In this experiment we compare
the cutting result of MDF by autonomous position control
and master/slave-combined control with virtual hard wall
display. The cutting profile is simplified as a half circle.
The radius of the boundary is 20mm and cutting speed is
about 6mm/sec.
10 12 14 16 18 20 22 24 26 28
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
time (sec)
error (mm)
10 12 14 16 18 20 22 24 26 28
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
time (sec)
error (mm)
(a) Autonomous control (b) Virtual hard wall display
Fig. 13. Positional error at the surgical boundary
Fig. 13 shows the positional error at the surgical
boundary. The gap between the artificial hip implant and
the bone should be less than 0.3 to 0.5mm[13]. So the
surgical robot needs to have a position accuracy to less
than 0.3mm. By the autonomous control (12(a)), position
error is less than about 0.223 mm. In Fig. 12 (b) by the
virtual hard wall display, the maximum position error at
1127
the surgical boundary is 0.240 mm. From this result, it is
shown that the performance of the virtual hard wall
display is acceptable.
The third experiment shows how the master/slave-
combined manipulator cut an arbitrary profile by the
operator. The femoral bone of a cow is used and the
surgical boundary is defined as a circle with a 15mm
radius. Here, the admittance model of Region I is
b
d
= 0.4
[Nsec/mm] and the radius of Region I is 12mm. In Region
II, the admittance model is
b
d
= 1.0 [Nsec/mm] and k
d
=
0.05 [N/mm].
First, the operator cut around the surgical boundary and
then cut arbitrary spiral profile inside the surgical
boundary as shown in Fig 14. The cutting profile is made
well by the operator using the master/slave-combined
manipulator. In each surgical region, the operator feels a
different force and velocity relationship by a different
admittance model. In Region I, the human guide force is
relatively low and the cutting speed is high with respect to
Region II. At the surgical boundary, human guide force is
larger than that of other regions and the operator can feel
the surgical boundary and the position accuracy is
satisfied by the virtual hard wall display.
Fig. 14. Cutting profile
6. Conclusion
This paper presents an on-going development of a
surgery robot for total hip arthroplasty that includes a new
registration method and a control method. The proposed
gauge-based registration method simplifies the
registration procedure, and the parallel-type surgical robot
is mounted on the bone clamp, which is attached to the
femur. The surgical robot can be controlled along a pre-
programmed path; in addition, it can be directly controlled
by a surgeon with the master/slave-combined structure.
Preliminary experiments show that the proposed surgical
robot can be operated with an acceptable degree of
positional error.
Through experiment, it has been shown that a operator
can carve the femur of a cow into a desired profile,
changing the moving direction and velocity of the
proposed parallel-type surgical robot. Position accuracy at
the surgical boundary by a virtual hard wall display has
been shown to be acceptable.
Currently, plans are being made to design a more
compact second prototype for clinical test.
Acknowledgments
This work has been supported by the Human-friendly
Welfare Robot System Engineering Research Center
(HWRS-ERC), KAIST in Korea.
References
[1] Piston R, Engh C, “Osteonecrosis of the femoral head
treated with total hip arthroplasty without cement”,
J
bone Joint Surg AM
, 1994, 76A(2): 202~214
[2] Lee E H, Yoon Y S, Park Y B., “Study of surgical
robot registration method using block gauge”,
Proceedings of the first International Workshop on
Human-friendly Welfare Robotic Systems, KAIST, 2000
[3] J. Pransky, “Surgeons’ realizations of ROBODOC”,
Industrial Robot Vol.25 No.2 1998, pp105-108
[4] http://www.robodoc.com
[5] Meister, D.; Pokrandt, P., Both, A.,”Milling accuracy
in robot assisted orthopaedic surgery”,
Proceedings of
the 24th Annual Conference of the IEEE, Volume: 4 ,
1998 , 2502 –2505
[6] Yoon YS, Lee JJ, Kwon DS, Won CH, Hodgson AJ
and Oxland T, “Accurate Femoral Canal Shaping in
Total Hip Arthroplasty Using a Mini-Robot”,
Proc. Of
IEEE Int. Conf. on R&A, 2001, Volume 4,3214-3219
[7] Lee S H. Yi B J. Kwak Y K, “Optimal kinematic
design of an anthropomorphic robot module with
redundant actuators”,
Mechatronics, 1997, 7(5):
443~464
[8] Dong-Soo Kwon, Ki Young Woo and Hyung Suck Cho,
“Haptic Control of the Master Hand Controller for a
Microsurgical Telerobot System”,
Proc. Of the 1999
IEEE Int. Conf. On Robotics and Automation
, Detroit,
Michigan, 1999, May, pp.1722~1727
[9] C.L. Clover, “A Control-System Architecture for Robots
Used to Simulate Dynamic Force and Moment Interaction
Between Humans and Virtual Objects”,
IEEE Trans. On
Systems and Cybernetics,Vol.29, No. 4, November 1999
[10] J.E. Colgate, “Implementation of stiff virtual walls in
force-reflecting interfaces”,
IEEE Virtual Reality Annual
International Symposium, 1993
[11] S.C. Ho, R.D. Hibberd, B.L. Davies, “ Robot Assisted
Knee Surgery”,
IEEE Engineering in Medicine and Biology,
1995
[12] Joseph D. Bronzino, “ The Biomedical Engineering
HANDBOOK”, CRC PRESS, pp. 706, 1995
[13] L. Carlsson, T. Rostlund, B. Albrektsson and T.
Albrektsson, “Implant Fixation Improved by Close Fit.
Cylindrical Implant- Bone Interface Studied in Rabbits”,
Acta Ortho. Scand., Vol. 59, 272-275, 1988
1128
