129

A review of robotics in surgery

B Davies
Mechatronics in Medicine Group, Department of Mechanical Engineering, Imperial College of Science, Technology
and Medicine, Exhibition Road, London SW7 2BX, UK

Abstract: A brief introduction is given to the definitions and history of surgical robotics. The
capabilities and merits of surgical robots are then contrasted with the related field of computer assisted
surgery. A classification is then given of the various types of robot system currently being investigated
internationally, together with a number of examples of different applications in both soft-tissue and
orthopaedic surgery. The paper finishes with a discussion of the main difficulties facing robotic surgery
and a prediction of future progress.

Keywords: robotic surgery, computer assisted surgery, active robots, passive robots, safety, imaging,
pre-operative planning, registration, fiducials, haptics

1 WHAT IS A SURGICAL ROBOT? of the robot and tools to both the patient and to the
imaging.
Definitions of industrial robots vary widely. The Robot These benefits (and requirements) are equally applicable
Institute of America defines a robot as ‘A reprogrammable to the area called computer assisted surgery (CAS) (see
multifunctional manipulator, designed to move material, Section 3), but robots tend to provide greater accuracy
parts, tools or specialized devices through variable pro- and precision than CAS. However, it is mainly in their
grammed motions for the performance of a variety of ability to constrain the tools that robots are superior to
tasks’. A reasonable definition of a surgical robot would CAS. The surgeon holds the tools in CAS and could ignore
be ‘a powered computer controlled manipulator with arti- all warnings to the contrary and cut into unsafe regions.
ficial sensing that can be reprogrammed to move and pos- The robot, on the other hand, can be programmed to pre-
ition tools to carry out a range of surgical tasks’. It could vent motions into critical regions or only allow motions
be argued that this definition implies that a robot has a along a specified direction (e.g. in orthopaedic surgery, to
similar functionality to that of a surgeon. This functional drill an angled hole or cut to an inclined plane). Thus,
similarity is intentional. It is the externally powered com- provided the robot itself is considered to be safe, robots
puter controlled mechanism, with sensing and repro- could be said to enhance the safety of the procedure com-
grammable motions, that distinguishes the robot from pared with conventional surgery and to CAS. However,
both the related area of computer assisted surgery and the difficulty is that medical robots do not have generally
from the surgeon. Thus, although the general functions agreed safety recommendations. Industrial robots are
are similar to a surgeon’s, the properties that result are required to operate inside a cage, away from people, and
different. The general intention is that such robots should are only powered up when all personnel are excluded. This
not replace the surgeon, but that the robot should ‘assist’ is clearly inappropriate for surgical robots and agreed
the surgeon while under his/her supervision. international safety guidelines are urgently needed. The
One way for the robot to assist the surgeon is to carry author has made proposals for safety recommendations
out repetitive motions automatically, thus relieving the [1, 2] in an attempt to promote a consensus, since uncer-
surgeon of a tiring task (e.g. making small repetitive tainties over the needs for safety are causing robot sup-
increments of motion for diathermy of a region). Another pliers to be reluctant to provide commercial systems.
is for the robot to position tools very accurately at a pre- While the above definition of surgical robots requires
defined location or to move them with micromotions or that they be powered and under computer control, some
through a complex path. This means that the target tissue commentators include simple unpowered manipulator
must also be accurately defined and implies the need for arms as ‘robots’ [3]. These manipulators are a type of
accurate imaging, computer modelling and for registration localizer (i.e. a means of tracking tools), which is used
to hold the tool and point it in a particular direction.
The MS was received on 23 April 1999 and was accepted after revision With the addition of brakes, the manipulator can be
for publication on 22 October 1999. used to clamp the tool at a location. However, the
H01799 © IMechE 2000 Proc Instn Mech Engrs Vol 214 Part H

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

130 B DAVIES

inclusion of such manipulators as ‘medical robots’ can that time marketed by Unimation Limited. Shortly after
cause confusion with CAS ‘pointing’ devices or with this, the company was sold to Westinghouse Limited,
simple clamps and is, in the author’s view, unhelpful to who refused to allow the robot to be used for surgery
the concept of powered surgical robots. purposes on the basis that it was unsafe, since the indus-
Compared with CAS systems, the potential benefits trial robot was designed to be used inside a barrier away
available to a well-designed robotic surgery system are: from all contact with people. This position has continued
with the present owners, Staubli Automation Limited.
1. The ability to move in a predefined and reprogramm-
Thus, Kwoh’s work has ceased, in spite of the encourag-
able complex three-dimensional path, both accurately
ing preliminary results which indicated that, compared
and predictably.
with a conventional stereotactic frame, the robot could
2. The ability accurately and repeatedly to position and
position itself automatically and very accurately.
orientate at a reprogrammable point or at a series of
In parallel with this, Taylor at IBM was developing
points. While CAS systems may also have this ability,
an industrial robot system for hip surgery [5]. This was
robot accuracy is generally higher.
based on an IBM ‘Scara’ style of robot, used to hold a
3. The ability to make repetitive motions, for long
rotating cutter which reamed out the proximal femur to
periods, tirelessly.
take the femoral stem of a prosthetic implant for a total
4. The ability to move to a location and then hold tools
hip replacement. Following laboratory studies, the robot
there for long periods accurately, rigidly and with-
was used on dogs in animal studies and was subsequently
out tremor.
transformed into a ‘veterinarian robot’ for replacing the
5. The ability actively to constrain tools to a particular
hips of pet dogs under the direction of veterinary surgeon
path or location, even against externally imposed
Dr Hap Paul. The IBM robot was then replaced by a
forces, thus preventing damage to vital regions. This
Scara industrial robot from the Japanese company,
can lead to safer procedures than those achieved
whose Sanko-Seiki control system had been specifically
using CAS.
modified by the manufacturers to incorporate additional
6. To be able to move, locate and hold tools within
safety structures for surgery.
hazardous environments without damage to the sur-
It was not until late 1991 that the modified Sanko-
geon (e.g. from fluoroscopic or radioactive sources).
Seiki robot system, now called ‘Robodoc’, was tried
7. To be able to make precise micromotions with pre-
clinically on human patients. Prior to this, in 1988, the
specified microforces.
author’s group at Imperial College carried out labora-
8. To be able to respond and adapt very quickly and
tory studies into using a Puma 560 industrial robot for
automatically, either in response to sensor signals or
soft-tissue surgery, in the transurethral resection of the
to changes in commands.
prostate ( TURP) for benign prostatic hyperplasia [6 ].
9. To be able to perform ‘keyhole’ minimal access sur-
However, the shapes that had to be resected required
gery, without the aid of vision and without ‘forget-
two additional frameworks to be mounted on to the
ting’ the path or the location.
Puma robot, resulting in an eight degrees-of-freedom
It can be seen from the above that it is no longer appro- system. The robot was required to move the cutting tool
priate to speak of the benefits of robots over conven- during the procedure, actively to remove tissue. At this
tional procedures. Rather, the use of robots should be time, robot surgery was so new that no such ‘active
justified over CAS procedures. motion’ robots had been attempted and there was no
precedent for the approach. The author felt that the use
of an industrial robot designed to have a large envelope
2 INTRODUCTION TO THE HISTORY OF of motions was intrinsically less safe than that of a
MEDICAL ROBOTS special-purpose mechanism whose motions and forces
were designed specifically for the task. The concept of
It may seem strange to talk of the ‘history’ of a subject such a special-purpose robot has gained credibility since
that is as new as robotic surgery. Indeed, when the that time, because, in addition to being able safely to
author first started research into a robot for prostate apply limited forces and motions, a dedicated system has
surgery in 1988, the only other work in clinical progress the possibility of using simpler software. In spite of all
was that of Kwoh, who in 1985 first used a standard the advanced computational techniques used to generate
industrial robot to hold a fixture next to the patient’s safe software, ‘keep the program small and simple’
head to locate a biopsy tool for neurosurgery [4]. The remains a major key to software provability [7].
robot was locked in position, with power removed, while After the Puma feasibility studies for TURP, a manually
the surgeon used the fixture in order to orientate drills powered special-purpose framework was designed to
and biopsy probes, which were inserted into the skull remove the prostatic adenoma and was used clinically on
manually by the surgeon. Thus the robot was relegated 40 patients to check that the kinematics of the frame were
to the role of a traditional stereotactic frame in neurosur- appropriate [8]. Based on the kinematics of this frame-
gery. The robot used was a ‘Puma 560’, which was at work, a robotic motorized system was developed which
Proc Instn Mech Engrs Vol 214 Part H H01799 © IMechE 2000

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

 A REVIEW OF ROBOTICS IN SURGERY 131

was eventually used clinically in April 1991 and prior to such as small screws are often added to the patient at the
the clinical use of Robodoc. This was the very first time imaging stage. These markers are also available intraoper-
that an active robot had been used automatically to ativley and can be used for tracking. This process ‘regis-
remove tissue from patients. Since that time a second- ters’ the current patient/tracker reference system to that of
generation prostate robot (called ‘‘Probot’) has been devel- the pre-operative image and model. In orthopaedic sur-
oped at Imperial College [9]. This was mounted on a large gery, it is usually adequate to clamp the appropriate bone
floor-standing counterbalanced framework which could of the patient and assume that the target anatomy does
be locked in position using electromechanical brakes. This not move, so that only a position sensor on the clamp is
seems to be the first robotic surgery application of a now required to act as a warning if the patient moves. By this
widely accepted concept in which a large ‘gross pos- technique the pre-operative models of the patient are
itioning’ system is used to support and move a smaller treated as fixed during the procedure, and it is only neces-
robot. The large positioner, often a passive manipulator, sary to superimpose on to them the current position of the
can be moved over a wide region and be locked in position cutting tools. At the pre-operative planning stage, the sur-
at the approximate location while the smaller, purpose- geon can take time to check that the information in the
built system carries out the task. Although resulting in computer display, showing the location of the target tissue
some redundancy of motion, this technique enables the and tools, is correct. However, for soft tissue (and other
powered robot to be small and to be designed with limited ‘compliant’ parts of the anatomy, such as in the spine), a
motions and forces just adequate for the tasks, which helps process called ‘dynamic referencing’ is used for intraoper-
to ensure both safety and accuracy. ative tracking of the moving tissue in real time and updat-
Robotic surgery systems have been slow to develop, ing of the computer database to provide the current target
and, at this time, Probot is one of only few soft-tissue location. For example, when drilling into a vertebra in the
surgery active robots to have been applied clinically [10]. spine to fix pedicle screws, drill forces often distort the
One reason for this slow rate of application has been the location of the vertebra relative to its neighbours. In CAS
parallel development of CAS, which can be seen to give systems, the drill location is monitored and displayed on
some of the benefits of robotics without the same degree a computer together with the relative location of the ver-
of concern for safety. Thus, no review of robotic surgery tebra. A separate tracking monitor on the vertebra updates
would be complete without also mentioning the parallel the display with its new location as it distorts during drill-
developments in CAS. ing. However, if, for example, the vertebra motion sensor
slips, it will display a false reading which shows motion of
the vertebra, with no time to check if the location is cor-
3 COMPUTER ASSISTED SURGERY (CAS) rect. Thus, a degree of ‘trust’ is required that the dynamic
referencing is correct, and this is why dynamic referencing
The main difference between the terms ‘robotic’ and ‘com- is potentially one of the most safety-critical areas of CAS.
puter assisted’ surgery is that robots are moved by some As discussed in Section 1, passive manipulators can also
sort of motorized system while computer assisted systems be used to carry and track tools. Unlike camera-based
are generally manually powered by the surgeon. In sur- ‘localizers’, manipulator arms have the problem that they
gery, the majority of computer-based systems are tracking can be cumbersome and restrict the surgeon in the free
systems. These may be used to track tools or parts of the motion of the attached tools. The use of a manipulator,
anatomy, either using a sensor-based system or by clamp- however, can help damp out unwanted surgeon tremor
ing the tool onto a manipulator arm whose joints are and, with the addition of electromagnetic brakes, can be
monitored for position. The sensor-based systems usually used to lock the tools in position while, for example,
use an array, either of light emitting diodes (LEDs) or of X-rays are taken. Camera-based systems which are both
optical reflectors, attached to the tool. The position and accurate and have a wide field of view are generally
orientation of the array in three-dimensional space can be expensive. They also have the problem that they can cease
tracked by a group of cameras. The tool and its three- to function when the surgeon leans over the patient and
dimensional coordinates can then be represented on a obscures the view of the target LEDs when seen from the
computer screen in relation to the coordinates of the target camera. However, other types of tracking system can also
anatomy which is also represented. To represent the target be inaccurate, such as those using electromagnetic coils
location in the computer, it is necessary for the appropriate (which can be rendered inaccurate if ferrous materials are
anatomy to undergo preliminary pre-operative imaging present) or ultrasound-based range finders (whose values
(usually computer tomography (CT ) or magnetic reson- can vary with environmental temperature). Most sensor
ance (MR) imaging). These three-dimensional scans are and computer systems are more susceptible to inaccuracies
used to form a three-dimensional model of the anatomy in the operating room, e.g. owing to the presence of elec-
intraoperatively. Recognizable features in the anatomy tromagnetic interference from sources such as the dia-
can then be located by the tracking system to ‘register’ the thermy used for cutting and cauterization. Advocates of
tracker to the patient anatomy. If no obvious anatomical such systems suggest arranging the operating room (OR)
markers are available, artificial markers (or ‘fiducials’) to exclude all sources of distorting influence. However, in
H01799 © IMechE 2000 Proc Instn Mech Engrs Vol 214 Part H

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

132 B DAVIES

practice it is difficult to ensure this, particularly in emer- Table 1 Typical stages in robotic knee surgery
gencies, and so safety is likely to be compromised.
Pre-operatively:
The CAS systems, unlike robots, rely on the surgeon Image patient
for motive power. However, they too are vulnerable to Edit images and create three-dimensional model of leg
hardware and software errors in the data provided by the Create three-dimensional model of prostheses
Superimpose prostheses over three-dimensional model of leg
tracking systems for the tools and tissue, in which case Adjust and optimize location
it is necessary for the surgeon to detect that there is a Plan operative procedure
problem and to take corrective action or stop the proced- Intraoperatively:
Fix and locate patient on table
ure. In addition, just as for robotic surgery, most CAS sys- Fix and locate robot (on floor or on table)
tems use a pre-operative planning system. This allows the Input three-dimensional model of cuts into robot controller
surgeon to take images of the patient, form them into three- Datum robot to patient
Carry out robot motion sequence
dimensional models and display them on a computer to- (Monitor for unwanted patient motion)
gether with the various tool locations. The surgeon can Post-operatively:
then simulate the whole procedure and ensure that the pro- Remove robot from vicinity
Release patient
posed protocol is correct, removing a lot of the worry and Check quality of procedure
strain from the actual operation. The safety issues for pre- If further cuts are necessary:
operative planners in both CAS and robots are broadly Reclamp patient
Reposition and datum robot to patient
similar. For CAS, just as for robot surgery, the accurate Repeat robotic procedure
registration of the pre-operative three-dimensional models
to the intraoperative position of both the patient and the
tools being tracked is important to ensure safety. ally by touching the robot tip to the markers. These same
The surgeon computer interface, the associated software fiducials will have been observable in the pre-operative
and the underlying assumptions built into the algorithms imaging and three-dimensional models, and so this process
all have a major impact on safety of CAS, as they do can register the current patient fudicial location to that on
for robotic surgery. Thus, although at this time CAS is the pre-operative images and models, as well as to the
considered safer than robotic systems, it is probable that intraoperative robot location. The fiducials are usually
the inherent safety issues and problem areas for CAS are small screws inserted into the bone in orthopaedic surgery
actually not significantly different. It is the perception that or are small discs stuck to the skin, e.g. over boney promi-
surgeons are more likely to take responsibility for a nences in neurosurgery. The more recent use of anatomical
CAS procedure, compared with using an autonomous features for registration can avoid the need for artificial
robot, that has tended to make the passive CAS systems markers, but they need to be carefully applied. This is
more favoured by equipment developers at this time. because the robot is touched on to 20 to 30 points on the
anatomy at the time of surgery, and the points are then
used to generate surfaces that are matched with surfaces
4 ROBOTIC SYSTEMS in the pre-operative model. Since the interpolation of
points and surface matching is a statistical process, the
As we have seen, the robot is only one aspect of an results need to be applied carefully to maintain accuracy.
integrated surgical system. Such systems have three Whether artificial or natural markers are used, the overall
phases: registration process is one of the greatest sources of error
in both CAS and robotic procedures.
(a) pre-operative planning,
In order to ensure that the robot is being correctly
(b) intraoperative intervention,
applied, an intraoperative display of robot motions is
(c) post-operative assessment.
required to guide the surgeon. This should show a three-
Table 1 shows a typical sequence for robotic total knee dimensional schematic of the correct position of the tool
replacement. It is only in the intervention aspects of the (together with the planned extremes of tool motion),
intraoperative phase that the robot is of direct benefit. As superimposed over simplified views of the tissue that has
outlined in the previous section, the pre-operative plan- been removed and of that remaining. These simplified
ning phase is also necessary for CAS procedures. schematic views are necessary for real-time viewing of
However, when a robot is to be used, the planning aspect often complex motions. More complex views, e.g. of sur-
can also include a computer simulation sequence of the face or volume rendered images of the tissue, should be
robot motions. When the surgeon is satisfied that the provided on separate displays. The robotic display needs
sequence is correct and the robot will not impinge on the to be kept to simple schematics, with only basic robot
patient or adjacent equipment, then the motion sequence parameters on the screen, so as not to confuse the sur-
can be downloaded directly to the robot controller. geon in an emergency. Full diagnostics, however, should
In the intraoperative phase, it is necessary to fix the be available on the screen when needed, so that, say,
robot with reference to the patient and then ‘register’ the when a procedure is interrupted, the full status is avail-
robot to specific markers or ‘fiducials’ on the patient, usu- able to judge if it is safe to continue or if it is first
Proc Instn Mech Engrs Vol 214 Part H H01799 © IMechE 2000

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

 A REVIEW OF ROBOTICS IN SURGERY 133

necessary to re-register the robot to fiducial markers. In in Grenoble, France, where an industrial robot was fitted
an emergency, it may be necessary to abort the robotic with additional large-ratio gear boxes so that the robot
procedure and it must be ensured that at all times it is could move slowly and safely. The addition of a pre-
possible to finish the surgery using a safe manual operative planning facility based on CT imaging has
procedure. made this a powerful system. Recently, a special-purpose
An assessment phase is usually required immediately robot called ‘Neuromate’, to be used ‘passively’, has been
post-operatively. This requires that the robot can be developed commercially by IMMI Limited, Lyon,
readily removed and the patient unclamped so that the France, and is integrated with the planner for neuro-
patient can be moved around. Rapid robot removal is surgery. These systems have the potential to give a more
also essential for safety reasons, so that if the robot mal- stable platform and be more accurate for deep-seated tum-
functions, it can be quickly removed and the procedure ours than equivalent camera-based localizers or localizers
completed manually. Should the assessment show that based on unpowered manipulator arms. However, they do
further action is required, it will be necessary to reclamp tend to be more costly than their CAS equivalent.
the patient and reposition and re-register the robot. This
implies that any fiducial markers should remain in pos-
ition throughout and should not have been machined 5.2 Active robots
away by earlier robot actions. The use of a powered robot actively to interact with the
patient can potentially allow more complex motions than
5 CLASSIFICATION OF SURGICAL ROBOTS the above example of a powered robot used passively.
However, safety concerns are greater, and for this reason
most active robots have been developed specifically for
While it is possible to classify robots according to the
the task.
surgical tasks for which they are intended, it is helpful
first to define the technology basis for the different types. 5.2.1 Laparoscopic camera robots
A major division is whether the powered robot is used
in a passive, power-off mode or in an active mode for Probably the largest sales of a commercial system for
active movement of the tools to perform the surgery. robot surgery have been in the area of the manipulation
of laparoscopes, mostly for abdominal, ‘minimally invas-
ive’ surgery [12, 13]. Traditionally, a surgeon uses an
5.1 Powered robots used as passive tool holders assistant who moves the laparoscopic camera and tries
Some of the earliest applications of powered surgical to anticipate what the surgeon wishes to view. The
robots were to use them passively, as a means of holding cramped surroundings and inability to predict the sur-
fixtures at an appropriate location, so that the surgeon geon’s needs often makes this a fraught task. However,
could insert tools into the fixture. The early work of Kwoh some would argue it is ideal training for future ‘minimal
et al. [4] (briefly mentioned in Section 2) used an industrial access’ surgeons. Nevertheless, commercial robots have
Puma robot in this way to position a fixture next to the been developed for this task. The requirement for a robot
head so that a surgeon could insert drills and biopsy to hold and manipulate the laparoscope is very
needles at a desired location for neurosurgery. The patient demanding for abdominal procedures, since a robot
wore a standard neurosurgery stereotactic frame which must move the laparoscope (typically providing pitch,
was also used as part of a pre-operative CT scan. On the yaw, roll and in/out motions) about a remote pivot point
day of the operation, the tip of the robot was able to be located at the abdomen wall. To avoid obscuring the
touched on to the stereotactic frame, thus ‘registering’ the operation site, this means that the robot requires a
robot to the patient, and at the same time registering to ‘remote centre’ motion about the entry point, with a long
the pre-operative CT scans and the three-dimensional power transmission mechanism linking the laparoscope
models of both the brain and the target tumours. The to the powered robot. The need for a small ‘footprint’
robot was then moved slowly to the desired position at of the mechanism within the operation site further com-
which an entry hole in the skull could be located, and plicates the design. Input commands for the robotic
locked in position with all power removed to make it safe. camera can be achieved by the surgeon using either foot
The surgeon then used the locating fixture, at the robot pedals or (more recently) head motion sensors or by
tip, to orientate a drill to produce the entry ‘burr’ hole and voice control. Because the robot is not used to cut
then to insert a biopsy probe to make a straight line of directly or to move cutting tools, the motions are not
access into the tumour. Thus, the surgeon’s actions were considered as potentially dangerous and safety concerns
simple and limited to straight-line insertions and axial in the use of this application are much reduced.
rotations. However, the unmodified industrial robot could 5.2.2 ‘Robodoc’ orthopaedic surgery
be said to be used safely, since it was unpowered and
locked in position during the surgical procedure. A similar A further active robot that is available commercially is the
approach was subsequently taken by Lavallee et al. [11] ‘Robodoc’ hip surgery robot from Integrated Surgical
H01799 © IMechE 2000 Proc Instn Mech Engrs Vol 214 Part H

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

134 B DAVIES

Supplies Limited, Sacramento, United States [14] (as procedures. Once the patient is clamped, the hip is
briefly mentioned in Section 2). The robot is instrumented opened by the surgeon and the femoral head removed,
with force sensing on all axes, as well as using a six-axis as in conventional surgery. At this point the robot tip,
force sensor at the wrist. The tip of the robot carries a carrying a high-speed rotary cutter mounted on a force
high-speed rotary cutter which can accurately ream out sensor, is moved into the appropriate position on the
the femoral cavity for the stem of a particular hip implant femur head and the sequence of motions executed to
(see Fig. 1). A separate pre-operative planner called resect the appropriate shape for mounting the implant
‘Orthodoc’ can be used, which allows a computer model stem. The sequence of motions can be displayed simul-
of the appropriate size and shape of implant to be pos- taneously on a computer to ensure that all is well. Force
itioned over a three-dimensional model of the hip, recon- levels from each joint, as well as the wrist sensor, are
structed from a series of CT scans. The position and also monitored for safety and the procedure is halted if
orientation of the implant can be adjusted until the sur- forces rise above a predefined level.
geon is satisfied. The resulting femoral cavity can then be An important step, as in all CAS and robotic surgery,
displayed and the sequence of robot motions automati- is the ‘registration’ of the pre-operative MR, CT or ultra-
cally generated so that the surgeon can ensure that the sound scans to the intraoperative location of the patient
procedure will cause no difficulties. bone, as well as the current intraoperative position of
Once the planning has been completed, the intraoper- the robot. In hip surgery, this is generally achieved by
ative phase begins with the patient’s leg being clamped embedding ‘fudicial’ markers into the bone in both the
to a rigid framework mounted on the pedestal of the proximal head of the femur and the distal femoral con-
robot. A further clamp holds a pin located in the femoral dyles, so that their coordinates show clearly in the pre-
head so that any motion greater than 2 mm of the leg operative CT scans and three-dimensional models. The
relative to the robot stand can automatically halt the markers have a conical recess into which a ball can be
procedure. In this way the femur is treated as a fixed, located. The ball is held on the end of the robot arm
static object, in which the predefined motions (planned and positioned into the cone under force control to
pre-operatively) can be executed. This is a much simpler ensure repeatability. Thus, the fiducial location on the
procedure than say, soft-tissue surgery, where tissue pre-operative CT scans is registered to the current
motions may require intraoperative adaptation of the patient position and also to the robot coordinate system.
Because of patient complaints of pain from the knee
fiducials, attempts are being made to replace them by
using anatomical features as markers. This is achieved
by touching the robot tip to a series of 20–30 closely
related boney points. A surface map of the points is
statistically generated and matched to the pre-operative
model of the surface. This has two problem areas:
1. The statistical matching of surfaces is prone to error,
which is exaggerated at surfaces far from the
located points.
2. The exact location of the bone surface (as distinct
from soft tissue) as probed by the robot can be in
error when compared with the CT scanned surface.
Thus, although anatomical markers are gradually
being introduced, fiducial markers still remain the
‘gold standard’.
The Robodoc system underwent trials at three clinical
centres in the Untied States between 1991 and 1994 in
an attempt to satisfy the needs of the Federal Drugs
Administration (FDA), which requires that clear clinical
benefits be shown before the use of expensive technology
can be sanctioned. The difficulty is that the claimed ben-
efits for robotic procedures are good alignment of the
implant stem in the femur and a very good contact area
between bone and stem (better than 98 per cent for the
robot, compared with typically 23 per cent by conven-
tional manual surgery). Both benefits are claimed to give
improved long-term performance of the prosthesis as
Fig. 1 ‘Robodoc’ hip surgery robot well as improved bone growth. Such benefits would
Proc Instn Mech Engrs Vol 214 Part H H01799 © IMechE 2000

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

 A REVIEW OF ROBOTICS IN SURGERY 135

require a 10–15 year period to be demonstrated. Short- a localized endoscopic view is available and the sequence
term benefits, however, were more difficult to demon- of motions has to be ‘remembered’ in order to locate the
strate as the time for the procedure was longer, resulting resectoscope tip within the gland. Also, a number of fea-
in increased anaesthesia times and increased blood loss. tures must remain unharmed to avoid impotence and
Also, post-operative patient pain was reportedly greater incontinence. Although prostatectomy is a soft-tissue sur-
owing to the use of the knee fiducials. gical procedure, it is largely a ‘debulking’ process not
In the summer of 1994, Robodoc was introduced to requiring high accuracy. Also, the prostate is held rela-
Frankfurt Hospital, Germany, where a large number of tively immobile by the pelvic anatomy. It can thus form
operations have been conducted (over 2000 to date). an ideal procedure for robotic soft-tissue surgery.
This has resulted in improvements in protocols, so that, As mentioned in Section 2, the Mechatronics in
even though the robot is substantially unchanged, times Medicine Group at Imperial College has been concerned
for the procedure have been considerably reduced. This with the development and clinical implementation of an
indicates the dangers of long-term assessment of CAS active robotic system for prostatectomies, called ‘Probot’
and robotic surgery during the early years of implemen- [9]. This project started in 1987, with an approach by
tation, when both hardware and protocols are rapidly the Institute of Urology in London to ask if a robot
changing. Increased patient demand has now led to the system could be developed for resection of the pros-
introduction of 28 Robodoc systems in Europe. tate. Following preliminary feasibility studies, a special-
Recently, 250 pinless registration procedures, using a purpose ‘safety frame’ was developed to give the required
separate digitizer to locate anatomical features, have motions with the minimum degrees of freedom. This was
been successfully performed in Frankfurt [15]. It is manually powered and was tried clinically on forty
hoped that, when this experience has been further con- patients with good results [8]. Having proved the kin-
solidated, applications for FDA approval will be made ematics, the system was powered under computer control
to allow this system to be used in the United States. and applied clinically in 1991 to five patients (see Fig. 2).

5.2.3 Additional orthopaedic systems

Another recent commercial system, initially aimed at hip
implants, is called ‘Caspar’ by Orto-Maquet [16 ]. This
utilizes a robot based on an anthropomorphic Staubli-
Automation industrial clean-room robot, which has been
fundamentally modified for orthopaedic surgery. The
system has been used on 75 patients in the Erlangen
University Hospital. It is perhaps not surprising that the
two commercially available active robot surgery systems
have been developed for orthopaedic use in the hip, where
the bone can be treated as a fixed, clamped object to
which pre-operative imaging can be applied, with none of
the concerns of tissue motion and distortion that are inher-
ent in soft-tissue surgery. Other recent research projects
using robots for orthopaedic surgery include Rizzoli
Orthopaedic Institute, Bologna [17], which has used a
Puma 560 robot, and Helmholtz-Institute, Aachen, which
is developing a special-purpose parallel link robot for hip
surgery [18]. A robot is ideal for orthopaedic surgery
since it can generate the high forces needed to create
accurate cuts, even though the bone resistance can vary
widely. The constrained robot will also not bounce off
hard surfaces and cut into vulnerable soft tissue.

5.2.4 ‘Probot’ prostatectomy robot

The reduction in urinary flow owing to a benign adenoma
blocking the urinary duct is a common problem in males
past middle age. The usual treatment is to remove the
adenoma using a ‘hot wire’ diathermic loop resectoscope.
This is passed down the centre of the penis and is used to
chip away the adenoma. This minimally invasive pro- Fig. 2 Imperial College ‘Probot’ prostatectomy robot being
cedure is difficult to learn as, like all such procedures, only clinically applied
H01799 © IMechE 2000 Proc Instn Mech Engrs Vol 214 Part H

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

136 B DAVIES

required. This type of remote centre motion, beneficial

for minimal access surgery, is also being used by the
present author (in conjunction with Fokker Control
Systems BV, the Netherlands ) in a special-purpose four-
axis robot for neurosurgery (called Neurobot) which is
funded by the European Commission as part of a simu-
lation, imaging and robotic surgery project called
‘Roboscope’ [20].

5.2.5 ‘Minerva’ neurosurgery robot

A further example of an active robot is ‘Minerva’ which
has been applied clinically for neurosurgery [21] (see
Fig. 4). This is a novel special-purpose system developed
by the precision mechanisms group at the University of
Fig. 3 Johns Hopkins University remote centre motion Lausanne. A powered robot, in association with a dedi-
(RCM ) robot designed for kidney puncture cated CT imaging system, has been used in limited
neurosurgery clinical trials. The robot system employs a
series of special tools, located on a rotary carousel, each
This was the first time that an active robot had been
of which can then be locked into position on a single-
used to remove tissue from a patient, preceding the
axis travel. This single axis then advances the tool lin-
Robodoc clinical human trials by some five months. A
early into the region of the patient’s head which is
subsequent Engineering and Physical Sciences Research
datumed to a stereotactic frame. These actions take place
Council ( EPSRC ) UK Government grant for laboratory
adjacent to the CT machine to ensure easy intraoperative
studies gave the opportunity to add a transurethral ultra-
imaging. The robot is extremely accurate, with an overall
sound probe to the robotic frame for direct measurement
positional accuracy, including CT imaging, of just under
of the gland size at the start of the procedure [9]. The
1 mm. It will be some time before the clinical benefits of
probe could be interchanged with a diathermic cutter to
Minerva’s increased accuracy will be seen clinically to
remove the prostatic adenoma. The ultrasound images
justify the need for a dedicated CT scanner and the
are marked up by the surgeon to identify the tissue to
increased cost and complexity compared with, for
be removed. These ‘slices’ are then built into a three-
example, a computer assisted surgery ‘localizer’ system.
dimensional model of the resectable volume, which is used
to generate the cutting trajectories for the robot. Clin-
ical trials of the new prostatectomy robot have been car- 5.3 Synergistic systems—the ‘Acrobot’ active constraint
ried out at the Minimally Invasive Therapy Unit at Guy’s robot
Hospital, London, with very good results. It has thus
been shown that a fast, accurate and safe prostatectomy A novel control system for robotic surgery is being
can be carried out robotically. The anatomy of the pros- implemented at Imperial College, London, for prosthetic
tate minimizes motion of the soft tissue, as does careful implant knee surgery [22]. This system will allow the sur-
selection of the cutting protocol. This, together with the geon to hold a force-controlled lever placed at the end of
fact that the prostatectomy is primarily a debulking pro- the robot which also carries a motorized cutter. The sur-
cess, for which great accuracy is not required, has meant geon can use the lever to back-drive the robot motors
that imaging at the start of the procedure is adequate, within software constraints provided by the robot so that
in spite of this being a soft-tissue procedure. Further an appropriate shape is machined into the knee bones. This
research is being undertaken to ensure intraoperative programmable software constraint system gives rise to the
imaging of soft tissue distortions. concept of an active constraint robot (known as ‘Acrobot’)
A recent innovation by a group at Johns Hopkins which is shown in Fig. 5. Within a central predefined
University in the Untied States is to provide a small but region, low-force control is provided. This control strategy
versatile robot for low-force procedures such as kidney allows the surgeon to feel directly the forces experienced
biopsy [19] ( Fig. 3). Such systems need to have a remote by the cutter. Thus, if the surgeon cuts a hard piece of
centre of motion in pitch and yaw about the point where bone, the forces that are experienced rise and he can slow
the tool enters the skin. This could be provided by down or take a lighter cut. The force-controlled handle is
software through a compound motion of several axes. supplemented with a ‘deadman’s handle’ switch which,
However, it could be said to be safer to provide a power when released, can automatically bring the robot and cutter
transmission system where kinematics are arranged to to a safe state. Towards the edge of the low-force region,
provide the pitch and yaw motions from two dedicated the robot impedance gradually increases until, at the limit
motor axes. A further in/out motion and tool rotation of the permitted region, the control system switches into
about the pitched/yawed axis complete the four axes high-gain position control. Thus, the robot gives an active
Proc Instn Mech Engrs Vol 214 Part H H01799 © IMechE 2000

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

 A REVIEW OF ROBOTICS IN SURGERY 137

Fig. 4 ‘Minerva’ neurosurgery robot in position, with the patient adjacent to a CT scanner

constraint within an accurately preprogrammed area, pro-

viding accuracy as well as avoiding damage to vulnerable
areas, while the surgeon stays in control of the procedure.
It is felt that this strategy will be more acceptable to the
surgeon than conventional position control of an auto-
mated active robot. A series of phantom and cadaver trials
has demonstrated the accuracy of the system and its ease
of use. A pre-operative planning system, based on a low-
cost PC, provides a simple method for planning where to
place the appropriately sized prosthesis.
Acrobot represents a new type of robotic system for
surgery, known as a ‘synergistic’ system, in which the
surgeon’s skills and judgement are combined with the
robot’s constraint capabilities to form a partnership that
enhances the performance of the robot acting alone. A
variation of this concept has also been applied by a
French group who have produced a passive arm system
that uses a series of motorized clutches to allow motion
[23]. In this instance, the motorized clutches allow the
surgeon only to move the manipulator in a prepro-
grammed direction. Since the arm motions rely totally
on the surgeon to move them, and the power is used
only in the clutching mechanism and not for powering
motions, the system (called PADYC, after Passive Arm,
Dynamic Control ) is said to be safer than an active
robot. However, the fact that motorized clutches have
to be switched on and off many times a second can imply
that this will not be an easy mechanism to provide Fig. 5 Imperial College ‘Acrobot’ knee surgery robot with
smooth three-dimensional control. force-controlled handle
The process of using technology to aid in surgery is
primarily one of integration. It is only when robotic and systems to be used in the operating theatre, so that the
CAS mechanisms are included in a total system within total system with its imaging, modelling, sensing, regis-
the operating theatre that their viability can be correctly tration and motion mechanisms (all suitably sterile) can
judged. Thus, there is a considerable need for integrated be tried out using an appropriate ‘human/computer
H01799 © IMechE 2000 Proc Instn Mech Engrs Vol 214 Part H

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

138 B DAVIES

interface’ for the surgeon in a clinical setting. Even then, One type of autonomous robot that operates in a soft-
the complexity of the system in the operating theatre tissue, semi-disordered environment is a colonic crawler
environment means that a number of development or ‘inch-worm’ robot. This is used to inspect and sample
changes will inevitably be required to perfect the system. the colon for possible disease. It is generally based upon
This is a relatively new requirement for medical systems, a worm concept in which a concertina segment advances
and new funding mechanisms are needed internationally along the colon and attaches itself to the wall, usually
for these integrated robotic systems in order to enable by expansion or suction. A second section is advanced
medical and engineering personnel to communicate and to the first and then in turn anchored. The first section
work together to develop the equipment to an appro- is detached and the process repeated. This sequential
priate level. Only then can the efficacy of the robotic or process is usually pneumatic, under computer control.
CAS systems be correctly evaluated. The flexibility and variable structure of the colon require
a number of sensors and adaptive control. In order to
cope with sharp bends, more than two segments are usu-
5.4 Master–slave ‘telemanipulator’ systems ally required. Among a number of variants of this device
that are under investigation, the work of Professor Ng
Acrobot, which uses a force-controlled lever moved by at Nanyang University, Singapore, is unusual in using a
the surgeon, can be regarded as a type of master–slave number of miniature ‘feet’ to grip the colon wall and
system in which the master (the force lever) is, unusually, negotiate bends without slipping. This device has been
attached to the slave (the moving robot structure). used successfully on live pigs [27].
However, for these telemanipulator systems (sometimes
called telepresence) it is more usual to mount the slave
separately from the master. The master may consist of
a simple joystick input system or, more usual for surgery, 6 CONCLUSIONS
may be a kinematic mimic of the slave robot. It is poss-
ible to locate the master many miles from the slave, and In this brief overview, it has not been possible to cover
have a connection via high-speed telephone line or a all aspects of the rapidly developing area of robotic sur-
satellite link. Such systems have been proposed for sur- gery. As we have seen, the various types of surgical robot
gery, but it is more likely that they will find more can carry out all the tasks that can be performed by
immediate application in diagnostics, where the ability CAS systems. In addition, robots have the very useful
to transmit a sense of ‘feel’ remotely will be of value. In property of being able to constrain and guide surgical
surgery, however, it is possible also to place the master interventions in a way that is not possible with normal
controller nearby, alongside the slave in the OR. This CAS systems. Robots have the potential to be auton-
will permit the use of scaled motions so that large move- omous and to carry out repetitive actions tirelessly, as
ments of the master will result in micromotions, with well as move through complex paths with considerable
small forces, applied by the slave. Two examples of this accuracy. However, since they tend to involve additional
are the ‘da Vinci’ system being developed by Intuitive components for the system, the use of robots will inevi-
Surgical Incorporated [24] and the ‘Zeus’ system of tably make the equipment more costly and complex than
Computer Motion Incorporated [25], both of which are CAS systems. This cost and complexity will be easier to
being used clinically for minimally invasive ‘closed’ heart justify in those procedures where the benefits of robotic
surgery. In both systems, a robotic arm carries an endo- interventions provide a clear advantage over CAS. Thus,
scope while two other manipulator arms carry inter- just as it is difficult in some procedures to justify the use
changeable tools, such as scissors and grippers. An of CAS as compared with conventional surgery, so there
innovative feature is the ‘wrist’ inside the body, which will be specific procedures that can justify the use of
can angle tools. This feature is of particular value in robotics as compared with CAS. When considering the
tying knots for sutures inside the body. However, at this different types of robotic system, there are immediate
time there is no sense of feel fed back from the master benefits in using robotic systems passively; e.g. the ability
to the slave, and the surgeon relies upon the high-quality safely to lock off a relatively unmodified industrial robot
endoscopic vision for monitoring the process. This sense so that it can be used as a guiding fixture by a surgeon.
of feel or ‘haptics’ is a complex issue at the forefront of This limited role for ‘passive’ robots will be less attract-
research and requires force-sensing systems at the slave ive once the safety requirements for ‘active’ medical
to apply appropriate feedback forces to the master and robots have been agreed. Active robots, which perform
hence to the surgeon [26 ]. A realistic sense of feel, how- autonomous interventional actions while being super-
ever, requires more than simple force information. Rates vised by the surgeon, are likely to have a healthy future.
of change in force and motion, as well as their inter- It should be emphasized that it is not envisaged that
action, are equally important in determining such aspects these robots will be used to ‘automate’ a procedure with-
as tissue ‘texture’. The best way to input this information out the surgeon being present. They will be assistive
back to the surgeon is also a research topic. devices augmenting the capabilities of the surgeon. They
Proc Instn Mech Engrs Vol 214 Part H H01799 © IMechE 2000

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

 A REVIEW OF ROBOTICS IN SURGERY 139

may be industrial-style robots, which will need to be a safe region or to an accurate plane, a path or a
extensively modified for safety by the manufacturers, or location. The surgeon thus uses his innate sensing and
special-purpose devices configured for individual tasks. judgement while the robot constrains, providing safety
It is the present author’s view that the special-purpose and quality. This synergy between the best robot and
systems are likely to be lower cost, smaller, simpler and surgeon qualities has considerable potential for both
easier to make safe. There is a worrying tendency for soft-tissue and orthopaedic surgery. Recent develop-
some research workers to purchase standard industrial ments in imaging will benefit CAS and robotic surgery.
robots, on the basis that these are the same as those used The lower costs and higher definition of both MR
in surgical systems. However, while the kinematics may and CT imaging, as well as the availability of three-
be similar, the surgically approved versions have extens- dimensional ultrasound imaging with good resolution,
ive modifications to allow their safe use next to people. have improved information about the target tissue
The much lower-cost industrial versions could be used location. Developments in imaging systems and in endo-
in the laboratory to demonstrate the kinematics and inte- scopes and cameras have meant that there has been a
gration concepts prior to use in the OR, but even in this preponderance of vision-based sensing, associated with
environment the safety of research personnel must sensing position. Other senses, such as haptics, have been
remain paramount. much neglected and are an area of current research. The
Specific procedures that will benefit from robotic inter- use of the Acrobot concept is a way of supplementing
vention in the near future are various orthopaedic cut- currently poor artificial sensing with the surgeon’s own
ting and drilling procedures, where the forces generated innate sensory capability. This ‘hands-on’ robot forms
can be resisted by the robot, preventing the cutters from an intermediate type of robot which bridges the gap
bouncing off hard bone and damaging other areas, such between autonomous systems such as Robodoc and
as soft tissue. In addition, the robot will provide con- Probot and the master–slave telemanipulator such as
siderable accuracy to the cuts, which will often need to Zeus.
be made repetitively. Orthopaedics is a good application The future of robotic surgery will ultimately depend
area since, once the bone is clamped, it can be treated not just on technology but also on the abilities of the
as a fixed object. Robotics will also be of benefit in soft- engineers, computer scientists and medical physics
groups to communicate and collaborate effectively with
tissue surgery, particularly in minimally invasive ‘key-
medical personnel. The engineering and medical disci-
hole’ procedures. The use of a robot will overcome many
plines are very different in training and orientation. The
problems of visualizing where the tips of the tools are
effective application of robotic surgery can only be
located, which is such a problem with conventional
achieved with understanding, dedication and enthusiasm
endoscopic techniques. Many of these procedures also
from all personnel. The history of robotic surgery is now
require a remote centre motion, which adds complication
just over a decade old. Developments in intraoperative
to the design of the robot. A particular type of keyhole
imaging, microsurgery and in sensory perception
surgery that will benefit from the use of robots is in
additional to vision (such as haptic sensing) will con-
neurosurgery, where the needs of a precise path and a
siderably change robotic surgery by the year 2010.
final precise location of the tool are both critical. Here,
because of the need to target features such as tumours
and track them as they distort and move during the
intervention, it is essential to have image guidance intra-
operatively, at least intermittently but preferably con-
REFERENCES
tinuously. This tendency for soft tissue to move when
pressed or cut and to change shape makes robotic soft-
1 Davies, B. L. A discussion of safety issues for medical
tissue interventions particularly difficult. However, in
robots. In Proceedings of 2nd International Workshop on
some procedures, such as prostatectomy, the anatomy Computer Assisted Robotic Medical Interventions, Bristol
can be sufficiently constraining that the need for continu- (Eds A. Di Gioia, T. Kanade and P. N. T. Wells), June
ous imaging can be reduced. There is also considerable 1996, Appendix H (Ctr. Ortho. Res. Shadyside Hosp.,
potential for telemanipulator master–slave systems in Pittsburgh, Pennsylvania).
soft-tissue surgery, particularly where forces or motions 2 Davies, B. L. The safety of medical robots. In Proceedings
can be scaled down for microsurgery. of 29th ISR Conference on Advanced Robotics: Beyond
A potentially beneficial type of robotic system is that 2000, Birmingham, April 1998.
3 Taylor, R. Robots as surgical assistants. Lecture Notes in
of an active constraint ‘hands-on’ robot, such as the
Artificial Intelligence No. 1211, March 1997, pp. 3–11
Acrobot used at Imperial College for knee surgery. The
(Springer-Verlag).
benefit of this concept (in which the surgeon drives a 4 Kwoh, Y. S., Hou, J., Jonckheere, E. A. and Hayall, S. A
force-controlled lever attached to the robot) is that the robot with improved absolute positioning accuracy for CT
surgeon has the potential to feel the forces exerted by guided stereotactic brain surgery. IEEE Trans. Biomed.
the robot tool while being constrained by the robot to Engng, February 1988, 35(2), 153–161.
H01799 © IMechE 2000 Proc Instn Mech Engrs Vol 214 Part H

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016

140 B DAVIES

5 Taylor, R. H. et al. Robotic hip replacement surgery in International Symposium on Computer Assisted
dogs. In Proceedings of IEEE EMBS International Orthopaedic Surgery, Davos, Switzerland, March 1999.
Conference, 1989, pp. 887–889. 17 Marcacci, S. et al. Computer-assisted knee arthro-
6 Davies, B. L., Hibberd, R. D., Coptcoat, M. J. and plasty. In Computer-Integrated Surgery ( Eds R. H. Taylor
Wickham, J. E. A. A surgeon robot prostatectomy—a lab- et al.), 1996, pp. 417–423 (MIT Press, Cambridge,
oratory evaluation. J. Med. Engng Technol., November Massachusetts).
1989, 13(6), 273–277. 18 Brandt, G., Rademacher, K., Zimolong, A. and Rau, G.
7 Davies, B. L. Safety of medical robots. Safety of Software Development of an integrated compact robot system for
Book, Spring 1993, Ch. 3 (Software Safety Club, DTI ). orthopaedic surgery. In Proceedings of 29th ISR
8 Davies, B. L., Hibberd, R. D., Timoney, A. G. and Wickham, Conference on Advanced Robotics: Beyond 2000,
J. E. A. A surgeon robot for prostatectomies. In Birmingham, April 1998.
Proceedings of 2nd Workshop on Medical and Healthcare 19 Stoianovici, D., Whitcomb, L. L., Anderson, J. H., Taylor,
Robots, Newcastle, September 1989, pp. 91–101. R. H. and Kavoussi, L. R. A modular surgical robot system
9 Ng, W. S., Davies, B. L., Hibberd, R. D. and Timoney, A. G. for image guided percutaneous procedures. In Proc.
A firsthand experience in transurethral resection of the Medical Image Computing and Computer-Assisted
prostate. IEEE, EMBS J., March 1993, 120–125. Interventions (Eds W. T. Wells et al.), November 1998,
10 Davies, B. L., Harris, S. J., Arambula-Cosio, F., Mei, Q. pp. 404–410 (Springer, Cambridge, Massachusetts).
and Hibberd, R. D. The Probot—an active robot for pros- 20 Auer, L. M. et al. Visualization for planning and simulation
tate resection. Proc. Instn Mech. Engrs, Part H, Journal of of minimally invasive neurosurgical procedures. In
Engineering in Medicine, 1997, 211(H4), 317–326. Proceedings of Medical Image Computing Assisted
11 Lavallee, S., Brunie, L., Mazier, B. and Cinquin, P. Image Interventions ( Eds C. Taylor and A. Colchester), September
guided operating robot: a clinical application in stereotactic 1999, Vol. 1679, pp. 1199–1209 (Springer, Cambridge,
neurosurgery. In Computer Integrated Surgery (Eds R. H. UK ).
Taylor et al.), 1996, pp. 77–98 (MIT Press, Cambridge, 21 Glauser, G., Flury, P., Epitauz, M., Piquet, Y. and
Massachusetts). Burckhardt, C. Neurosurgical operation with the dedicated
12 Finlay, P. A. and Ormstein, M. H. Controlling the move- robot Minerva. IEEE, EMBS J., March 1993, 347–351.
ment of a surgical laparoscope. IEEE Engng in Med. Biol. 22 Davies, B. L., Lin, W. J., Hibberd, R. D. and Cobb, J. C.
Mag., May 1995, 14(3), 289–291. Active compliance in robotic surgery—the use of force con-
13 Sackier, J. M. and Wang, Y. Robotically assisted laparo- trol as a dynamic constraint. Proc. Instn Mech. Engrs, Part
scopic surgery: from concept to development. In H, Journal of Engineering in Medicine, 1997, 211(H4),
Computer-Integrated Surgery ( Eds R. Taylor et al.), 1996, 285–292.
pp. 577–580 (MIT Press, Cambridge, Massachusetts). 23 Troccaz, J., Peshkin, M. and Davies, B. Guiding systems
14 Mittelstadt, B. D., Kazanzides, P., Zuhers, J., Cain, P. and for computer-assisted surgery. Med. Image Analysis, 1998,
Williamson, B. Robotic surgery: achieving predictable 2(2), 101–119 (Oxford University Press).
results in an unpredictable environment. In Proceedings of 24 Carpentier, A., Loulmet, D., Aupecle, B., Berrebi, A. and
6th International Conference on Advanced Robotics, Rellard, J. Computer-assisted cardiac surgery. Lancet,
Tokyo, November 1993, pp. 367–372. 1999, 353, 379–380.
15 Wiesel, U., Lahmer, A., Borner, M. and Skibbe, H. 25 www.computermotion.com/zeus.html.
Robodoc at B.G. Frankfurt—experiences with the pinless 26 Dillman, R. and Salb, T. In Proceedings of 1st International
system. In Proceedings of 3rd Annual North American Workshop on Haptic Devices in Medical Applications, Paris,
Program on Computer Assisted Orthopaedic Surgery, France, June 1999 (IPR, University Karlsruhe, Germany).
Pittsburgh, Pennsylvania, June 1999, pp. 113–117 ( UPMC 27 Phee, S. J., Ng, W. S., Chen, I. M., Seow-Choen, F. and
Shadyside). Davies, B. Locomotion and steering aspects in automation
16 Grueneis, C. O. R., Ritcher, R. H. and Hening, F. F. Clinical of colonoscopy—a literature review. IEEE Engng Med.
introduction of the CASPAR system. In Proceedings of 4th Biol., 1997, 16(6), 85–96.

Proc Instn Mech Engrs Vol 214 Part H H01799 © IMechE 2000

Downloaded from pih.sagepub.com at PENNSYLVANIA STATE UNIV on September 15, 2016