Compact laparoscopic assistant robot using a bending...

Advanced Robotics, Vol. 21, No. 5–6, pp. 689–709 (2007) VSP and Robotics Society of Japan 2007.Also available online - www.brill.nl/ar

Full paper

Compact laparoscopic assistant robot usinga bending mechanism

SEONG-YOUNG KO 1, JONATHAN KIM 1, WOO-JUNG LEE 2

and DONG-SOO KWON 1,∗1 Department of Mechanical Engineering, Korea Advanced Institute of Science and Technology,

373-1 Guseong-dong, Yuseong-gu, Daejeon, South Korea2 College of Medicine, Yonsei University 134 Sinchon-dong, Seodaemun-gu, Seoul, South Korea

Received 27 June 2006; accepted 9 August 2006

Abstract—This paper presents the development of a compact laparoscopic assistant robot. Therobot was designed to increase convenience and reduce possible interference with surgical staff byconfining the majority of motions inside the abdomen. Its size was miniaturized as much as possiblefor convenient handling. A bending mechanism composed of several articulated joints was introducedto produce motions inside the abdomen. The proposed assistant robot can generate 3-DOF motion,including 2-DOF internal bending motion and 1-DOF external linear motion. Since the robot itselffunctions as a laparoscope, a small CCD camera module and a bundle of optical fibers were integratedas part of the system. For accurate control, mathematical modeling of the bending mechanism and amethod of hysteresis compensation were introduced and implemented. For the control of the robot,a voice interface and a visual-servoing method were implemented. The performance of the developedsystem was tested through solo-surgery of in vivo porcine cholecystectomy. It was found that theviews generated by the bending mechanism were sufficient throughout the surgery. Since the robothas functions comparable to the previously developed systems, while retaining its compactness, it isexpected to be a useful device for human cholecystectomy.

Keywords: Laparoscopic assistant robot; robot assisted surgery; bending mechanism; laparoscopy;visual-servoing.

1. INTRODUCTION

Laparoscopy became one of the most popular surgical techniques in the 1990s due toits surgical effectiveness, fast recovery and good cosmetic outcome. The most com-mon application of laparoscopy is cholecystectomy, which is the surgical removalof the inflamed or stoned gallbladder. Recently, laparoscopic cholecystectomy is

∗To whom correspondence should be addressed. E-mail: [email protected]

690 S.-Y. Ko et al.

widely performed and its rate is reported to be continuously increasing [1]. Due tominimal incision, patients can regain health without much hospitalization; however,the operating surgeons suffer from a limited range of motion, reduced flexibility,loss of tactile sensation and limited depth perception compared to open surgery.Additionally, cooperation between the operating surgeon and the assistant becomesan important issue as it can determine how well the surgeon can perform surgi-cal tasks. Manipulating vessels and organs using long tools without direct visualfeedback requires the utmost attention, and it would be desirable that the assistantmaneuvers the laparoscope to the surgical site without disrupting the operating sur-geon. Novice assistants often suffer from following the surgeon’s commands due to:(i) the difficulty in properly positioning the laparoscope in three-dimensional spacebased on the projected images on a TV monitor, (ii) the presence of the fulcrumeffect at the trocar insertion point and (iii) hand tremor caused by fatigue. In or-der to solve these problems, some surgical robotic systems [2–6] and laparoscopicassistant robot systems were developed.

Among the laparoscopic assistant robots, AESOP [7], EndoAssist [8] and thesystem developed by Russell et al. [9] utilize the conventional rigid laparoscope.The design concept is based on replacement of the human assistant’s hand motionwith rigid robotic arms of multiple links. Despite their applicability in real surgeries,these systems address some important issues that must be resolved. These systemsare known to occupy a voluminous space in the operating room, and the externalmotion of links tends to interfere or come in close contact with the surgeon andsurgical staff. In order to solve these issues, Berkelman et al. [10] and Kobayashiet al. [11] have attempted to miniaturize the system without loss of the requiredfunctions. However, their systems also change the laparoscopic view by some ofthe external motions of the robotic arm and the issues are not completely resolved.

To develop a compact robot, we adopted an internally bending mechanism. Thisinternally bending mechanism will not require a wide range of motion external to theabdomen for visualizing a wide range of surgical views. The proposed laparoscopicassistant robot system called KaLAR (KAIST Laparoscopic Assistant Robot) makesan internal bending motion for changing the viewing direction and an external linearmotion for magnifying or reducing the surgical view. The remainder of this paperwill present the implementation, the control method and the experimental results ofthe proposed KaLAR system.

2. WORKSPACE REQUIREMENTS

To determine the viewing angle range in conventional laparoscopy, we madeobservations during human cholecystectomies. In general, 4-DOF motion isavailable in conventional laparoscopic surgery [12]. There are mainly two rotations(up/down and left/right) about two axes on the incision surface, a translation (in/out)along the axis perpendicular to the incision surface and an axial rotation. Since

Compact laparoscopic assistant robot 691

Figure 1. Range of motions in cholecystectomy.

the axial rotation is not fully utilized, we have not implemented this feature in oursystem.

Our observation shows that the ranges of up/down and left/right movementsare within 30◦, while the range of in/out movement is approximately 100 mmduring normal operation, as shown in Fig. 1. The range of this in/out motion isin accordance with the result reported by Riener et al. using an electromagneticposition sensor [13]. Based on this observation of the necessary workspace, wehave developed a laparoscopic assistant robot that can cover the full range of viewrequired for human cholecystectomy.

3. COMPACT LAPAROSCOPIC ASSISTANT ROBOT

To keep the size of the robot compact and enable the change of views insidethe abdomen, several designs were considered and the bending mechanism wasdetermined to be the best candidate. So far various bending mechanisms havebeen developed to apply to laparoscopic surgery. Ikuta et al. [14] developeda 6-DOF hyper-redundant manipulator that is driven with wires to increase thedexterity, and Yamashita et al. [15] developed a multi-slider linkage mechanism thatis driven through small rigid linkage to achieve high repeatability and to improvethe power capability. The 2-DOF 4.2-mm-diameter bending mechanism consistingof some disks and four super-elastic NiTi tubes was proposed by Simaan et al. [16].Recently, Meer et al. [17] presented a disposable plastic bending mechanism usinga small partially locked ball joint. These mechanisms were mainly focusing on asurgical manipulator and improved the mechanical characteristics of the bendingmechanism. However, since it is necessary that the wires of both the small CCDmodule and the optical bundle of the light source on the tip of our system passthrough inside the bending mechanism, we adopted a simple bending mechanism

692 S.-Y. Ko et al.

Figure 2. The developed compact laparoscopic assistant robot.

consisting of hollow cylindrical short links [18]. The commercial version of themechanism was modified so that it has an appropriate workspace for our application.

The overall design of the developed robot is as shown in Fig. 2 [19]. The directionof views can be altered by changing the alignment of the articulated joints whilemagnification/reduction of the view can be altered by moving closer or further awayfrom the surgical site using a linear actuator. Unlike other previous laparoscopicassistant robots, KaLAR itself functions as a laparoscope. A CCD camera moduleand a bundle of optical fibers are installed on the tip of the bending section as shownin Fig. 2 and they are directly connected to the image-capturing unit and a xenonlight source.

3.1. Bending motion

The bending mechanism consists of hollow cylindrical short links which connect tothe next links with small joints and each link has two or four guiding holes insideas shown in Fig. 3. The axis of rotation of each joint lies at right angles. For theinternal bending motion, the most distal joint of the multiple joints is connected totwo wheels through two pairs of steel wires, which are guided by guiding holes inthe joints. There are two guiding holes inside each link except for the two links onboth ends of the bending mechanism. The two links on both ends have four guidingholes inside individually and the wires are fixed to the guiding holes inside the linknearest to the CCD camera. The two wheels are put in a line, have different size andare attached to corresponding motors directly as shown in Fig. 4. For controllingthe bending mechanism, rotation of the motor changes the tension in the wires and,thus, changes the orientation of each joint as shown in Fig. 5. For the sake of safetyand initialization, two stoppers and photo sensors are placed to fix the bending rangeas shown in Fig. 4.

We have determined the range of motions of a rigid laparoscope in Section 2.A comparable range must be possible with the bending mechanism composed ofmultiple joints. To determine how much bending is required for a comparableviewing range, we have simulated the motion using approximated parameters inlaparoscopy and design constraints. For installation of a CCD module and a bundle


Figure 3. The real image and the design image of the bending mechanism.

Figure 4. The configuration of the wheels and the motors for bending motion.

Figure 5. Wire-driven bending mechanism with motors.

694 S.-Y. Ko et al.

Figure 6. Workspace comparison between a rigid scope and the proposed system.

of optical fibers at the tip, approximately 70 mm of length is required. From theobservation, the distance between the navel and the gallbladder is approximately200 mm and in some cases the laparoscope may be placed 30–50 mm apart fromthe gallbladder during surgery. The length of the bending section is 23 mm and itis composed of seven circular links connected by six joints. In conjunction with anassumption that bending of links will form a circular shape with a constant radius,we can compute the required bending angle for a comparable viewing range. Asshown in Fig. 6, about 30◦ of bending angle will be have an equivalent viewingangle as a rigid laparoscope rotating 15◦ about the insertion point. If the robot ispositioned further than 30 mm from the region of interest, the viewable range willbe greater than that of a rigid laparoscope. To make the installation procedure moreflexible, we have configured the limit sensor so that bending can take place from−60◦ to +60◦ in each direction.

3.2. In/out motion and sterilization

For moving closer and further away from the surgical site, a linear actuatorconsisting of a linear motion guide, a ball screw and a brushed DC motor wasinstalled. In order to cover the necessary workspace, we chose a linear actuator with130-mm stroke length. This linear actuator is connected to a passive laparoscopeholder by a connector similar to the one used to join a camera and a tripod as shownin Fig. 7. The use of this passive holder allows the surgeon to readily install therobot to the bedside.

It might not be accurate to state that linearly moving a bent tip will provide anenlarged or reduced view of the site; however, considering the size of the gallbladder(approximately 10 cm long by about 5 cm wide in a conical or pear-shaped sac) issmall and initial placement of the robot does not deviate much from the site, anenlarged/reduced view can be obtained by linear motion.


Figure 7. The 2-DOF unit/linear actuator/connector/laparoscope holder.

(a) (b)

Figure 8. The control performance of the zooming motion before/after the sterilization. (a) Themotion trajectory. (b) The error of zooming motion.

With regard to the quality of the laparoscopic view, we measured the field of viewof KaLAR. The view ranges of the up/down and the left/right directions are 36.3◦and 48.4◦, respectively. The focus length was adjusted to make the view distinctwithin the range from 5 to 13 cm. The surgeon performing in vivo tests commentedthat the quality of the view was sufficient to perform the surgery.

To sterilize the KaLAR system, the moving portion of the robot, which includesthe upper part and the linear actuator, is made separable from the passive holderunit, and thus both the robot and the passive holder can be easily sterilizedwith ethylene oxide gas. The effect of sterilization with ethylene oxide gas wasinsignificant, considering the maximum errors before/after sterilization were 0.129and 0.172 mm, and both of them are sufficiently small, as shown in Fig. 8.

696 S.-Y. Ko et al.

4. MODELING AND HYSTERESIS COMPENSATION AT THE JOINTS

4.1. Modeling of the bending mechanism

The bending mechanism adopted in the KaLAR system consists of seven short linksand six paired rotary joints. The axis of rotation of each joint lies at right angles asshown in Fig. 9. This means that three joints are used for the bending tip to moveeither up or down and the others are to move either left or right. In this section weexplain the modeling of the bending mechanism with the following assumptions:(i) the movement of each joint group (up/down group or a left/right group) isindependent of each other, (ii) there are no plays in the joints and the guiding holesfor wires and (iii) the bending angle of each joint in the same group is equal toeach other. Based on these assumptions, we will derive the relationship betweenbending angle (θup/down, θleft/right) and length variation (�lwup/down, �lwleft/right) ofa corresponding wire.

Figure 9 shows the situation when only the (i+1)th joint is rotating. The positionsof the guiding holes can be expressed by (1)–(4), based on the coordinate systemlocated on the (i + 1)th joint. In these equation, l and r indicate the length ofeach link and the distance from the link center to each guiding hole, respectively.Rotation of the wheel pulls one wire on one side while the wire on the otherside is released. This variation of either wire length causes bending to occur andits mathematical relationship to the bending angle can be expressed by (5) usingeach guiding hole’s position indicated by (1)–(4). Equations (6) and (7) show theoverall length variation of the wires, and (8) and (9) show the overall bending angleproduced by the length variation of the two wires. Figure 10 shows the relationshipbetween the wire length variation by (6) and the bending angle by (8) when l and r

are 3.3 and 4 mm, respectively. The relationship is almost linear (R ≈ 1.00) withcoefficient of 14.35◦/mm and thus the bending mechanism is expected to move inproportion to the wire movement. However, since the real bending mechanism hasa play in joints and guiding holes, in addition the presence of a hysteresis, the real

Figure 9. Mathematical model of an ideal bending mechanism.


Figure 10. Ideal relationship between the wire length variation and the bending angle.

coefficient value is expected to be a bit different from the ideal one.

Hi,1 = (−l, r) (1)

Hi,2 = (−l, −r) (2)

Hi+2,1 = (l cos θ − r sin θ, l sin θ + r cos θ) (3)

Hi+2,2 = (l cos θ + r sin θ, l sin θ − r cos θ) (4)

�lwi+1 = 1

2(|Hi+2,2 − Hi,2| − |Hi+2,1 − Hi,1|) (5)

�lwup/down =∑

i=2k

�lwi+1 = 1

2

∑

i=2k

(|Hi+2,1 − Hi,1| − |Hi+2,2 − Hi,2|),

where k = 0, 1, 2 (6)

�lwleft/right =∑

i=2k+1

�lwi+1 = 1

2

∑

i=2k+1

(|Hi+2,1 − Hi,1| − |Hi+2,2 − Hi,2|),

where k = 0, 1, 2 (7)

θup/down =∑

i=2k

θi+1 (8)

θleft/right =∑

i=2k+1

θi+1. (9)

698 S.-Y. Ko et al.

4.2. Hysteresis modeling of the bending mechanism

In the previous section, we established a simplified mathematical model and it wasshown that the relationship between the wire length variation and the bending angleis highly linear. We would like to confirm that the aforementioned relationship isindeed true and assess the effect of the hysteresis through experimentations using theoptical three-dimensional position measurement system Optotrak3020 [20]. Fouroptical markers are placed on the moving tip and the rigid portion of the robot asshown in Fig. 11 and the movements of each segment are measured for a giveninput. Input consists of a set of sinusoidal inputs with increasing magnitude aftercompletion of a cycle. Figure 12 shows the measured relationship between thedesired variation of wire length and the bending angle of the tip during the up/downmotion and the left/right motion, respectively.

An interesting point to note is that unlike normal hysteresis characteristics, inwhich its effect is equally dispersed about the origin, our bending mechanism has

Figure 11. Experimental setup to measure the characteristic of bending motion.

(a) (b)

Figure 12. Measured relationships between the wire length variation and the bending angle.(a) Up/down motion. (b) Left/right motion.


Table 1.Result of hysteresis analysis

Up/down motion Left/right motion

Ratio between angle and wire(deg/mm)

11.5 10.9

Effect of hysteresis between thedesired angle and the real bend-ing angle

Difference between trendlines 1 and 2

0.22 0.23

Difference between trendlines 1 and 3

0.45 0.30

Average of offset values (hys-teresis) (mm)

0.34 0.26

its effect shifted to the right as shown in Fig. 12. The cause of this shift lies in theinitialization process, in which the motor connected to the bending tip first movesin a positive direction until a photo diode switch is activated and then comes backto the predetermined position in a negative direction. Consequently, if the bendingtip moves in a negative direction, i.e., it follows the trend line 1, the hysteresiscompensation is not required. We define ‘hysteresis’ as an average value of thedifferences between the trend lines 1 and 2 and between the trend lines 1 and 3.The cause of two different trend lines 2 and 3 during increasing the wire length isdue to imperfect alignment of circular links in the bending mechanism. The effectof the hysteresis is summarized in Table 1 and the indicated values are used ascompensating factors in controlling the bending mechanism.

4.3. Hysteresis compensation and the low-level control structure

In this section, a low-level control method for the bending motion with the hysteresisand its hysteresis compensation are discussed based on the results provided inSection 4.2. Figure 13 shows the block diagram of a low-level control structureincluding the hysteresis compensation. The compensation is made by adding theaverage offset value from Table 1 to the desired input if the input is increasing. Thiscompensation scheme can be expressed by (10). In case of the linear (zoom-in/-out)motion, no compensation is made due to no observed hysteresis effect.

�comp,i = �hys,i × sign (•θdes) + 1

2, (10)

where �comp,i indicates the value of hysteresis compensation of each motor, �hys,i

indicates the value of predefined hysteresis that is defined in Table 1 and•θdes

indicates the velocity of the desired input.This simple hysteresis compensation scheme may produce a discontinuity in the

desired value. Since the discontinuity causes an abrupt and unstable transitionbetween views, maximum deviation of the desired value is limited to a predefined

700 S.-Y. Ko et al.

Figure 13. Block diagram of low-level control system.

maximum speed (•θmax). We have determined these values to be roughly 11.2◦/s for

the bending motion and 8 mm/s for the linear motion. This method can be expressedby (11)–(13):

�d(k) = θdes(k) + �comp(k) − �beforeLPF(k − 1) (11)

if (|�d(k)| >•θmax ×�T )

�beforeLPF(k) = �beforeLPF(k − 1) + •θmax ×�T × sign (�d(k)), (12)

else

�beforeLPF(k) = θdes(k) + �comp(k), (13)

where �beforeLPF indicates the desired position value that is regulated not to exceedthe maximum velocity, �comp indicates the value calculated by (10), θdes indicatesthe desired input and �T is the sampling time.

The �beforeLPF obtained by (12) or (13) goes through a first-order low-pass filerwith τ = 0.03 s for eliminating the discontinuity of the velocity and the resultingvalue is regarded as the final desired input �des for a PD controller. As shown inFig. 13, the low-level controller is not a perfect closed loop, but it is sufficientlycontrollable under the assumptions that there is no external force acting on themoving tip and that the surgeon can see the laparoscopic view on the monitor.Figure 14 shows the tracking performance during the left/right swing motion beforeand after the hysteresis compensation.

5. HIGH-LEVEL CONTROL METHOD: A USER INTERFACE

This section explains a higher-level control method of the KaLAR system, which isrelated to the generation of the desired position (θdes) from the surgeon’s command.We adopted both a voice interface and a visual-servoing method to control thesystem. Voice recognition is implemented based on a speaker-independent softwaremodule and thus requires no training. Since it is one of the most intuitive methods,it has been successfully utilized to control many laparoscopic assistant systems


(a) (b)

Figure 14. Tracking performance with and without hysteresis compensation. (a) Without hysteresiscompensation. (b) With hysteresis compensation.

[21, 22]. However, the voice interface has a limitation of requiring a lot ofvoice input when there is a need for continuous view change. To overcome theshortcoming of the manual control, a visual-servoing method [7, 23–25] had beenimplemented. However, it also has some difficulty of adding some additionalcommands and changing the magnifying ratio of the surgical view. Thereupon,we have combined the voice interface and the visual-servoing. These two controlmethods are selected by the operating surgeon by voice commands during theoperation.

5.1. Voice interface

As shown in Fig. 15, voice commands are utilized to determine the robot’s stateand the control mode. After an initialization process, the system is in the pausingstate and waits for the surgeon’s command. Upon the ‘start’ command by thesurgeon, the robot system is placed in the controlling state where the robot can bephysically activated for specific movement and image processing. To pull the robotout of the controlling state, the ‘pause’ command is required. In the controllingstate, the surgeon can choose the control mode using commands ‘tracking mode’or ‘voice mode’. In voice command mode, the surgeon manipulates the surgicalview using the commands ‘go up’, ‘go down’, ‘go left’, ‘go right’, ‘zoom in’, and‘zoom out’. These commands move the robot toward the corresponding directionby predetermined amounts, about 4◦ per command for bending and 20 mm percommand for a linear motion. Auto-tracking mode is for tracking the primarysurgical instrument marked with color markers. In this mode, 2-DOF bendingmotion is controlled by visual-servoing while the in and out motion with respectto the abdomen is still controlled by the voice commands ‘zoom in’ and ‘zoom out’.For additional convenience, a two-position memory function is also implemented

702 S.-Y. Ko et al.

Figure 15. Robot states and available voice commands.

using commands ‘remember position 1’, ‘remember position 2’, ‘go to position 1’and ‘go to position 2’.

With regard to voice recognition, the experiments with the KaLAR system in thelab environment show that the recognition rates with and without the prerecordednoise of the surgical environment were 87.4% (σ = 4.5%) and 98.1% (σ = 1.5%),respectively. The average time delay from voice recognition to the bending/zoomingmotion was 0.74 s (σ = 0.12 s). The magnitude of this time delay turned out tobe tolerable by the surgeon. Five subjects participated in these experiments and allcommands were issued in Korean, while both Korean and English commands canbe recognized.

5.2. Visual-servoing

Visual-servoing is expected to alleviate the surgeon from issuing a great number ofvoice commands in times of frequent change of camera views. The visual-servoingalgorithm is based on the result of other researchers [7, 23–25]. Unlike the previousworks, to locate the tip of the instrument in the captured image and to identify thetool’s type at once, a color marker composed of a two-color band is placed at the tipas shown in Fig. 16 [26]. The two-color band is composed of three parts: near(P1),middle(P2) and far(P3) part, named by the distance from the tip. The near andthe far parts from the tip are marked with bright cyan as it is rarely found in theinternal organs [23]. These parts have different thicknesses, and are used to locatethe direction and position of the tool tip. Since the real distances (D1t and D12)between markers and a tip in Fig. 16b are known, we can obtain the tip positionwith a simple equation (14), in which the effect of a perspective view is neglectedfor the sake of the simplification. The color of the middle part is used for identifyingthe type of tool that is inserted and, thus, is utilized to verify the marker detection


Figure 16. Color markers on surgical instruments. (a) Real marker image. (b) Schematic diagram.

and to upload the geometric information of the tool.

Pt = P1 + |P2 − P1|D1t

D12

P1 − P3

|P1 − P3| . (14)

To avoid the surgeon’s motion sickness, the visual-servoing is activated only whenthe tip is moved out of the small portion at the center of a TV screen. The size of theportion and the maximum bending speed during in vivo porcine cholecystectomieswere determined by the operating surgeons’ preference before the surgery began,and thus were approximately 11.2◦/s and 30% of the TV monitor, respectively.

6. IMPLEMENTATION AND EVALUATION

6.1. Overall system configuration

The main controller is based on a Pentium 4 2.8-GHz PC running under Windows2000. A Model 626 board from Sensoray is utilized for performing low-levelposition control and for generating hardware interrupts. VoiceEZ software fromVoiceware is utilized for recognizing the surgeon’s voice commands and forsynthesizing voice instructions. For convenience, a wireless headset from Inter-Mis used. A Matrox Meteor-II frame grabber board from Matrox and a small CCDcamera (IK-M43S) from Toshiba are used for image processing. In the developedsoftware module, three threads were implemented, each accounting for positioncontrol, voice recognition and image process. Sampling in the PD controller isconducted at 1000 Hz and image processing is done at a minimum rate of 25 Hz.Since the CCD camera module can support multiple outputs, the laparoscopic viewis delivered simultaneously to the image grabber and a super VHS recorder, whichis connected to a high-definition monitor, as shown in Fig. 17.

6.2. In vivo porcine cholecystectomy

To evaluate the performance of KaLAR, three cases of porcine cholecystectomywere performed. The objective of these trials were (i) to determine if the workspace

704 S.-Y. Ko et al.

Figure 17. Overall system configuration.

(a) (b)

Figure 18. Porcine cholecystectomy with KaLAR. (a) Solo surgery. (b) Internal view.

covered by the robot is sufficient for cholecystectomy, (ii) to see if solo-surgeryis possible with the proposed control scheme and (iii) to determine if the timerequired to complete the surgery is comparable to robot-assisted and human-assistedcholecystectomy. The materials used were three female pigs, 3–4 months old andweighing approximately 30 kg. The size of their abdominal cavity was smallerthan that of an adult person and, thus, the trocar had to be placed below the navel.Since KaLAR’s initial position influences the motion range, it is necessary to placeit carefully during the initialization procedure. The cholecystectomy mainly dealswith the gallbladder that is under the liver, so we fixed KaLAR to head for thelower part of the liver by adjusting the passive laparoscopic holder. With respectto the in/out position, we marked the long hollow tube of KaLAR with a colortape to indicate that the bending mechanism is fully inserted into the abdomen. Allthree surgeries were performed by one surgeon as shown in Fig. 18a and all thesurgical procedures were in accordance with the guidelines enforced by the localethics committees.


Table 2.Surgical time for porcine cholecystectomy

Target and material Time (min)

On a pig with Kobayashi’s system [11] 38 (one case)On a pig with a human assistant [11] 41.6 (10 cases)On a pig with the KaLAR system 26.7 ± 8.3 (three cases)

Through animal tests, we were able to confirm that the workspace covered byKaLAR is sufficient for porcine cholecystectomy, and the control of the robotusing voice commands and visual-servoing is effective enough for solo-surgery.The surgical time comparison with other robot-assisted and human-assisted porcinecholecystectomies is summarized in Table 2, where the surgical time was definedas the time from inserting the laparoscope or KaLAR to extracting it. The surgicaltimes described in Ref. [11] were recalculated in terms of our definition, i.e., wehave subtracted the trocar insertion time from the total operating time. Althoughwe only have a limited number of surgeries and the time measurement can onlybe used for a rough estimate of the robot’s performance, the time spent forporcine cholecystectomy can be said to be comparable to other robot-assisted andhuman-assisted porcine cholecystectomy. Note that the time difference betweenexperiments with KaLAR and the other system seems to be mainly caused by thesurgeon’s expertise level, considering our surgeon’s operating time (23.3 ± 9.9 minfor four cases) for the human cholecystectomy with a conventional rigid scope isslightly shorter than for these experiments using KaLAR.

Next, the other objective of in vivo tests was verifying whether or not KaLARcould cover all the views that a surgeon wants to see during operations. Figure 19and Table 3 show the trace of KaLAR’s motions during in vivo tests. The maximumone-directional range of bending motion and maximum range of linear motion were34.50◦ and 120 mm, respectively, and they were within the workspace of KaLAR.We can conclude that the workspace of KaLAR is sufficient to provide necessaryviews during porcine cholecystectomy. Table 3 includes the motion range duringthe second and third experiments as no position data were measured during the firstexperiment.

In the case that the tip was bending, the zooming motion may not be able to keepthe center position of the view. Since the tip was adjusting automatically duringthe auto-tracking mode, the aforementioned effect was not considerable. However,during the voice-command mode, it may make the surgeon uncomfortable. How-ever, considering most of the bending motion is not large, as shown in Fig. 19, thedeviation by the zooming motion was acceptable without any compensation eventhough the tip is bending a little. In only a few cases did the surgeon issue addi-tional commands to adjust the center position.

706 S.-Y. Ko et al.

Figure 19. The desired positions during the third in vivo test.

Table 3.The motion of range of KaLAR during real operations

Up Down Left Right Zoom in Zoom out(deg) (deg) (deg) (deg) (mm) (mm)

Second experiment 5.08 −14.49 19.19 −3.27 −80 0Third experiment 8.28 −34.50 16.89 −19.62 −100 20Maximum motion range 42.78 36.51 120Maximum one-directional range 34.50 100

7. CONCLUSIONS AND FUTURE WORKS

This paper proposes a new compact laparoscopic assistant robot, KaLAR. TheKaLAR system can generate 3-DOF motion, including 2-DOF internal bendingmotion and 1-DOF external linear motion. Unlike previous robotic systems, therobot makes use of an internally bending mechanism and the constraining ofmotions within the abdomen is expected to reduce the potential risk of interferingwith surgical staff. A mathematical modeling of the bending mechanism wasperformed and the inherent hysteresis characteristics of the mechanism werecompensated for more accurate control. In order to facilitate easy operation, a voiceinterface and a visual-servoing method were introduced and implemented. To verifythe applicability of KaLAR, three solo-surgeries on porcine cholecystectomy wereperformed. These in vivo animal tests showed that the mechanical structure of


the KaLAR system is acceptable in the surgical environment and has sufficientworkspace to provide necessary views during cholecystectomy.

It was noticed during our preliminary animal study that the KaLAR system canprovide the laparoscopic view as long as the operating surgeon places KaLAR in thevicinity of the gallbladder. This placement procedure is not a difficult task becausethe surgeon always checks the entire abdominal cavity before the surgery begins andcan roughly estimate where the robot should be placed. There was no big differencein operating time in comparison to other robot-assisted and human-assisted porcinecholecystectomy.

Despite KaLAR’s applicability, there are several shortcomings that must beimproved. The length of the bending tip should be shortened because its lengthinesscan restrict the reachable area. With regard to the bending mechanism, the multiplelinks produced a considerable amount of hysteresis and its effect could not becompletely eliminated by the compensation. Therefore, it is necessary to reducethe play in the joints and improve the alignment of links by modifying the bendingmechanism. There is the potential of harm caused by the robot’s malfunction,especially during the zooming motion. To reduce the possibility of malfunction,we will introduce an additional sensor to confirm the actuators’ position measuredin the zooming motion. In addition to these mechanical improvements, more in vivotests are required for more quantitative evaluation and for its extensibility to otherminimally invasive surgeries.

Acknowledgments

This work was supported by the SRC/ERC program of MOST/KOSEF (grant #R11-1999-008). The authors gratefully acknowledge the assistance of Yun-Ju Lee andWon-Ho Shin.

REFERENCES

1. M. R. Ferreira, R. L. Bennett, S. C. Gilman, S. Mathewson and C. L. Bennett, Diffusion oflaparoscopic cholecystectomy in the veterans affair health care system, 1991–1995, Effect. Clin.Pract. 2, 49–55 (1999).

2. G. S. Guthart and J. K. Salisbury Jr., The IntuitiveTM Telesurgery system: overview andapplication, in: Proc. IEEE Int. Conf. on Robotics and Automation, San Francisco, CA, pp.618–621 (2000).

3. G. H. Ballantyne and F. Moll, The da Vinci telerobotic surgical system: the virtual operativefield and telepresence surgery, Surg. Clin. North Am. 83, 1293–1304 (2003).

4. M. Ghodoussi, S. E. Butner and Y. Wang, Robotic surgery—the transatlantic case, in: Proc.IEEE Int. Conf. on Robotics and Automation, Washington, DC, pp. 1882–1888 (2002).

5. R. J. Franzino, The Laprotek surgical system and the next generation of robotics, Surg. Clin.North Am. 83, 1317–1320 (2003).

6. N. Koizumi, S. Warisawa, H. Hashizume and M. Mitsuishi, Development of a remote minimally-invasive surgical system with operational environment transmission capability, in: Proc. IEEEInt. Conf. on Robotics and Automation, Taipei, pp. 2663–2670 (2003).

708 S.-Y. Ko et al.

7. Y.-F. Wang, D. R. Uecker and Y. Wang, Choreographed scope maneuvering in robotically-assisted laparoscopy with active vision guidance, in: Proc. 3rd IEEE Workshop on Applicationsof Computer Vision, Sarasota, FL, pp. 187–192 (1996).

8. P. A. Finlay, Clinical experience with a goniometric head-controlled laparoscope manipulator,in: Proc. IARP Workshop on Medical Robotics, Vienna (1996).

9. R. H. Taylor et al., A telerobotic assistant for laparoscopic surgery, in: IEEE Eng. Med. Biol. 14,279–288 (1995).

10. P. Berkelman, P. Cinquin, J. Troccaz, J. Ayoubi, C. Letoublon and F. Bouchard, A compact,compliant laparoscopic endoscope manipulator, in: Proc. IEEE Int. Conf. on Robotics andAutomation, Washington, DC, pp. 1870–1875 (2002).

11. E. Kobayashi, K. Masamune, I. Sakuma, T. Dohi and D. Hashimoto, A new safe laparoscopicmanipulator system with a five-bar linkage mechanism and an optical zoom, Comput. AidedSurg. 4, 182–192 (1999).

12. M. C. Çavusoglu, I. Villanueva and F. Tendick, Workspace analysis of robotic manipulatorsfor a teleoperated suturing task, in: Proc. Conf. on Intelligent Robots and Systems, Maui, HI,pp. 2234–2239 (2001).

13. R. Riener, S. Reiter, M. Rasmus, D. Wetzel and H. Feussner, Acquisition of arm and instrumentmovements during laparoscopic interventions, Minimally Invasive Ther. Allied Technol. 12, 235–240 (2003).

14. K. Ikuta, T. Hasegawa and S. Daifu, Hyper redundant miniature manipulator ‘hyper finger’ forremote minimally invasive surgery in deep area, in: Proc. IEEE Int. Conference on Robotics andAutomation, Taipei, pp. 1098–1102 (2003).

15. H. Yamashita, D. Kim, N. Hata and T. Dohi, Multi-slider linkage mechanism for endoscopicforceps manipulator, in: Proc. IEEE Int. Conf. on Intelligent Robots and Systems, Las Vegas,NV, pp. 2577–2582 (2003).

16. N. Simaan, R. Taylor and P. Flint, High dexterity snake-like robotic slaves for minimally invasivetelesurgery of the upper airway, Lecture Notes Comput. Sci. 3217, 17–24 (2004).

17. F. V. Meer, A. Giraud, D. Esteve and X. Dollat, A disposable plastic compact wrist for smartminimally invasive surgical tools, in: Proc. IEEE Int. Conf. on Intelligent Robots and Systems,Edmonton, pp. 919–924 (2005).

18. H. Tanaka, Articulated, four-way bendable tube structure, US Patent 4,108,211 (1978).19. J. Kim, Y.-J. Lee, S.-Y. Ko, D.-S. Kwon and W.-J. Lee, Compact camera assistant robot for

minimally invasive surgery: KaLAR, in: Proc. IEEE/RSJ Int. Conf. on Intelligent Robots andSystems, Sendai, pp. 2587–2592 (2004).

20. http://www.ndigital.com/.21. M. E. Allaf, S. V. Jackman, P. G. Schulam, J. A. Cadeddu, B. R. Lee, R. G. Moore and L. R.

Kavoussi, Laparoscopic visual field: voice vs. foot pedal interfaces for control of the AESOProbot, Surg. Endosc. 12, 1415–1418 (1998).

22. A. Nishikawa, D. Negoro, H. Kakutani, F. Miyazaki, M. Sekimoto, M. Yasui, S. Takiguchi andM. Monden, Design of an endoscopic solo surgery simulator for quantitative evaluation of thehuman–machine interface in robotic camera positioning systems, Adv. Robotics 17, 503–521(2003).

23. G.-Q. Wei, K. Arbter and G. Hirzinger, Real-time visual servoing for laparoscopic surgery:controlling robot motion with color image segmentation, IEEE Eng. Med. Biol. 16, 40–45 (1997).

24. A. Nishikawa, S. Asano, R. Fujita, T. Yohda, F. Miyazaki, M. Sekimoto, M. Yasui, S. Takiguchiand M. Monden, Robust visual tracking of multiple surgical instruments for laparoscopicsurgery, in: Proc. Computer Assisted Radiology and Surgery, London, Poster 092 (2003).

25. A. Casals, J. Amat and E. Laporte, Automatic guidance of an assistant robot in laparoscopicsurgery, in: Proc. IEEE Int. Conf. on Robotics and Automation, Minneapolis, MN, pp. 895–900(1996).


26. S.-Y. Ko, J. Kim, D.-S. Kwon and W.-J. Lee, Intelligent Interaction between Surgeon andLaparoscopic Assistant Robot System, in: Proc. IEEE Int. Workshop on Robots and HumanInteractive Communication, Nashville, TN, pp. 60–65 (2005).

ABOUT THE AUTHORS

Seong-Young Ko received the BS degree in Mechanical Engineering from KoreaAdvanced Institute of Science and Technology (KAIST), South Korea, in 2000,and the MS degree in Mechanical Engineering from KAIST, in 2002. Heis currently working towards the PhD degree in the Human–Robot InteractionResearch Center at KAIST. His research interests include medical robotics,human–robot interaction and telerobotics.

Jonathan Kim received the BS degree in Mechanical Engineering from TheCooper Union for the Advancement of Science and Art, New York, USA, in 1993,and the MS degree in Mechanical Engineering from the State University of NewYork at Buffalo, New York, USA, in 1994. He has worked as a ManufacturingEngineer and IT consultant. He is currently working towards the PhD degree inthe Human–Robot Interaction Research Center at Korea Advanced Institute ofScience and Technology (KAIST), South Korea. His research interests includemedical robotics and human-robot interaction.

Woo-Jung Lee received the Medical degree from the College of Medicine atYonsei University, Seoul, South Korea in 1982, the Diploma from the KoreanBoard of General Surgery in 1987, the Master of medicine from Yonsei University,Seoul, in 1992 and the PhD degree in the Department of Surgery from KoreaUniversity, Seoul, in 1996. He served the Korean Army as a captain from 1987to 1990. From 1997 to 1999, he was a Research Fellow at the Vanderbilt MedicalSchool, Tennessee, USA. He is currently a Professor of the Department of Surgeryand the Director of Yonsei Medical Minimally Invasive/Robotic Surgery Center at

Yonsei University, Seoul. His current research interests include minimally invasive surgery and roboticsurgery. He is a member of the KMA, KAGS, HBPS and SLS.

Dong-Soo Kwon received the BS degree in Mechanical Engineering from theSeoul National University, South Korea, in 1980, the MS degree in MechanicalEngineering from Korea Advanced Institute of Science and Technology (KAIST),Seoul, in 1982 and the PhD degree in mechanical engineering from GeorgiaInstitute of Technology, Atlanta, Georgia, USA, in 1991. From 1991 to 1995,he was a Research Staff at Oak Ridge National Laboratory. He is currentlya Professor of Mechanical Engineering and the Director of the Human–RobotInteraction Research Center at KAIST. His current research interests include

medical robots, human–robot/computer interaction, telerobotics and haptics. He is a member of theIEEE, KSME and ICASE.

Date post:	25-Mar-2018
Category:	Documents
Upload:	phungtram
View:	216 times
Download:	2 times

Compact laparoscopic assistant robot using a bending...

Documents