Abstract—In recent years, due to the increasing rate of elderly people in Japan, the needs to detect adults’ diseases at the early stage becomes a high priority. In particular, an increased interest in detecting heart and cerebrovascular diseases at an early stage may allow clinicians to begin treatment sooner (when more treatment options are available), when interventions are generally more effective and less expensive. For this purpose, the introduction of robotic-assisted technology has advantages in terms of repetitiveness and accuracy of the measurement procedure. Therefore, at Waseda University, we have proposed the development of a carotid blood flow measurement system to support doctors while using ultrasound diagnostic equipments to detect arteriosclerosis and myocardial ischemia by measuring wave intensity. In this paper, the Waseda-Tokyo Women’s Medical-Aloka Blood Flow Measurement System No. 1 Refined II (WTA-1RII) is detailed. The WTA-1RII is composed by an ultrasound diagnostic system and a probe-supporting robotic device. The robotic device is composed by a parallel link manipulator (for fine adjustment) and a passive arm (for rough positioning). The WTA-1RII has improved the design of the gravity compensation mechanism (composed by a constant force spring attached to a slide guide). In addition, a genetic algorithm has been implemented to determine the optimal link’s position of the 6-DOF parallel manipulator to increase the workspace. Finally, a set of experiments were carried out to determine the usability of the proposed system.

I. INTRODUCTION HE heart and cerebrovascular diseases are considered among the main causes of death in developed countries. In particular, it is estimated that about 30% of the

Japanese die due to heart and cerebrovascular diseases. Henceforth, due to the effects of the elderly society, it is

Manuscript received January 23, 2009. A part of this research was done at the Humanoid Robotics Institute (HRI), Waseda University.

Ryu Nakadate, Hisato Uda, Hiroaki Hirano is with the Graduate School

of Advanced Science and Engineering, Waseda University (r-nakadate@takanishi.mech.waseda.ac.jp)

Jorge Solis is with the Department of Modern Mechanical Engineering, Waseda University; and a researcher at the Humanoid Robotics Institute (HRI), Waseda University (solis@kurenai.waseda.jp)

Atsuo Takanishi is with the Department of Modern Mechanical Engineering, Waseda University; and one of the core members of the Humanoid Robotics Institute (HRI), Waseda University (contact@takanishi.mech.waseda.ac.jp)

Eiichi Minagawa is with ALOKA Co. Ltd., Japan (http://www2.aloka.co.jp/)

Motoaki Sugawara is with Department of Medical Engineering, Himeji Dokkyo University, Japan

Kiyomi Niki is with the Department of Biomedical Engineering, Tokyo City University, Japan

important to detect and treat arteriosclerosis and myocardial ischemia, which cause the abovementioned adult diseases, at an early stage. For this purpose, recently a methodology to measure the force of cardiac contraction has been proposed. Such methodology is based on the measurement of the index Wave Intensity (WI). Basically, the WI is a hemodynamic index that can be defined any location in the circulatory system proposed by Parker et al. [1]. It can be defined at any site in the circulatory system, and provides information about the interaction of forward and backward traveling waves, and hence, information about the dynamic behavior of the heart and vascular system and their interaction. Wave intensity was originally defined as the product of △P and △U, where △P and △U are the changes in blood pressure P and velocity U, respectively, during constant short time intervals. Wave intensity has the dimension of rate of energy flux per unit area and corresponds to acoustic intensity. This original wave intensity depends on the sampling interval, △t, which makes it difficult to compare data measured at different sampling rates. Therefore, Sugawara et al. normalized wave intensity by dividing △P and △U by △t [2]. Thus, the time-normalized wave intensity (WI) is given by as Eq. 1.

=

dtdU

dtdPWI (1)

In particular, the conventional WI measurement is done

manually by a doctor or technician. Basically, the operator manipulates the ultrasound diagnostic system while fixing a probe to the measurement point (in which the patient must adopt a supine posture). However, the measurement time takes long time (around 18 minutes) and the overall measurement accuracy may be low. More recently, a conventional solution has been proposed based on a probe-holding device consisting of a passive arm was used. However, the measurement time and accuracy is affected due to the lack of user-friendliness [3-5].

Therefore, at Waseda University, we have proposed the development of a robotic system which aims in reducing the required time to measure the WI, increase the accuracy of the measurement and enhance the user-friendliness of the assisted-robotics system.

As a result of our research, the WTA-1 has been developed in 2006 [6]. The WTA-1 is composed by an ultrasound diagnostic system (from ALOKA, Ltd.), a probe-supporting device and a controlling device. The probe supporting devices consisted of passive arms for the approximate positioning of

Development of a Robotic Carotid Blood Measurement WTA-1RII: Mechanical Improvement of Gravity Compensation Mechanism and

Optimal Link Position of the Parallel Manipulator based on GA Ryu Nakadate, Hisato Uda, Hiroaki Hirano, Jorge Solis, IEEE Member, Atsuo Takanishi, IEEE Member, Eiichi

Minagawa, Motoaki Sugawara, Kiyomi Niki

T

2009 IEEE/ASME International Conference on Advanced Intelligent MechatronicsSuntec Convention and Exhibition CenterSingapore, July 14-17, 2009

978-1-4244-2853-3/09/$25.00 ©2009 IEEE 717

the probe and a compact actuated parallel manipulator for fine adjustment.

Although the configuration of the link part resembles the Stewart platform, a different part is used for the division of the link and actuator parts. Consequently, it becomes possible to reduce the thickness of the link. Since this configuration allows the reduction of the thickness of the link, the probability of interference with each link is low; hence, a wide movable range is possible.

Even that the WTA-1 has demonstrated to reduce the required time for doing the measurement of the WI and enhance the user-friendliness, further mechanical improvements are required to optimize the workspace of the parallel manipulator which holds the probe [6]. For this purpose, in this paper, we present the Waseda-Tokyo Women’s Medical-Aloka Blood Flow Measurement Robot System No. 1 Refined II (WTA-1RII) which has been developed at Waseda University. In particular, the parallel manipulator of the WTA-1RII has been designed to optimize the workspace by means the Genetic Algorithm. Furthermore, a gravity compensation mechanism has been proposed to eliminate any force while the operator is grasping the manipulator. A set of experiments are proposed to verify the effectiveness of the redesigned manipulator of the WTA-1RII.

II. WTA-1RII SYSTEM

A. System Overview The overall system is composed of the ultrasound

diagnostic system (Pro Sound II SSD-6500SV that is commercialized by ALOKA Ltd.), a probe supporting robot and its controller as shown in Fig.1. The probe supporting

robot is composed by 6-DOFs passive serial link arm, 3-DOFs passive ball joint and 6-DOFs active parallel link manipulator (Figure 2). The detailed configuration of the mechanism is shown in Fig 3.

All passive joints of the serial link arm include magnetic brakes which are used to lock/unlock the joints controlled by foot switch controller. On the other hand, a push switch controller is attached to the probe holder. The magnetic brakes unlock the joints when power supply is on, so that risk of injury caused by power shut down is minimized.

The active manipulator employs the linear parallel link mechanism, which each contains an actuation system composed by a DC servo motors, a ball screw and universal joint (Figure 4). The actuation mechanism is used accurately control the positioning and posture of the end-effector (where the ultrasound probe is attached). The control system calculates the inverse kinematics of linear parallel link mechanism according to the target position commanded a joystick-styled controller designed at our laboratory.

In particular, the joystick-styled controller is attached with a set of sensors to measure the position and posture of the tip of the device. In particular, the following sensors were embedded: accelerometers, gyro, photo reflectors and laser mice. Thanks to the design of the controller device, the adjustment of the probe position can be easily done with high level of precision. By using the proposed WTA-1RII, the operator can unlock the passive joints and accurately control the positioning the manipulator close to measurement point by hand (rough positioning), then lock the joints, and finally, adjust the positing with precision by means of the joystick-styled controller.

Fig. 1. Waseda-Tokyo Women’s Medical-Aloka Blood Flow Measurement Robot System No. 1 Refined II (WTA-1RII)

718

B. Gravity Compensation Mechanism The position control of passive arm is done by operator’s

hand. It is important for better controllability that operators do not feel the gravity force by weight of the arm, so we have installed the mechanical gravity canceller to the passive arm. The gravity canceller should be passive, linear with any angle of the joint, and have small friction so that the operator can move end-effector freely. We have employed constant force spring.

The model of our first prototype is shown as Fig.5. The constant force spring is attached to the base and adds force to the horizontal link of the arm. This model does not compensate the gravity force by the end-effector completely when the angle of the joint (θ) is not zero. When θ equals to zero, the moment force added by constant force spring and the one added by gravity are the same so that Eq. (2) is obtained, where F is force of constant force spring, M is the weight of the end-effector.

MgLFl = (2)

In the case θ is different to zero; the moments are calculated

as follows. The force F directed from point C to A in the Fig.5 can be divided into horizontal and vertical force. The vertical

force f is calculated as Eq. (3). The remaining force after cancellation at the point of end-effector is then obtained by Eq. (4).

22 )sin()cos1(

sin

θθ

θ

Lhl

lhFf−+−

−= (3)

LflMgLError θθ coscos −

= (4)

By using Eq. (4), the remaining force is computed and

plotted in Fig.6. The parameters are, l=135[mm], L=300[mm], h=100[mm], M=3.6[kg]. By this model remaining force after cancellation is maximum 1700[gf].

100

340～400

Fig.4. Detail of the 6-DOFs Parallel Link Manipulator of the WTA-1RII

Constant Force Spring

x

z

x

z

Ll θ

Mg

Fh

fA B

C

Fig.5. Principle of the first prototype of the gravity compensation mechanism composed by a constant force spring

-0.2

0.2

0.6

1

1.4

1.8

-60 -40 -20 0 20 40 60

Link's Angle θ(deg)

Err

or

Forc

e (

gf)

Fig. 6. Computed result of the compensation error by link’s angle of the first prototype.

Fig. 2. Details of the probe supporting robot; where the passive arm is used to roughly positioned the probe on the patient body and the parallel manipulator is used to refine the location of the probe.

Fig. 3. Details of the mechanism configuration of the probe supporting robot which is composed by a passive arm mechanism and a 6-DOFs parallel mechanism manipulator

719

x

z

x

z

Ll θ

Mg

FC

Slide Guide

Constant Force Spring

Fig.7. Detail of the improved gravity compensation mechanism proposed to be implemented on the WTA-1RII. In this model, a slide guide is used so that the moment arm length of the right and left side of the link is not affected by the change of the angle of the joint.

The improved model is shown in Fig. 7. The constant force

spring is attached on a slide guide located on the base of the WTA-1RII. In this model, force of constant force spring F is always vertical, and not affected by change of the angle of joint. By this model, the remaining force after cancellation is reduced to 100[gf]. However, the constant force spring adds non-linearity components. In order to minimize such effect, the angle of the slide guide respect to the base of the arm can be adjusted, so that the effect of the non-linearity of the output of constant force spring is effectively cancelled. As a result, not only the gravity force was effectively compensated but also a linear response was achieved thanks to the proposed mechanism (Figure 8). From these improvements stated above, a high level of controllability of the passive arm was assured.

C. Link’s Position Optimization based on GA The maneuverability of the probe supporting robot is highly

related to the designed workspace of the active manipulator. If the working space of the 6-DOFs parallel manipulator is small, the operator is required to accurately position the passive arm to assure the effectiveness of the measurement. In contrast, if the active manipulator is designed to have a wider working space, the time for adjustment of the positioning of the passive arm can be considerably reduced. In order to increase the working space of the active manipulator, there are two possible strategies: to increase the dimensions of the manipulator or to optimize the link position of the parallel links. The first strategy may cause a

considerable increase of the weight of the active manipulator. Therefore, in this paper, we have proposed to apply the Genetic Algorithm in order to optimize the workspace of the active manipulator by setting the position of the links. In particular, as it is shown in Table I, we have defined the target workspace of the active manipulator based on our discussion with doctors.

Table I. Required specifications of the working space of the active

manipulator designed for the WTA-1RII

Parameter Axis Specifications(mm)

x ±10 y ±10 Position z 0 ~ -10

roll ±7.5 pitch ±5.0 Orientation yaw ±10.0

The chromosome (C) is defined as in Eq. (5) where i

represent the link number, b denotes the base of the link and e denotes the end-side of the link [7]. The positions of three of the links of the manipulator are defined on the zx plane as it is shown in Fig. 9 (considering the symmetry properties regarding the other three links).

{ } 3,2,1 ;,,,,, == ixzxzC eieibibi LL (5) In particular, in this paper, we focused on finding the

optimal position of each of the links so that the resultant workspace of the parallel manipulator satisfies the specifications shown in Table I. For the present model, GA has been developed in real code representation using Roulette wheel selection, arithmetic crossover and uniform mutation [7]. In particular, the mutation rate has been experimentally determined (0.05). The cost function is the volume of the working space in which the target posture (Table I) is achieved. Moreover, in order to avoid the singularity problem of the proposed parallel mechanism, the following considerations were taken into account.

),( 11 ee zx

),( 11 bb zx

),( 22 bb zx

),( 22 ee zx

),( 33 ee zx−

),( 33 bb zx

),( 11 ee zx−

),( 11 bb zx−

),( 22 bb zx−

),( 22 ee zx−

),( 33 ee zx

),( 33 bb zx−

Fig. 9. Definition of the chromosomes to apply the Genetic Algorithm to optimize the workspace; where each of the chromosomes represents the position of each of the links of the 6-DOFs parallel manipulator.

Fig. 8. Screenshot of the passive arm designed for the WTA-1RII

720

The computed length of each link (hi) corresponding to the movement of the end-effector along the x-axis can be expressed as hx= (h1 h2 h3 h4 h5 h6). Based on the same principle, the other five vectors hy, hz, hφx, hφy and hφz can be defined. As a result, the motion of the end-effector can be easily expressed as a linear combination from those vectors, as is shown in Eq. (6).

zzyyxxzzyyxx ΦΦΦΦΦΦ +++++= hhhhhhh λλλλλλ (6) If we consider that the six vectors are linear independent, the

link position will not have singularity. For this purpose, we have rejected those solutions obtained by the GA algorithm when the following condition is fulfilled, as is shown in Eq. (7). The parameter k is a threshold that is determined as twice the minimum value of the determinant measured on our first prototype which had no singularity.

( ) kzyxzyx <ΦΦΦ hhhhhhdet (7)

However, in some cases the above determinant may not be

satisfied; particularly if any two vectors are nearly parallel depending on the other vectors. As a result, we have implemented an additional condition as is shown in Eq. (8).

　θ1cos1),(

1 −−=⋅

−=yx

yxdhh

hh (8)

By using Eq. (8), those solutions which fulfills the condition

d < do are rejected. In this case, do is a threshold which is determined as twice of the measured d of our first prototype. Therefore, for each computation step of the GA algorithm, the resultant working space is calculated. If the resultant working space does not satisfy the target requirements, a further computation step is performed by slightly increasing the link length until the target working space is satisfied.

III. EXPERIMENTS & RESULTS

A. Gravity Compensation Mechanism Before carrying out tests with technicians with the

WTA-1RII, we focused on verifying if the proposed design effectively cancels the gravity force as well as the non-linear components. This is an important issue as technicians should not feel any additional force while manipulating the parallel manipulator of the WTA-1RII. For this purpose, we have plotted the response of the gravity compensation mechanism to verify if the proposed design effectively cancels the gravity force as well as the non-linear components. The experimental results are shown in Fig. 10. The force at the end-effector was measured by spring weight meter manually. During the gravity compensation experiment, other joints were fixed in order to make moment length constant. As we may observe, thanks to addition of the constant spring the gravity force is effectively cancelled. Moreover, no-linear components were found due to the linear guide.

B. GA’s Optimization Results In this experiment, we have verified if the proposed Genetic

Algorithm could effectively determine the link length while optimizing the workspace of the parallel manipulator. Basically, after each iteration step of the GA, the workspace was calculated. If the resultant workspace do not satisfies the target specifications, the link length is re-calculated and a iteration step of the GA is repeated. This process was done until the target specifications are satisfied. The experimental results are shown in Fig. 11. As we may observe, the target specifications were satisfied when the link length was set to 70mm.

C. Preliminay Clinical Tests Finally, we have carried a clinical experiment to verify the

usability of the WTA-1RII. In particular two subjects (one doctor and one sonographer) were asked to measure the carotid blood flow by using the WTA-1RII for fifteen times. In order to measure the usability of the designed workspace of parallel manipulator, we have registered the location of the end-effector while the subjects were performing the task. The results are shown in Fig. 12; where we have compared the

Fig. 10. Experimental results while measuring the gravity force on the designed parallel manipulator. As we may observe, the proposed compensation mechanism effectively reduces the gravity force.

-90

-80

-70

-60

-50

-40

-30

-20

-10

0

-40 -20 0 20 40

x mm

z m

m

Base

End

2345

12

3 4

5

616

Fig. 11. Experimental results while applying the GA to optimize the workspace of the parallel manipulator designed for the WTA-1RII. The length of the links was designed to 70mm and the position of each link is shown in the graph.

721

percentage of use of the workspace of the previous mechanism and the one designed based on the GA. As we may observe, the subjects used 63% of the times the new designed workspace while the rest of the times they used the previous designed workspace.

Finally, we have performed preliminary clinical tests in order to determine the effectiveness of the mechanical improvements of the WTA- 1RII while performing the measurement of the carotid blood flow. In particular, we have asked a doctor and a sonographer to perform the measurements with a human volunteer by using the WTA-1 (before expansion of the workspace) and the WTA-1RII (after expansion of the workspace). During the experiment, the time required to perform the task was recorded for both rough positioning using the passive joint and fine positioning by means of the joystick-styled controller.

The experimental results are shown in Fig. 13. As we may observe, both subjects used the expanded workspace most of the time while performing the positioning adjustment task. However, as we also may notice, the overall time required for doing the measurement was not reduced significantly. This is may cause because of lack of training from the operators to use the joystick-styled controller.

IV. CONCLUSION & FUTURE WORK In this paper, we have presented the mechanical

improvements on the Waseda-Tokyo Women’s Medical-Aloka Blood Flow Measurement Robot System No. 1 Refined II (WTA-1RII). In particular a new gravity compensation mechanism has been designed by using a constant force spring and a sliding guide. Moreover, the

location of each of the links of the parallel manipulator of WTA-1RII has been optimized by using the Genetic Algorithm to assure that the target workspace.

A set of experiments were carried out to verify the effectiveness of the proposed mechanism. From the experimental results, we have confirmed the effectiveness of the proposed gravity compensation mechanism to almost complete cancel the gravity force. In addition, the optimization did with the GA assured the target workspace.

As a future work, we will extend the usability of the proposed robotic system to measure other parts of the body. Moreover, an automated measurement system will be proposed to improve the repetitiveness of the ultrasound diagnostics by implementing a feedback control based on the ultrasound image processing.

ACKNOWLEDGMENT This study has been conducted as a part of the humanoid

project at the Humanoid Robotics Institute, Waseda University

REFERENCES [1] K.H. Parker and C.J.H. Jones, "Forward and backward running waves

in the arteries: analysis using the method of characteristics," ASME J Biomech. Eng.Vol.112, pp. 322-326, 1990.

[2] C.J.H Jones, M. Sugawara, R.H. Davis, Y. Kondoh, K. Uchida, and K.H. Parker, Arterial wave intensity: physical meaning and physiological significance. In: S. Hosoda, T. Yaginuma, M. Sugawara, M.G. Taylor, and C.G. Caro (eds) , Recent progress in cardiovascular mechanics, Hanvood Academic Publishers, Chur, Switzerland, pp.129-148, 1994.

[3] Niki K., Sugawara M., Uchida K., Tanaka R., Tanimoto K., Imamura H., Sakomura Y., Ishizuka N., Koyanagi H., and Kasanuki H.., “A noninvasive method of measuring wave intensity, a new hemodynamic index: application to the carotid artery in patients with mitral regurgitation before and after surgery.”, Heart Vessels, vol. 14, pp.263-271, 1999.

[4] Harada, A., Okada, T., Sugawara, M, Niki, K. “Development of a Non-invasive Real-time Measurement System of Wave Intensity,” in IEEE Ultrasonics Symposium, pp. 1517-1520, 2000

[5] Sugawara M, Niki K, Ohte N, Okada T, Harada A., “Clinical usefulness of wave intensity analysis,” in Medical and Biological Engineering and Computing, doi: 10.1007/s11517-008-0388-x, 2008

[6] Sadamitsu Y., Fujita A., Arino C., Takanishi A., Harada A., M., Sugawara M. , Niki K., “Development of a Robotic Carotid Blood Flow Measurement System - A Compact Ultrasonic Probe Manipulator Consisting of a Parallel Mechanism, “ in Proc. of the IEEE International Conference on Robotics, Automation, and Mechatronics, pp. 1-6, 2006.

[7] Holland, H.J. Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control and Artificial Intelligence (2nd ed.), MIT Press, Cambridge, 1992.

Fig. 12. Experimental results while verifying the usability of the workspace while comparing the workspace between the previous mechanism and the one designed based on the GA.

Fig. 13. Experimental results while doing preliminary clinical tests to measure the carotid blood flow with the WTA-1R (before expansion) and the WTA-1RII (after expansion).

722