3D Ultrasound Navigation System with Reconstruction of ... - J-Stage

Advanced Biomedical Engineering 3 : 29-36, 2014.

Original Paper

3D Ultrasound Navigation System with Reconstruction of Blood Vessel Network for Microbubble Delivery Therapy Shinya ONOGI,*, # Tuan Hung PHAN,* Antoine BOSSARD,** Naoto HOSAKA,* Ren KODA,* Takashi MOCHIZUKI,* Kohji MASUDA*

Abstract This paper provides a 3D ultrasound navigation system with a blood vessel network (BVN) reconstruction algorithm for microbubble delivery therapy, which manipulates microbubbles in blood vessel by acoustic radiation force of ultrasound, aiming to increase the efficiency of high-intensity focused ultrasound (HIFU) and realize acoustic drug/gene delivery. To apply the technique in vivo, reconstruction of the BVN from 3D volume, such as bifurcation positions and flow direction, and visualization of the relative positions of the BVN and ultrasound field for microbubble manipulation are required. To address these issues, we have developed a 3D ultrasound navigation system to guide microbubbles in blood vessel. The novel navigation system consists of an ultrasound imaging device with a 2D array probe, an optical tracking device, a focused ultrasound transducer, and a Windows workstation with in-house navigation software. The system visualizes three-dimensionally both the BVN and the focal position of the transducer. When a 3D volume is acquired by an imaging device, the volume is automatically sent to the navigation system. Then, the system visualizes the volume and its bifurcations, which are estimated by 3D thinning processing. In this study, we examined the feasibility of the system by evaluating the guidance accuracy and microbubble induction rate. From the results, we confirm that microbubbles can be navigated by the system. Keywords : ultrasound navigation, blood vessel network, acoustic drug/gene delivery. Adv Biomed Eng. 3 : pp. 29-36, 2014.

1. Introduction Microbubbles are widely used as an echo contrast agent in ultrasound imaging to visualize blood vessel or flow velocity. On the other hand, microbubbles are also used for therapeutic purpose. For example, high-intensity focused ultrasound (HIFU) ablation is a minimally invasive therapy, but HIFU has very low efficient. Heat efficiency of HIFU can be increased by microbubbles because of their heat amplification1, 2. Another example is targeted gene or drug delivery. Microbubbles can be used as a carrier of gene or drug, which is activated around a target area by sonoporation3-5. However, the efficacy of these techniques is still limited because the microbubbles emitted diffuse throughout the whole body. To address this issue, we have researched on “acoustic microbubble delivery”, a method of manipulating microThis study was presented at the Symposium on Biomedical Engineering 2013, Fukuoka, September, 2013. Received on July 25, 2013 ; revised on November 1, 2013 and December 24, 2013 ; accepted on February 3, 2014. * Department of Bio-Applications and Systems Engineering, Tokyo University of Agriculture and Technology, Tokyo, Japan. ** Faculty of Information Systems Architecture, Advanced Institute of Industrial Technology, Tokyo, Japan. # BASE #522, 24-2 Naka-cho, Koganei-shi, Tokyo 184-8588, Japan. E-mail : [email protected]

bubbles in blood vessel using ultrasound. When microbubbles or micro-particles are radiated with ultrasound, the acoustic radiation force or primary Bjerknes force is applied to the bubbles or particles in the direction of ultrasound propagation. In addition, microbubbles can be aggregated by mutual interaction among microbubbles, an interaction induced by the secondary Bjerknes forces. By aggregate formation, we previously confirmed that microbubbles can be efficiently induced in artificial blood vessel because the primary force depends on the size of microbubbles6-9. To apply the technique in vivo, guidance provided by a navigation system to visualize the positions of bifurcations in blood vessel and sound field is required. Several navigation systems have been reported to serve as a guide during surgery ; for example, orthopedics surgery, endoscopic surgery, and interventional needle insertion for ablation10, 11. However, many of these systems require preoperative 3D volumes obtained from CT or MRI to visualize bones, organs, and tissues. While these volumes have high contrast and resolution, they are not only costly but also require intraoperative registration between the preoperative volume and the patient as well as compensation for deformation. As interventional procedures, 2D images by X-ray fluoroscope are used to monitor the process of needle insertion. However, guidance by 2D image is not intuitive and X-ray exposure of both surgeons and patients is inevitable. Recently, ultrasound imaging is also considered for intraoperative guidance. Ultrasound has less contrast and shorter depth than other modalities,

（30）

Advanced Biomedical Engineering. Vol. 3, 2014.

Fig. 1

System configuration of 3D ultrasound navigation system for acoustic microbubble delivery.

but has shorter time lag and is more portable. Therefore, we previously reported a navigation system using augmented reality for acoustic microbubble delivery12. We confirmed that microbubbles could be aggregated at a target position of an artificial blood vessel with guidance provided by the system. However, the system provided only the relative position of the B-scan plane and the focus of ultrasound, and the projected position of the focus on echogram. To realize acoustic microbubble delivery for therapy, 3D information, especially 3D blood vessel network structure, is required. A 3D navigation system using ultrasound is therefore needed for acoustic microbubble delivery. In this study, we developed a novel navigation system for manipulating microbubbles in blood vessel. Moreover, we implemented an algorithm for detection of blood vessel bifurcations from 3D ultrasound volume in order to visualize the bifurcation positions as targets. The algorithm and microbubble induction were validated.

2. Materials and Methods 2.1 System description To realize 3D ultrasound navigation for acoustic microbubble delivery, we developed a navigation system using 3D ultrasound imaging (Fig. 1). The navigation system consists of a 3D echography (iU22, Phillips Inc.) with a 2D array phased 3D probe (x6-1), an optical tracking device (Polaris Spectra, NDI Inc.), ultrasound transducers for microbubble manipulation, and a Windows workstation for the navigation software (Windows 7, Microsoft Co. Inc.). The system can visualize the focal points and sound axes of multiple transducers for aggregate formation and induction. The coordinate systems of the probe and the transducers to which trackers (four infrared reflective markers) were attached were integrated. Also, calibration between each device and the tracker is required. As

for transducer calibration, sound axis directions and focal positions of respective transducers were calibrated by an in-house transducer calibration jig12. As for ultrasound probe calibration, the fixed transformation between the coordinate systems of the attached tracker and the ultrasound volume was estimated. In this study, the method for a 2D probe was used 13. To apply this method to a 3D probe, the sweeping angle was fixed at 0 degree during calibration. Then, the center of the slices of a volume can be determined. Therefore the coordinates of 3D ultrasound volume can be obtained by simple translation of the calibrated position in the slice direction (Fig. 2). The navigation software was developed using C++ language (Visual Studio 2012, Microsoft Co. Inc.) and the Visualization Toolkit (VTK5.10.0, Kitware Inc.). The software has the following features : coordinate system transformation, bifurcation detection of blood vessel volume, visualization of spatial objects such as ultrasound volumes, sound axis and focal area of each transducer and bifurcation points, and communication with the tracking device. The coordinate transformations are updated at 60 Hz, which is the frame rate of the tracking device. For guidance of microbubble delivery, the system provides the relative transformation between a target bifurcation and the focal spot of a transducer (Fig. 3). Let  T be a transformation matrix from any coordinate system 2 to any coordinate system 1, the transformation is expressed as :  (1) T = T  T  T   T   T  where S, B, Tr, G, Pr, and Vol are the coordinate systems of a sound field, a bifurcation, a transducer, the global (the tracking device), the probe, and the volume, respectively. The transformation G TTr and G TPr are provided by the tracking device, Tr TS and Pr TVol are determined by the transducer and probe calibrations, respectively, and VolTB

Shinya ONOGI, et al : 3D US Navigation System for Microbubble Delivery Therapy

（31）

Fig. 2 3D ultrasound calibration by 2D calibration method.

Fig. 4

Reproducibility of bifurcation detection by pointbased registration.

different positions. From the volumes, the respective surfaces of the rubber ball were segmented, and the positions were estimated by a spherical fitting algorithm. Finally, the positions were transformed to the coordinate system G (tracking device) by the calibration matrix. The positions derived from both the tracking device and the echograph were compared.

Fig. 3

Relative transformation between bifurcations and a focal point of a transducer.

is obtained by the bifurcation detection algorithm described below. As for bifurcation detection, centerlines of blood vessel were computed from power Doppler volume acquired by 3D thinning processing. The 3D thinning algorithm has been reported previously 14. Then, a voxel connected with the other three voxels is labeled as a bifurcation.

3. Validations 3.1 3D probe calibration For 3D probe calibration, a steel ball 10.0 mm in diameter was used as a fiducial marker. The transformation matrix between the coordinate systems of the probe and the volume was estimated using 50 point sets measured by both the tracking device and the echograph. To validate the accuracy of the transformation, another ball made of rubber was used to acquire the ball surface clearly. First, the rubber ball position was measured by the tracking device. Next, volumes of the ball were measured at eight

3.2 Assessment of bifurcation detection To validate the reproducibility of the bifurcation detection algorithm, relative positions of the bifurcations detected were evaluated by point-based registration technique (Fig. 4). Twelve volumes of an artificial blood vessel made of poly(vinyl alcohol) were acquired under perfusion of Doppler test fluid, and bifurcations were detected by the bifurcation detection algorithm. Then, the bifurcations of the respective volumes were registered to the original bifurcation positions by point registration, and the registration error was computed. 3.3 Assessment of Navigation Accuracy For the evaluation of navigation accuracy, a cylindrical projection 2.5 mm in diameter was used as a target. The target point was selected manually in the ultrasound volume, and the point was picked using a calibrated pointer tool of the tracking device under guidance of the navigation (Fig. 5). Let Pointer be the coordinate system of the pointer, VolTPointer be the transformation between the coordinate systems of the volume and the pointer via the calibration matrix and the optical trackers, and Volp be the target position. The relative position between the pointer and the target TargetpPointer is described as :

（32）


Fig. 6

Fig. 5

Evaluation of navigation accuracy, and the target projection as a target.



p= T  T  (2) (3) p= T   T   T  The position pointer ptarget was measured as the navigation error when the pointer was placed at the target. The measurement was repeated five times using different volumes. To evaluate the orientation accuracy, the pointer orientation was measured by both the tracking device and ultrasound volume. In the volume, the needle was segmented manually, and the best fitting line was estimated by least square linear fitting. The tests were conducted 20 times. 

3.4 Microbubble induction by the system The final validation was the microbubble induction test under guidance of the navigation system. Microbubble induction rate was compared with the optimal result using jigs as reported previously7, 8. In this test, two ultrasound transducers were used ; one was for aggregate formation (Tu), and the other was for induction of aggregated microbubbles (Tb). Through aggregate formation by the transducer Tu, higher induction rate is achieved because of more effective primary Bjerknes force generated by the transducer Tb. A Y-shape poly(ethylene glycol) (PEG) vessel and a suspension of microcapsules (F-04E, Matsumoto Oil) were prepared. As shown in Fig. 6, optical image of the bifurcation and the observation area were respectively recorded by an optical microscope (KH-7700, Omron) and an inverted microscope (DMRIB, Leica) with a high-

Scheme of microbubble induction test under guidance of the navigation system.

speed camera (EX-F1, Casio) ; the former was for observation of microbubble behavior and the latter for calculation of microbubble induction rate. According to a previous report 8, both paths were extended using semitransparent tubes. Both tubes could be observed within a single view of the inverted microscope. Also, microbubble induction indexes (％) of both paths were computed from the average brightness values of both paths in the regions of interest (ROI) using the following formula7 : I −I −I −I  ×100, (4) ξ = I −I +I −I  where IA0 and IB0 are the average initial brightness without microbubbles in ROI A and B, respectively, to cancel the effect of background ; and IA and IB are the average brightness with microbubbles in the respective ROIs. We previously confirmed that the induction index reflects the relative ratio of microbubbles passing through ROI B to total microbubbles.

4. Results For 3D probe calibration, the time required for calibration was approximately 30 min, which included collecting 50 point sets and computing a calibration matrix. As for the calibration accuracy tested using the rubber ball, the error of the ball position was 1.40±0.36 mm (n=8), where the ground truth was measured by the tracking device. As for bifurcation detection, the registration results of the volumes are shown in Fig. 7. The registration error was 0.23±0.11 mm. For the evaluation of navigation accuracy, Fig. 8 shows a photograph of the experiment and the navigation view when the pointer was pointing at the target position.


Fig. 7

Fig. 8

（33）

Result of bifurcation detection. The detected bifurcations are shown by yellow dots.

Evaluation of navigation accuracy. Top : Navigation views (front and top). Bottom : photograph of the experiment.

The error between the target and the pointed position was 0.93±0.09 mm. As for the orientation accuracy, a sample of the acquired volume and the segmented surface is shown in Fig. 9. Mean error±standard deviation of 20 trials was 1.12±0.52 degree. Finally, the results of microbubble induction under guidance of the navigation system are given below. The transducers were guided by the navigation system (Fig. 10). When ultrasound was generated from both transducers, microbubble aggregation was observed, and the microbubbles entered the induced path B rather than path A (Fig. 11). The induction rate using the navigation was 24.6±2.5％. On the other hand, the induction rate at optimal positioning by the jig was reported to be within 22.9±2.3％, as shown in Fig. 12. No significant difference

Fig. 9

A sample of the volume and surface of the pointer for evaluation of orientation accuracy.

Fig. 10

Microbubble induction test under guidance of the navigation system.

was observed between the navigation and the jig.

5. Discussion In this study, a navigation system using 3D ultrasound was developed and validated with the objective to apply acoustic microbubble delivery in vivo. The navigation provides intuitive 3D information in nearly real-time 3D. To validate the feasibility of the navigation, we evaluated the accuracy of the 3D probe calibration, the reproducibility of the bifurcation detection algorithm, the accuracy of the navigation, and the basic performance of the transducer positioning by the microbubble manipulation test. As for the 3D probe calibration, the accuracy was 1.40 mm. The error includes the tracking error and the imaging error of the echograph. We calibrated in 40°C degassed hot water, thus the error derived from sound

（34）


Fig. 11 Microscope view of the experiment.

Fig. 12 Comparison of microbubble induction index between the jig and the navigation.

velocity was minimized. Also, the nominal error of the tracking device was 0.30 mm in root mean square. Therefore, the major error should be derived from the imaging resolution, especially the slice thickness. To improve the accuracy, a method of volume-based calibration would be effective, such as calibration using an orthogonal wired phantom15. As for the bifurcation detection algorithm, we confirmed that the detected bifurcation positions did not depend on probe position and were reproducible. However, the algorithm performance has to be validated using more complicated blood vessels with various diameters, curvatures, and morphologies. In the future, we plan to develop an algorithm for detecting error pattern based on anatomical information such as covering deficits, and eliminating error connections such as a loop. For the validation of navigation accuracy, the pointing error was 0.93 mm and 1.12 degree. The error includes the tracking error, calibration error, and pointing error. Segmentation error is also included in the orientation accuracy. The reason that the pointing error was smaller than the calibration error is that the cylindrical projection volumes were acquired around the

center of the volume because complete lateral surfaces of the projection were required. Meanwhile, the ultrasound beam width is 2 mm to 5 mm in diameter. Microbubbles can be manipulated when a target vessel is covered by the beam, therefore the required accuracy would be 1 to 2 mm. As for the required orientation accuracy, this error affects the induction efficiency because acoustic radiation force is reduced by cos Δθ, where Δθ is orientation error. Even if orientation error is 8 degree, the acoustic radiation force is reduced by 1％ (cos 8°=0.99). On the other hand, a therapeutic target should be acquired around the center of the volume for imaging accuracy in the clinical setting, thus the navigation accuracy reflects the actual accuracy in clinical situation. In the microbubble manipulation test, the transducers were positioned to the target positions under guidance of the navigation system. Also, microbubble behaviors including movement to propel around the target positions and aggregate formation were observed. The induction index was not significantly different between the jig and the navigation. Therefore, the navigation accuracy is sufficient for microbubble manipulation in 2.0-mm blood vessels. Our approach is not yet ready to be applied to in vivo use. One of the issues is the unstable movements caused by heartbeat, breathing, and body movements. To address this problem, we are considering two approaches ; one is update of target positions by continuous volume acquisition, and the other is robotic transducer positioning for the real-time targets. As for the robotic solution, we have developed a parallel link robot for this purpose16. While two transducers (Tu, Tb) were used in the present study, more transducers are required to manipulate microbubbles in multiple bifurcations. Therefore, we have developed a 2D array transducer, which can generate 3D sound field distribution17. In the future, these systems will be integrated to realize microbubble delivery therapy. Another issue is that the imaging probe and ultrasound transducers have to be in contact with the patientʼ s surface. Therefore, we are in the process of developing force and visual feedback control by a robot to automatically maintain appropriate contact condition. Finally, navigation error will increase in vivo because of inhomogeneous structures. An echograph constructs image or volume based on predicted average sound velocity in the body. Therefore, the difference between the average and actual velocities results in image distortion. However, like ultrasound for imaging, ultrasound for therapy is also distorted. Therefore, the effect of the imaging error would be small for ultrasound therapy with ultrasound navigation. This cancellation effect occurs because ultrasonic therapy is conducted under ultrasound guidance.

6. Conclusion This report presents a novel 3D navigation system to realize image guidance for acoustic microbubble delivery therapy. The system is capable of detecting bifurcation positions, visualizing the 3D blood vessel network, and


guiding transducers to the target bifurcation. Also, we have confirmed that microbubbles can be manipulated by the system as efficiently as the jig. These results suggest that the system has the potential to be applied to the clinical setting.

12.

Acknowledgement This study is supported by Japan Society for the Promotion of Science (JSPS) through the Funding Program for Next Generation World-Leading Researchers (NEXT Program).

13.

14.

References 1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

Stride S, Saffari N : The potential for thermal damage posed by microbubble US contrast agents. Ultrasonics. 42, pp. 907-913, 2004. Liu GJ, Moriyasu F, Hirokawa T, Rexiati M, Yamada M, Imai Y : Expression of heat shock protein 70 in rabbit liver after contrast-enhanced US and radiofrequency ablation. Ultrasound Med Biol. 36(1), pp. 78-85, 2010. Kudo N, Okada K, Yamamoto K : Sonoporation by singleshot pulsed US with microbubbles adjacent to cells. Biophys J. 96(12), pp. 4866-4876, 2009. Osawa K, Okubo Y, Nakao K, Koyama N, Bessyo K : Osteoinduction by microbubble-enhanced transcutaneous sonoporation of human bone morphogenetic protein-2. J Gene Med. 11(7), pp. 633-641, 2009. Juffermans LJ, van Dijk A, Jongenelen CA, Drukarch B, Reijerkerk A, de Vries HE, Kamp O, Musters RJ : US and microbubble-induced intra- and intercellular bioeffects in primary endothelial cells. Ultrasound Med Biol. 35(11), pp. 1917-27, 2009. Masuda K, Muramatsu Y, Ueda S, Nakamoto R, Nakayashiki Y, Ishihara K : Active path selection of fluid microcapsules in artificial blood vessel by acoustic radiation force. Jpn J Appl Phys. 48(7), 07GK03, 2009. Masuda K, Watarai N, Nakamoto R, Muramatsu Y : Production of local acoustic radiation force to constrain direction of microcapsules in flow. Jpn J Appl Phys. 49, 07HF11, 2010. Masuda K, Koda R, Watarai N, Shigehara N, Ito T, Kakimoto T, Miyamoto Y, Enosawa S, Chiba T : Experimental study of active control of microbubbles in blood flow by forming their aggregations. Proc 2011 IEEE Int Ultrason Symp. (IUSʼ11), Orlando, pp. 2021-2024, 2011. Masuda K, Nakamoto R, Watarai N, Koda R, Taguchi Y, Kozuka T, Miyamoto Y, Kakimoto T, Enosawa S, Chiba T : Effect of existence of red blood cell in trapping performance of microbubbles by acoustic radiation force. Jpn J Appl Phys. 50, 07HF11, 2011. Seth T, Chaudhary V, Buyea C, Bone L : A virtual interactive navigation system for orthopaedic surgical interventions. Proc 4th Int Symp Appl Sci Biomed Commun Technol. (ISABELʼ11). 71, 2011. Lin F, Lim D, Wixson RL, Milos S, Hendrix RW, Makhsous M : Validation of a computer navigation system and a CT

15.

16.

17.

（35）

method for determination of the orientation of implanted acetabular cup in total hip arthroplasty : A cadaver study. Clin Biomech. 23, pp. 1004-1011, 2008. Onogi S, Taguchi Y, Sugano Y, Shigehara N, Koda R, Bossard A, Masuda K : Navigation system with augmented reality for ultrasonic microbubble delivery therapy. Adv Biomed Eng. 1, pp. 16-22, 2012. Onogi S, Sugano Y, Yoshida T, Masuda K : An accurate calibration method of US images by center positions of a metal ball. Conf Proc IEEE Eng Med Biol Soc 2012 (EMBCʼ12), San Diego, pp. 468-471, 2012. Bossard A, Kato T, Onogi S, Masuda K : Towards realtime 3D reconstruction of the blood vessel network with automatic detection of vessel bifurcations and flow directions by ultrasound data analysis. Proc World Congress Med Phys Biomed Eng, Beijing, CD-ROM, 2012. Bergmeir C, Seitel M, Frank C, De Simone R, Meinzer HP, Wolf I : Comparing calibration approaches for 3D ultrasound probes. Int J Comput Assist Radiol Surg. 4(2), pp. 203-213, 2009. Irisawa S, Onogi S, Masuda K : Robotic 3D position control of therapeutic ultrasonic field by ultrasound image information. Proc 5th Biomed Eng Int Conf, Ubon Ratchatani, BME-2012-63, 2012. Koda R, Koido J, Hosaka N, Ito T, Onogi S, Mochizuki T, Masuda K, Ikeda S, Arai F : Active control of microbubbles stream in multi-bifurcated flow by using 2D phased array ultrasound transducer. Conf Proc IEEE Eng Med Biol Soc 2013 (EMBCʼ13), Osaka, pp. 6277-6280, 2013.

Shinya ONOGI Shinya ONOGI received the B.Eng. and M.Eng. degrees from Hokkaido University, Japan in 2002 and in 2004, respectively, and the Ph.D. degree in Science from the University of Tokyo, Japan in 2007. From 2007 to 2009, he was a Project Researcher of Intelligent Modeling Laboratory at the University of Tokyo. From 2009 to 2010, he was a Postdoctoral Fellow of Mechanical Engineering at Johns Hopkins University, USA. He is currently a Specially Appointed Assistant Professor (NEXT Program, JSPS) of Bio-Applications and Systems Engineering at Tokyo University of Agriculture and Technology, Japan. His research interests are computer assisted surgery/therapy based on robotics and information science. Tuan Hung PHAN Tuan Hung PHAN received the B. Eng. degree from Tokyo University of Agriculture and Technology, Japan in 2013. He has been working about medical image processing for the ultrasonic therapy for his master degree.

（36）


Antoine BOSSARD Antoine BOSSARD is an assistant professor of

Takashi MOCHIZUKI Takashi MOCHIZUKI received the B.Eng. and

the

Systems

Ph.D. degrees from Tokyo University of

Architecture at the Advanced Institute of

Agriculture and Technology, Japan, in 1976

Industrial Technology, Tokyo Metropolitan University. His research is focused on graph theory, interconnection networks and dependable systems. He received the B.E. and M.E. degrees from Université de Caen Basse-Normandie in 2005 and 2007, respectively, and the Ph.D. degree from the Tokyo University of Agriculture and Technology in 2011. He is a member of ACM.

and 1994, respectively. From 1976 to 2012, he worked for Hitachi Aloka Medical, Ltd. After the company retirement, he established a startup company to consult the ultrasound technology and has become CEO. Also, he has been a Specially Appointed Professor (NEXT Program, JSPS) of Graduate School of BioApplications and Systems Engineering, Tokyo University of Agriculture and Technology. His research interests include therapeutic applications in ultrasound and bioeffects of ultrasound.

Faculty

of

Information

Naoto HOSAKA Naoto HOSAKA received the M.E. degree in Engineering from Tokyo University of Agriculture and Technology, Japan in 2014. He is currently working about microbubble manipulation by four-dimensional ultrasound field using a two-dimensional matrix transducer in blood vessel for therapy for his Ph.D. degree. Ren KODA Ren KODA graduated Department of Engineering, Chiba University, Japan in 2009, received Masterʼs (Engineering) degree from Graduate school of Chiba University, Japan in 2011, and received Ph.D. degree from Tokyo University of Agriculture and Technology, Japan in 2014. He is currently a Research Assistant Professor of Graduate School of Science and Engineering for Research, University of Toyama, Japan. His research interest is ultrasonic therapy using microbubbles.

Kohji MASUDA Kohji MASUDA received the B.Eng., the M.Eng. and Ph.D. degrees from Osaka University, Japan, in 1991, 1993, and 1996, respectively. From 1996 to 1999, he was an Assistant Professor with Nagoya University, Japan. From 1999 to 2002, he was an Assistant Professor with Ehime University, Japan. From 2002 to 2003, he was an Postdoctoral Fellow with Universite Grenoble I-Joseph Fourier, France. Since 2003, he has been an Associate Professor with Graduate School of Bio-Applications and Systems Engineering, Tokyo University of Agriculture and Technology, Japan. His research interests include therapeutic applications in ultrasound, medical robotics, and medical image processing.