©PROJECT NEUROARM,UNIVERSITY OF CALGARY,CALGARY, ALBERTA

A Modular Approachto MRI-Compatible Robotics

The soft tissue contrast of magnetic resonance imaging(MRI) and its lack of harmful radiation make it a verysensitive and powerful diagnostic technique [1]. In con-trast, the existence of strong magnetic fields, spatial and

temporal field gradients, and radio frequency (RF) pulses in themagnetic resonance (MR) environment imposes severe designconstraints for any device present [2], requiring a number ofrestrictions if safety and image compatibility is to be assured [3].1) Ferromagnetic and highly paramagnetic materials are

prohibited inside and around the scanner as high forcesand torques can be induced because of the main staticmagnetic field and its associated spatial gradient.

2) Traditional electromagnetic (EM) actuators are also prohib-ited because of the mutual interaction between the fields ofthe motors and those present in the scanner room, imped-ing the optimal functioning of both devices.

3) Because of the RF pulses and gradient fields, electricalcurrents may be induced in conductive materials, whichcan cause considerable heating, especially when loopsare formed, potentially burning patients.

4) RF noise, which often appears as static on the image, iscaused by EM emissions from electrical devices that arepicked up by the sensitive RF coils of the scanner.

These restrictions mean that standard mechatronic devices arenot suitable for the MR environment and require that safetyand compatibility of any system and its operation in the mag-netic field be verified before being introduced into the scannerroom. Further details about the concept of MR compatibilitycan be found in the introduction to this special issue [4].

Closed cylindrical MRI scanners are often used for obtain-ing quality images of patient anatomy as they have high-fieldstrengths (typically 1–3 T in clinical environments) and betterfield homogeneity in the imaging region than do open scan-ners and, thus, can produce images with higher signal-to-noiseratio (SNR) [1]. The diameter of the bore is usually around600 mm, which must contain both the patient and the scannertable. In addition, most scanner bores are more than 150 cm inlength so that access by the practitioner to the part of thepatient anatomy that is being scanned (which is generallylocated around the magnet isocenter) is especially difficult.

These spatial constraints make freehand image-guided inter-ventions in a closed-bore scanner much more challenging. Anumber of MRI-compatible manipulators have been devel-oped with the objective of overcoming these difficulties [5].

MRI-Compatible ManipulatorsConsiderable progress is being made in the development ofMR-compatible manipulators designed to perform image-guided interventions, and this is reflected in the increasingbody of literature. Systems developed for specific proceduresinclude a prostate intervention device [6], an in-bore breastbiopsy system [7], a manipulator for MR-guided neurosurgi-cal interventions [8], and a robot for thermotherapy of livertumor [9] among others. A recent trend in the research field isthe development of MRI-compatible manipulators that areversatile and able to be used for a variety of surgical interven-tions [10]–[12]. These systems can be used to perform a num-ber of different procedures (although their scope is stillsomewhat limited) and, thus, require an increased number ofdegrees of freedom (DoF), making them all quite large in size.

The majority of the reported devices use piezoceramicmotors as an alternative to EM actuators, which do not usemagnetic principles for operation [13]. Although the piezocer-amic elements of the motors are not affected by the magneticfield of the scanner, they are often embedded inside conductivenonmagnetic materials, which can result in image artifacts.

The electric circuits of the motor drivers can introduce noisein the image, severely degrading the SNR while the motors arein motion. Some systems have tried using pneumatics andhydraulics [11], [14], [15] as actuation principles. Conventionalpneumatic cylinders can be adapted for use inside the MR scan-ner, although precise motions are not easily obtainable. Thishas led to the development of other forms of converting thepneumatic flow into accurate controllable movements [15].Another concern with these actuation methods is the presenceof fluids at high pressures in close proximity with the patient.

The piezoceramic motor that is used most for actuation isthe USR-30. The manipulators that incorporate this technol-ogy have generally followed one of the following two strat-egies. 1) The first consists in placing the motors at a distancefrom the imaging volume in the scanner and transferring themotion to the actuated elements via a transmission mechanism

BY HAYTHAM ELHAWARY,ALEKSANDAR ZIVANOVIC, MARC REA,BRIAN L. DAVIES, COLLIN BESANT,DONALD MCROBBIE, NANDITA M. DESOUZA,IAN YOUNG, AND MICHAEL U. LAMPERTH

MRI

ROBO

TICS

Using Robotic Modules withInterconnectable 1-DoF Stages

Digital Object Identifier 10.1109/EMB.2007.910260

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE 0739-5175/08/$25.00©2008IEEE MAY/JUNE 2008 35

using cables, rods, or rigid links [7], [10]. This introducesadditional complexity to the system as well as unwanted flexi-bilities and compliance. By separating the motors from thefield of view, an excessive degradation of the SNR of the MRimages is avoided. A separation of�1 m [7] is usually a suffi-cient distance to prevent the motors having any impact on theimages, allowing for actuation while the imaging is takingplace. 2) The second strategy consists in introducing themotors inside the scanner bore into direct contact with theactuated elements [8]. In this case, the motor has to be sepa-rated from the imaged region of interest by at least the size ofthe artifact it produces. In addition, the noise level of the scansis extremely high, requiring the motors to be turned off whilethe scanning and data acquisition is taking place. This hindersreal-time image guidance as imaging and motion cannot besimultaneous.

A Modular Approach to MR-Compatible RoboticsAn important part of the system-development time for anyMR-compatible manipulator is invested in hardware designand manufacture. An effective measure for reducing researchredundancy would be the availability of standardized off-the-shelf MR-compatible hardware that offered variability indimensions and shape. The objective of the research describedin this article is to create individual MR-compatible modulesconsisting of 1-DoF stages complete with actuators and posi-tion encoders for implementation of the closed-loop positioncontrol. These modules can connect together to form multi-DoF assemblies that can be located inside the scanner borenear to the patient anatomy that requires the intervention. Thisavoids the problems associated with remote actuation andtransmission mechanisms, considerably reducing the size ofthe manipulator. As most robots consist of kinematic chains of1-DoF stages, these modules would be suitable for a widerange of interventions, and their design can be optimized forthe procedure for which they are applied to.

Actuation TechnologyTo actuate each 1-DoF stage, piezoceramic motors (PiezoLEGS L motor, Piezomotor AB, Uppsala, Sweden) wereselected. They present a maximum force of around 10 N, topspeed of 12.5 mm/s (although with current drive electronicsthe speed is �4 mm/s), and a resolution of up to 2 nm. Thesemotors were more compact than the USR-30 model andpresented smaller and simpler drive electronics. Its operatingprinciple consists of a friction drive between four piezocer-amic crystals, which act as legs, and a linear ceramic rod. A

motor along with its drive electronics is shown in Figure 1(a).Each motor consists of a base unit, in which the piezo-legs arelocated, and a block mounted on top, which contains twoberyllium copper bearings and a series of leaf springs thatexert a preload on the ceramic rod, as shown in Figure 1(b). Asthe objective of this work is to develop MR-compatible hard-ware that can be actuated inside the bore while imaging is tak-ing place, a number of measures were implemented to reducethe EM noise coming from the motor drive circuits.

Position EncodingTo track the position of the modules, encoding technology isrequired. Traditional optical encoders tend to contain ferro-magnetic materials and are, therefore, unacceptable.

To overcome this problem, previous solutions have includedthe custom development of new encoders from nonmagneticmaterials [11], thereby leading the optical signals outside thescanner room via fiber optic cables to optoelectronic circuits[10], which then convert them into electrical pulses. A simplerand more cost-effective solution was found by using surfacemount incremental optical encoders from the AEDR-8000series. The size of the encoders (6.2 3 4.4 3 3.2 mm) is suffi-ciently small so that any incompatible material contained in theunit would only produce a small artifact. The encoder mountedon a printed circuit board (PCB) is shown in Figure 1(c).

Design of 1-DoF ModulesEach module consists of a base unit and a slider unit, as shown inFigure 2(a). The base unit is made of four blocks that contain twomotors, on opposite sides of the slider (doubling the total forcedelivered), and an encoder displayed in Figure 2(b). The sliderunit can translate back and forth and has two ceramic rodsattached on opposite sides that are in contact with the motors. Inthis design, the bearing and spring block has been removed fromthe piezoceramic motor [see Figure 2(b)], and the preloadbetween the piezo-legs and the ceramic rod is produced by meansof beryllium copper leaf springs [Figure 2(b)]. This allowed theapplication of a higher preload between the piezo-legs and theceramic rod, yielding an increased force output (wear particlesfrom beryllium copper are considered a health hazard, and someasures must be taken to avoid the scattering of these particles,i.e., enclosing the motors). The encoder on the base unit of thestage reads the position from a reflective encoder strip, which isplaced on the slider [Figure 2(b)]. At both extremes of the slider,there is an end stop, which acts as a mechanical block when thestage reaches the end of its travel. To connect the modulestogether, one of the end blocks at the extreme of each slider

forms part of the base unit ofthe next module, as can beseen in Figure 2(c).

An important characteristicof these 1-DoF modules is theirdesign variability. The dimen-sions of the modules can bevaried and optimized for thespecific intervention in whichthey will be used. The sliderand the base unit can be madeout of most MRI-compatibleplastics, depending on therequirements of rigidity,manufacturing method, cost

Bearing andSpring Block

(a) (b) (c)

Block withPiezo-Legs

CeramicRod

Fig. 1. (a) Piezoceramic motor with drive electronics. (b) The motor drives a linear, ceramicrod, which can vary in length. (c) Incremental surface mount encoder soldered onto a PCB.

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE MAY/JUNE 200836

constraints, and size. Rapid prototyping (RP) methods can beused, which are quick, cheap, and allowing considerable designflexibility, although they are available for only certain materialsand often have a limited tolerance range. When materials withhigher rigidity are required (and which are not available forRP), more traditional manufacturing techniques have to beimplemented, which can also offer tighter tolerances.

A prototype of a linear modular stage was made from Delrinwith slider dimensions of 30 3 30 3 140 mm. Three moduleswere produced in this way and connected together to form anXYZ Cartesian stage as displayed in Figure 2(c). At present,modules providing only linear motion have been manufac-tured, although current work includes adding curvature, withthe objective of manufacturing curved stages (as curvedceramic rods are available), which is highly desirable for anumber of robotic interventions.

Mechanical Characteristics of 1-DoF Linear ModulesOne of the modules was tested to obtain its load-speed characteris-tic, which is presented in Figure 3. The curve was obtained by fix-ing the base unit and applying an external load to the slider, andthe velocity was derived from the encoder readings. The graphshows that the stage can take a maximum load of�17 N, and runsat a top speed of around 2.7 mm/s. The current speed allows plentyof reaction time in case of any unforeseen circumstances during anintervention, although acquiring higher-speed drivers is not dis-carded for future prototypes. The stage was fitted with an encoderstrip with a resolution of 75 lines per inch (�85 lm accuracy).Because of the slow speeds and low inertia of the piezoceramicmotors, a bang-bang feedback control loop was implemented forposition control, which was sufficient to ensure that the accuracywas within a single encoder count. It also ensured repeatability ofpositioning the stage at a desired position. Accuracy tests were per-formed by moving the module-specific distances and comparingthe encoder readings with an external measurement device. Inevery case, accuracy was within 0.1 mm.

MRI Compatibility of ModulesWhen evaluating a system in the MR environment, both its MRsafety and compatibility must be verified. MR safety requiresthat the device does not experience hazardous torques, forces,and heating due to RF. MR compatibility is accepted when adevice does not significantly distort the diagnostic information

produced by an image. A device may produce artifacts if itpresents susceptibility differences with that of tissue or if it per-turbs the main static and RF field. These artifacts must be quan-tified to evaluate the use of these devices inside the scannerbore. Tests were performed for a piezoceramic motor, anencoder on its PCB, and a complete linear modular stage.

Two aspects of MR compatibility were assessed to quantify theimpact on the images: 1) the maximum size of any image artifactproduced by the device, and 2) how the device affects the SNR ofthe image when compared to a control image with a phantom.

To evaluate the size of the artifact produced, the standardF2119 protocol of the American Society for Testing and Mate-rials was followed. It requires testing the motor, encoder, andmodule at different orientations to the main field with a gradi-ent echo and spin-echo sequence while submerged in a CuSO4

solution. The maximum artifact is then measured and dis-played under the worse set of conditions (according to thestandard, the size of the artifact is measured from the boundaryof the device to the edge of the artifact it produces). The testswere done in a 1.5-T Siemens Magnetom Avanto using a spinecoil, and they were performed with the devices disconnected

0 0.5 1 1.5 2 2.5 30

2

4

6

8

10

12

14

16

18

Speed (mm/s)

Load

(N

)

Experimental DataRegression Curve

Fig. 3. Force-speed curve of a single linear module. Theexperimental data can be approximated to a linear rela-tionship between the two variables.

(a) (b) (c)

Slider Unit Limit Switch

Leaf Spring

Ceramic RodPiezoceramicMotor

PiezoceramicMotor

OpticalEncoder

Module 2

Module 1Module 3

Support Blockand BearingEncoder

Strip

Base Unit

Fig. 2. (a) A computer-aided design (CAD) representation of the developed 1-DoF module with its parts. (b) Detailed CADview of the base unit showing the encoder on the PCB and one of the motors. (c) Three Delrin stages interconnected to forman XYZ Cartesian robot. Supports are required to keep the deflection under load to a minimum.

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE MAY/JUNE 2008 37

from their power source. The results are presented in Figure 4.The size of these artifacts show that the devices are required tobe located at a small distance away from the region of interestto avoid distorting valuable image information.

To quantify SNR degradation due to the presence of a mod-ule, the following definition was used as in [10]:

SNR ¼ Pcenter

SDcorner

ð1Þ

where Pcenter is the mean signal in the 40 3 40 pixel region inthe center of the phantom, and SDcorner the standard deviationof the signal in the 40 3 40 pixel region at the bottom right-hand corner of the image. By quantifying how the devicedegrades the SNR when placed at varying distances from theisocenter, a suitable distance can be selected which willproduce acceptable images for MR guidance. These tests wererepeated for two different imaging sequences: turbo spin echo(TSE) and gradient echo [true fast imaging with steady-stateprecession (FISP)]. All tests were done on a 1.5-T SiemensMagnetom Avanto, using the body coil.

To reduce the EM interference of the modules with thescanner in the form of noise, the motor drive electronics, three6-V lead acid batteries, and motion control hardware werelocated in an aluminium-shielded enclosure, which acts as aFaraday cage, placed at 1–2 m from the entrance to the bore.The motor driver was connected to the motor in the bore via ashielded, twisted pair cable of�3 m in length, after being pre-viously filtered (4209-053, TUSONIX) through the wall of the

Faraday cage. The enclosure was then grounded to the scan-ner-ground connection.

For evaluation of the SNR reduction, a control image wasobtained by scanning only a phantom. A modular stage wasthen placed on the phantom and images were taken under twodifferent conditions: 1) the motors and encoder powered butwithout actuation and 2) with the motors actuated at maximumspeed implementing position control. The module was thenseparated from the isocenter and located at 50, 100, 150, 200,and 500 mm, and images were taken each time. At each dis-tance, the SNR variation was calculated by comparing theSNR measured at a slice at the isocenter with the SNR of thesame slice in the control image.

In Figure 5, the variation of SNR due to the presence of a1-DoF module is presented. Figure 5(a) corresponds to theTSE sequence and shows a maximum reduction in the SNR of9.1% when the module is placed at 50 mm from the isocenter.In Figure 5(b), the results for the true FISP sequence areshown, with a maximum reduction of SNR of 4.9% with themotor at 200 mm from the isocenter. The results show that thevariation of the SNR is very small at any distance and, in mostcases, is within the random fluctuations of SNR presentbetween different image acquisitions.

These compatibility tests show the feasibility of locatingthese modules almost anywhere within the imaging region,while ensuring MR compatibility under test conditions. Withregard to MR safety, susceptibility differences between thematerials of the motor, encoder, and module with that of tissueand air in the bore are quite small. This means that the forcesand torques induced on these devices is equally minute (whichcan be verified when one goes into the scanner with any of theelements), and, thus, MR safety is assured. In addition, becausethe motors and encoder are enclosed inside the module and notin direct contact with the patient, problems due to RF heatingwere considered negligible. Additional tests showing any con-sequences due to image shift and effects of the RF and gradientfields on robot accuracy (i.e., encoder immunity to the imagingfields) may complement the compatibility claims.

Formation of Kinematic ChainsOne of the main advantages of the developed hardware is itsmodularity, which permits connecting the individual stages tocreate multi-DoF kinematic chains. These chains can then beused to make robotic systems capable of image guided interven-tions. As the modules are made of MR-compatible plastics, it isespecially important to consider rigidity and accuracy when con-necting the numerous stages together. The selection of materials,design, and manufacturing techniques will depend on the finalconfiguration of the modules and must be decided accordingly.To demonstrate the connectability of the developed modules,three individual 1-DoF stages were connected in perpendicular

(a) (b) (c)

Fig. 4. MR image of the slice containing the maximum sizeartifact for (a) piezoceramic motor, 39 mm; (b) encoder onits PCB, 20 mm; and (c) 1-DoF module, 36 mm. In all cases,the artifacts were maximum when the devices were imagedusing the gradient-echo sequence (TR/TE ¼ 100/15 ms;matrix size, 256 3 256; FA, 30�, BW, 32 kHz; FOV, 160 mm;3-mm slice thickness; slice spacing, 0.6 mm; phase encodedirection H � F, spine coil). Small artifacts can also beobserved because of the fabric mesh that was used to sus-pend the devices in the CuSO4 solution.

With this work an important step toward

decreasing the hardware complexity of

MR-compatible manipulators has

been achieved.

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE MAY/JUNE 200838

directions to form an XYZ Cartesian stage configuration, which isshown in Figure 2(c). Two tests were then performed with theresulting robotic system: 1) deflection of the system under differ-ent loads and 2) capability of hitting a target with a needle in aphantom under MR guidance.

To evaluate the rigidity of the overall robot, forces wereapplied at the end of Module 3 in different directions and itsdeflection measured with a dial gauge. For small forces under20 N (the robot will be in contact with patient soft tissue), thedeflection under load was always less than 1 mm.

A needle and a needle-firing stage were then rigidly attachedto the Cartesian robot and supported through a guide on a probe.The system was then introduced into the scanner bore of a 1.5-TSiemens Magnetom Vision, and tests were performed to see if theneedle could be guided to hit a 6-mm target in a phantom underMRI guidance, as displayed in Figure 6(a) and (b). The imagesused for guidance were a gradient-echo sequence, and tests wereperformed with the head coil and the body coil. By locating thecoordinates of the target in the images, the needle was driven tothe tip of the target numerous times. An image showing the voidcreated by the needle and the target is presented in Figure 6(c).

Discussion and Future Work in the Field ofMRI-Compatible RoboticsWith this work, an important step toward decreasing the hard-ware complexity of MR-compatible manipulators has been

achieved. No longer do the actuators need to be placed outsidethe scanner bore and complicated transmission mechanisms putinto place to transmit motion to the end effector. These simplemodules can be modified and connected to form multi-DoFstructures, which are adapted and optimized for specific proce-dures, reducing hardware-development time and research redun-dancy. Small, compact, actuated systems located inside thescanner bore are now a reality, while still ensuring MR compati-bility and real-time image guidance. By selecting the correctmaterials and design, deflection under load can be kept to accept-able levels. The current embodiment of the modules reduces itsuse to interventions that require only kinematic chains with linearmotion, although curved stages are under development, whichshould allow more procedures to be performed.

The validation of the use of these modules to create robotstructures capable of performing interventions should takeplace in the near future in the form of clinical trials with atransrectal prostate biopsy robot.

Although a number of compatible manipulators have beendeveloped, only a small number report clinical trials [7], [11],[16]–[19] of any type, and two commercial systems are avail-able [20], [21]. This shows that although some technical solu-tions have been proposed for overcoming compatibility issuesand spatial constraints associated with the MR environment, anumber of clinical aspects still need to be addressed. MRI-compatible manipulators have to be developed from a real

Small, compact, actuated systems located

inside the scanner bore are now a reality, while

still ensuring MR compatibility and real-time

image guidance.

TSE Sequence

−6.2%

−9.1%

−3.7%

1000 100 200 300Distance from Magnet Isocenter (mm)

400 500

105

110

115SN

R 120

125

130

135

−5.0%

−3.8%

−5.8%

24

25

26

27

28

29

30

31

32

0 100 200 300Distance from Magnet Isocenter (mm)

(a) (b)

400 500

SN

R

True FISP Sequence

+0.9%

−1.9%

+2.0%+0.1%

−4.9% −4.8%

PhantomMotor onMotor Actuated

PhantomMotor PoweredMotor Actuated

Fig. 5. (a) Quantification of the SNR variation at the isocenter for a control image, first with a phantom, then with a modulepresent with its motors only powered, and finally with motors actuated. The 1-DoF module presents a maximum reduction of 9.1%,using a TSE sequence (TR ¼ 2,000 ms, TE ¼ 85 ms, FA ¼ 180�, 9 slices, 5-mm slices, 2.5 mm spacing, 256 3 256, FOV ¼ 300 mm,BW ¼ 130 Hz/px). (b) SNR variation under the same conditions with a true FISP sequence (TR ¼ 4.68 ms, TE ¼ 2.33 ms, FA ¼ 80�, 9 sli-ces, 5-mm slices, 2.5 mm spacing, 256 3 256, FOV ¼ 230 mm, BW ¼ 130 Hz/px). The maximum reduction in SNR is 4.9%.

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE MAY/JUNE 2008 39

interventional need and used in procedures where the benefitwill overcome the significant costs and complexity derivedfrom introducing such systems into the already expensive andscarce MR suite. The enormous capabilities of MRI must beused in a way that will facilitate the interventional task, andthe integration of this information with the control and motionof the manipulator needs to be intuitive, well presented, andeffective, while not increasing in excess the workload of thepractitioner.

Another element to be addressed is the variety of scannersavailable and the development of systems that can be used ondifferent scanner types and models. In this context, it is impor-tant either to provide systems that do not need to interfacedirectly into the scanner software or to offer some other solu-tion that is easily transferable from one platform to the next. Inaddition, the current tendency in new cylindrical MRI scan-ners is to increase the diameter of the scanner bore whilereducing its length (the Siemens Magnetom Espree has a borediameter of 70 cm and a bore length of 125 cm), which willrelieve some of the severe spatial constraints of traditionalscanners, changing the requirements of MR-compatible manip-ulators, as free-hand intervention might then be possible. In thiscontext, manipulators will have more of a function to help guidethe freehand intervention, shifting the research on the side ofnavigation and computer-assisted intervention.

AcknowledgementThis study forms a part of the project funded by the New andEmerging Applications of Technology (grant code D083) pro-gram of the Department of Health, UK.

Haytham Elhawary received his M.Eng.in mechanical engineering from the Schoolof Engineering and Telecommunications(Technological Campus) of the Universityof Navarra, Spain. He is currently workingtoward his Ph.D. degree at the mecha-tronics and medicine laboratory at Impe-rial College, London, United Kingdom.

His current research interests are in the field of medicalrobotics and, more specifically, the design and development ofMRI-compatible mechatronic systems for intervention andguidance.

Aleksandar Zivanovic received his B.Sc.(Hons.) in computer systems engineering(informatics) and M.Sc. in electronic engi-neering, both from the University of Kentat Canterbury, United Kingdom. He re-ceived his Ph.D. degree in mechanicalengineering from the Imperial College,London, United Kingdom, where he is cur-

rently a research associate. His research interests are in theapplications of mechatronics in medicine, in particular, thedevelopment of MRI-compatible manipulators and haptic sys-tems for training and intervention.

Marc Rea received an M.Phys. degreefrom the School of Physics, LancasterUniversity, in 1999. He received twoM.Sc. degrees, one in interactive multime-dia systems from John Moores University,Liverpool, and another in medical physicsfrom Leeds University. Currently, he isworking as a trainee clinical scientist in

MRI while working toward a Ph.D. degree at Imperial Col-lege, London, United Kingdom. His current research interestsare in the field of MRI-compatibility of devices and systems,including the automatic tracking of mechatronic deviceswithin the MRI environment.

Brian L. Davies is the technical directorof the Acrobot Company, Ltd., London,United Kingdom, and emeritus professorof medical robotics at Imperial College,London, where he founded and leads theMechatronics in Medicine Research Group.Recently, he received the D.Sc. degree forhis international contribution to medical

robotics, and he was awarded the Maurice Muller Prize by theInternational Society of Computer-Aided Orthopedic Surgeryfor his contributions to medical research. He is also the recipi-ent of the IEE award for Outstanding International Contribu-tions to Computer-Aided Surgery.

Collin Besant obtained his engineering degree from LondonUniversity and completed his Ph.D. degree in nuclear

(a) (b) (c)

Target

Needle

Fig. 6. (a) The XYZ Cartesian robot placed inside the scanner, with a grapefruit phantom. The system is being prepared forneedle insertion into a target of 6 mm under MRI guidance. (b) The needle adaptor can be seen on the manipulator. (c) MRimage of the phantom and target, with the needle artifact leading to the void left by the target (TSE, FOV ¼ 170 mm; matrixsize, 256 3 256; TR/TE ¼ 2,000/120 ms; 5-mm slice thickness; 5-mm slice spacing, head coil).

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE MAY/JUNE 200840

engineering in 1960. He then joined Impe-rial College, London, United Kingdom,where he built a large research groupworking on engineering design and manu-facturing systems, with particular empha-sis on industrial needs. He was made afellow of the Royal Academy of Engineer-ing in 1987 and a professor of computer-

aided manufacturing in 1989. He is the CEO of the spin-offcompany Turbo Genset.

Donald McRobbie received his B.Sc. inphysics and M.Sc. in medical physics fromthe University of Aberdeen. He was awardedhis Ph.D. degree in magnetic resonanceimaging from the University of London in1992. He is currently the head of radiologi-cal and MR physics and deputy director ofthe radiological sciences unit in the Ham-

mersmith Hospitals, National Health Service Trust. He is ahonorary senior lecturer in imaging at the Imperial College,London. His research interests include functional MRI andvoxel-based morphometry, 3-D imaging of the airways, MRdevice-tracking and guidance, and dosimetry in computerizedtomography. He is the author of the award-winning textbookMRI from Picture to Proton (Cambridge University Press, 2003).

Nandita M. deSouza is a reader in imag-ing and the codirector of the MRI Unit atthe Institute of Cancer Research. Shereceived her B.Sc. (Hons.) in physiologyin 1978 from the University of Newcastleupon Tyne, United Kingdom. She receivedher MBBS, MRCP, FRCR, MD, and FRCPdegrees in 1986, 1991, 1996, 1983, and

2001, respectively. She has been a reader in imaging at theImperial College from 2000 to 2002 and a consultant in imag-ing at the Hammersmith Hospital, National Health ServiceTrust, from 2002 to 2004. Her main research interests are ingynecology, prostate and breast tumors, using MRI to under-stand biology, and improving staging and monitoring treat-ment response. She holds a Cancer Research U.K. programmegrant in MRI and several project and studentship grants. Shehas more than 90 peer-reviewed articles, several book chap-ters, and is the editor of a multiauthor book on endocavitaryMRI of the pelvis.

Ian Young is a senior research fellow in the Electrical andElectronic Engineering Department, Imperial College,London, United Kingdom. He has been working in thefield of MRI for over 30 years.

Michael U. Lamp�erth is currently a lecturerat the Mechanical Engineering Department,Imperial College, London, where he leads anumber of researchers in the Mechatronicsin Medicine Laboratory. He received hiseducation as a mechanical engineer from theMSW Winterthur Professional School andCollege of Technology, Switzerland, and

obtained his M.Sc. and Ph.D. degrees at Imperial CollegeLondon. His research interests are mainly in the field of medical

robotics and implants, with a focus on using MRI for real-timeguidance and planning. He also conducts research in hybrid elec-tric vehicles.

Address for Correspondence: Michael Lamp�erth, Depart-ment of Mechanical Engineering, Imperial College London,South Kensington Campus, SW7 2AZ London, UK. E-mail:m.lamperth@imperial.ac.uk.

References[1] J. Kettenbach, D. F. Kacher, S. K. Koskinen, S. G. Silverman, A. Nabavi,D. Gering, C. M. C. Tempany, R. B. Schwartz, R. Kikinis, P. M. Black, andF. A. Jolesz, ‘‘Interventional and Intraoperative Magnetic Resonance Imaging,’’Annu. Rev. Biomed. Eng., vol. 2, no. 1, pp. 661–690, 2000.[2] J. Kettenbach, D. F. Kacher, A. R. Kanan, B. Rostenberg, J. Fairhurst,A. Stadler, K. Kienreich, and F. A. Jolesz, ‘‘Intraoperative and interventionalMRI: Recommendations for a safe environment,’’ Minim. Invasive Ther. AlliedTechnol., vol. 15, no. 2, pp. 53–64, 2006.[3] FDA. (2005 Feb.). A primer on medical device interactions with magneticresonance imaging systems. [Online]. Available: http://www.fda.gov/cdrh/ode/primerf6.html.[4] R. Gassert, E. Burdet, and K. Chinzei, ‘‘A critical tool for image guided inter-vention, clinical diagnostics and neuroscience,’’ IEEE Eng. Med. Biol. Mag.,2007.[5] H. Elhawary, A. Zivanovic, B. Davies, and M. Lamperth, ‘‘A review of mag-netic resonance imaging compatible manipulators in surgery,’’ Proc. Inst. Mech.Eng., Part H: J. Eng. Med., vol. 220, no. 3, pp. 413–424, 2006.[6] A. Krieger, R. C. Susil, C. Menard, J. A. Coleman, G. Fichtinger, E. Atalar,and L. L. Whitcomb, ‘‘Design of a novel MRI compatible manipulator for imageguided prostate interventions,’’ IEEE Trans. Biomed. Eng., vol. 52, no. 2,pp. 306–313, 2005.[7] B. T. Larson, A. G. Erdman, N. V. Tsekos, E. Yacoub, P. V. Tsekos, andI. G. Koutlas, ‘‘Design of an MRI-compatible robotic stereotactic device for mini-mally invasive interventions in the breast,’’ J. Biomech. Eng. – Trans. ASME,vol. 126, no. 4, pp. 458–465, 2004.[8] K. Masamune, E. Kobayashi, Y. Masutani, M. Suzuki, T. Dohi, H. Iseki, andK. Takakura, ‘‘Development of an MRI-compatible needle insertion manipulatorfor stereotactic neurosurgery,’’ J. Image Guided Surg., vol. 1, no. 4, pp. 242–248,1995.[9] N. Hata, F. Ohara, R. Hashimoto, M. Hashizume, and T. Dohi, ‘‘Needle guid-ing robot with five-bar linkage for MR-guided thermotherapy of liver tumor,’’inLecture Notes in Computer Science, vol. 3217, Part 2, Berlin, Germany: Springer,2004, pp. 161–168.[10] K. Chinzei, N. Hata, F. A. Jolesz, and R. Kikinis, ‘‘Surgical assist robot forthe active navigation in the intraoperative MRI: Hardware design issues,’’ inProc. 2000 IEEE/RSJ Int. Conf. Intelligent Robots and Systems, Maui, HIpp. 727–732.[11] E. Hempel, H. Fischer, L. Gumb, T. Hohn, H. Krause, U. Voges,H. Breitwieser, B. Gutmann, J. Durke, M. Bock, and A. Melzer, ‘‘An MRI-compatible surgical robot for precise radiological interventions,’’ Comput. AidedSurg., vol. 8, no. 4, pp. 180–191, 2003.[12] N. V. Tsekos, A. Ozcan, and E. Christoforou, ‘‘A prototype manipulator formagnetic resonance-guided interventions inside standard cylindrical magneticresonance imaging scanners,’’ J. Biomech. Eng., vol. 127, no. 6, pp. 972–980,2005.[13] K. Uchino, ‘‘Piezoelectric ultrasonic motors: Overview,’’ Smart Mater.,vol. 7, no. 3, pp. 273–285, 1998.[14] R. Gassert, R. Moser, E. Burdet, and H. Bleuler, ‘‘MRI/fMRI-compatiblerobotic system with force feedback for interaction with human motion,’’ Mecha-tron. IEEE/ASME Trans., vol. 11, no. 2, pp. 216–224, 2006.[15] E. Taillant, J.-C. Avila-Vilchis, C. Allegrini, I. Bricault, and P. Cinquin,‘‘CT and MR compatible light puncture robot: Architectural design and firstexperiments,’’ presented at the Conf. Medical Image Computing and Computer-Assisted Intervention, Rennes, St. Malo, France, 2004.[16] D. Beyersdorff, A. Winkel, B. Hamm, S. Lenk, S. A. Loening, andM. Taupitz, ‘‘MR imaging-guided prostate biopsy with a closed MR unit at 1.5 T:Initial results,’’ Radiology, vol. 234, no. 2, pp. 576–581, 2005.[17] H. Fischer, S. Kutter, J. Vagner, A. Felden, S. O. R. Pfleiderer, andW. A. Kaiser, ‘‘ROBITOM II: Robot for biopsy and therapy of the mamma,’’presented at the 2004 IEEE Int. Conf. Systems, Man and Cybernetics.[18] W. A. Kaiser, H. Fischer, J. Vagner, and M. Selig, ‘‘Robotic system forbiopsy and therapy of breast lesions in a high-field whole-body magnetic reso-nance tomography unit,’’ Invest. Radiol., vol. 35, no. 8, pp. 513–519, 2000.[19] R. C. Susil, K. Camphausen, P. Choyke, E. R. McVeigh, G. S. Gustafson,H. Ning, R. W. Miller, E. Atalar, C. N. Coleman, and C. Menard, ‘‘System forprostate brachytherapy and biopsy in a standard 1.5 T MRI scanner,’’ Magn.Reson. Med., vol. 52, no. 3, pp. 683–687, 2004.[20] Insightec. (2005 Mar.) ExAblate 2000—Focused ultrasound ablation device.[Online]. Available: www.insightec.com[21] Innomedic. (2007 Dec.). Innomotion. [Online]. Available: http://www.innomedic.de/en/index.php

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE MAY/JUNE 2008 41