[ieee 2008 cairo international biomedical engineering conference (cibec) - cairo, egypt...

9
1 Operating Room of the Future Orthopedic Perspective Mohamed R. Mahfouz Mechanical Aerospace and Biomedical Engineering Department University of Tennessee, USA mmahfouz(@,cmr. utk. edu Abstract - The complexity of orthopedic procedures has mounted with increasing numbers of minimally invasive surgeries. Lack of optimal patient outcomes persist despite new techniques and improved implants. While improved training for physicians can enhance patient outcomes, computer assisted surgery (CAS) has the capability to provide even greater benefits to the patient by increasing control and repeatability. Cutting edge technology fused with new computerized techniques facilitates a full cycle of implant design and development by providing surgical preplanning and intraoperative guidance followed by post-operative gait analysis. In the pre-operative phase, these technologies can reconstruct patient specific bone models, automate cutting plane placement, and highlight anatomical abnormalities. In the intra-operative phase, the state-of-the-art CAS systems virtualize surgical protocols, provide real-time hyper-resolution micro-sensor feedback in ligament balancing, and wireless navigation guidance. Finally, in post-operative scenarios, disruptive technologies enable improved implant design through acquisition and analysis of 3D kinematic gait lab data. By introducing novel technologies and advanced computerized methods into the operating room (OR), the next generation of CAS systems will further surgeons' ability to control positive patient outcomes. I. OVERVIEW N Inimally invasive surgeries (MIS) require a new suite of IVitechnologies to lead the orthopedic surgeon through pre-operative, intra-operative, and post-operative procedures. Each phase in MIS will be discussed with highlights given to new technologies currently being designed at the Center for Musculoskeletal Research, University of Tennessee to address the new problems associated with MIS. A. Pre-Operative Phase With recent advances in surgical navigation systems, demand for higher accuracy and patient specific model creation has increased. CT as a method for patient specific model creation is very accurate; however, the high radiation dose and cost is a major constraint. Thus arise the need for a new technology for patient specific bone creation, such as biplanar reconstruction [1,2]. In biplanar reconstruction three dimensional patient specific models are created utilizing two x-ray images, AP and lateral, that do not have to be exactly perpendicular to one another. A calibration target is attached to the patient's femur and tibia to allow the algorithm to calculate the position of the x-ray machine relative to the patient. Biplanar reconstruction utilizes sex specific bone atlases to capture the variation across a population by varying the shape parameters, or principle components (PCA), extracted from the statistical atlas. Using this, the software can reconstruct new bones. The first step in reconstruction involves extracting both intrinsic and extrinsic camera parameters by extracting the calibration target from the images. These camera parameters are then used to setup the virtual scene. The mean model is then projected with the same camera parameters to simulate the x-rays. Shape parameters along with the pose parameters (3 rotations and 3 translations) are then optimized to find optimal shape and pose parameters. Fitness of the generated model is evaluated by comparing the projection of the generated model to the two x-ray images. This biplanar process has the potential to revolutionize preoperative surgical preplanning and will be further discussed in later sections of this paper. Next step after creation of patient specific bone models is surgical preplanning. This includes identifying surgical axes and landmarks along with sizing and determining the placement of the implant. These preplanning steps are done automatically utilizing gender specific statistical atlases [3]. We developed a 3D-3D surface matching algorithm which matches 3D surface meshes originating from high-resolution CT images with a high degree of accuracy. This initial step creates homologous point sets across all of the femora in the dataset, which is necessary for the creation of a statistical atlas. We will discuss how this method is used in automatically calculating the cutting planes after obtaining 3 -D patient specific bone models from biplanar reconstruction. These cutting planes are then used in the intra-operative phase to assist the surgeon in performing the surgery (we will focus on Total Knee Arthroplasty). B. Intra-Operative Phase New technologies are needed in the operating room to seamlessly incorporate orthopedic surgical navigation with minimally invasive surgical procedures. Current surgical navigation technology includes bone tracking via optical or electromagnetic trackers which must be fixated to the bone by drilling. Optical or electromagnetic trackers can also be used to track cutting blocks and pick clouds of points on the bone. Therefore, in many respects current surgical navigation technology is not compatible, or is lagging behind the needs of minimally invasive surgery. Only by introducing disruptive technologies to displace the current status quo will orthopedic surgical navigation systems be able to assist in difficult minimally invasive surgical procedures. Ultra Wideband (UWB) technology is used to introduce wireless instrument tracking into the operating room (OR). 978-1-4244-2695-9/08/$25.00 ©)2008 IEEE Proceedings of the 2008 IEEE, CIBEC'08

Upload: mohamed-r

Post on 28-Mar-2017

222 views

Category:

Documents


9 download

TRANSCRIPT

Page 1: [IEEE 2008 Cairo International Biomedical Engineering Conference (CIBEC) - Cairo, Egypt (2008.12.18-2008.12.20)] 2008 Cairo International Biomedical Engineering Conference - Operating

1

Operating Room of the FutureOrthopedic Perspective

Mohamed R. MahfouzMechanical Aerospace and Biomedical Engineering Department

University ofTennessee, USA

mmahfouz(@,cmr. utk. edu

Abstract - The complexity of orthopedic procedures hasmounted with increasing numbers of minimally invasivesurgeries. Lack of optimal patient outcomes persist despite newtechniques and improved implants. While improved training forphysicians can enhance patient outcomes, computer assistedsurgery (CAS) has the capability to provide even greater benefitsto the patient by increasing control and repeatability. Cuttingedge technology fused with new computerized techniquesfacilitates a full cycle of implant design and development byproviding surgical preplanning and intraoperative guidancefollowed by post-operative gait analysis. In the pre-operativephase, these technologies can reconstruct patient specific bonemodels, automate cutting plane placement, and highlightanatomical abnormalities. In the intra-operative phase, thestate-of-the-art CAS systems virtualize surgical protocols, providereal-time hyper-resolution micro-sensor feedback in ligamentbalancing, and wireless navigation guidance. Finally, inpost-operative scenarios, disruptive technologies enable improvedimplant design through acquisition and analysis of 3D kinematicgait lab data. By introducing novel technologies and advancedcomputerized methods into the operating room (OR), the nextgeneration of CAS systems will further surgeons' ability to controlpositive patient outcomes.

I. OVERVIEW

N Inimally invasive surgeries (MIS) require a new suite ofIVitechnologies to lead the orthopedic surgeon throughpre-operative, intra-operative, and post-operative procedures.Each phase in MIS will be discussed with highlights given tonew technologies currently being designed at the Center forMusculoskeletal Research, University of Tennessee to addressthe new problems associated with MIS.

A. Pre-Operative PhaseWith recent advances in surgical navigation systems,

demand for higher accuracy and patient specific model creationhas increased. CT as a method for patient specific modelcreation is very accurate; however, the high radiation dose andcost is a major constraint. Thus arise the need for a newtechnology for patient specific bone creation, such as biplanarreconstruction [1,2]. In biplanar reconstruction threedimensional patient specific models are created utilizing twox-ray images, AP and lateral, that do not have to be exactlyperpendicular to one another. A calibration target is attached tothe patient's femur and tibia to allow the algorithm to calculatethe position of the x-ray machine relative to the patient.

Biplanar reconstruction utilizes sex specific bone atlases tocapture the variation across a population by varying the shape

parameters, or principle components (PCA), extracted from thestatistical atlas. Using this, the software can reconstruct newbones. The first step in reconstruction involves extracting bothintrinsic and extrinsic camera parameters by extracting thecalibration target from the images. These camera parametersare then used to setup the virtual scene. The mean model is thenprojected with the same camera parameters to simulate thex-rays. Shape parameters along with the pose parameters (3rotations and 3 translations) are then optimized to find optimalshape and pose parameters. Fitness of the generated model isevaluated by comparing the projection of the generated modelto the two x-ray images. This biplanar process has the potentialto revolutionize preoperative surgical preplanning and will befurther discussed in later sections of this paper.Next step after creation of patient specific bone models is

surgical preplanning. This includes identifying surgical axesand landmarks along with sizing and determining theplacement of the implant. These preplanning steps are doneautomatically utilizing gender specific statistical atlases [3].We developed a 3D-3D surface matching algorithm whichmatches 3D surface meshes originating from high-resolutionCT images with a high degree of accuracy. This initial stepcreates homologous point sets across all of the femora in thedataset, which is necessary for the creation of a statistical atlas.We will discuss how this method is used in automaticallycalculating the cutting planes after obtaining 3-D patientspecific bone models from biplanar reconstruction. Thesecutting planes are then used in the intra-operative phase toassist the surgeon in performing the surgery (we will focus onTotal Knee Arthroplasty).

B. Intra-Operative PhaseNew technologies are needed in the operating room to

seamlessly incorporate orthopedic surgical navigation withminimally invasive surgical procedures. Current surgicalnavigation technology includes bone tracking via optical orelectromagnetic trackers which must be fixated to the bone bydrilling. Optical or electromagnetic trackers can also be used totrack cutting blocks and pick clouds of points on the bone.Therefore, in many respects current surgical navigationtechnology is not compatible, or is lagging behind the needs ofminimally invasive surgery. Only by introducing disruptivetechnologies to displace the current status quo will orthopedicsurgical navigation systems be able to assist in difficultminimally invasive surgical procedures.

Ultra Wideband (UWB) technology is used to introducewireless instrument tracking into the operating room (OR).

978-1-4244-2695-9/08/$25.00 ©)2008 IEEEProceedings of the 2008 IEEE, CIBEC'08

Page 2: [IEEE 2008 Cairo International Biomedical Engineering Conference (CIBEC) - Cairo, Egypt (2008.12.18-2008.12.20)] 2008 Cairo International Biomedical Engineering Conference - Operating

2

UWB has the potential to reduce sensor/probe size and increaseaccuracy compared to current optical and electromagnetictracking systems. At the same time, it does not have the sameline-of-sight (LOS) limitations of optical tracking systems orthe short range limitations of electromagnetic trackingsystems[4]. Additionally, A-mode ultrasound can be used tononinvasively track the position of the bones intraoperatively.This includes picking clouds of points on the bones by runningan ultrasound probe over skin and also tracking the orientationof the patient's bones through the use of an ultrasound brace.Combining this ultrasound technology for noninvasive bonetracking with UWB technology for wireless tracking of theinstruments, ultrasound probes, and ultrasound braces providesa complete solution for orthopedic surgical navigation systemsof the future.

The development of a smart provisional or spacer instrumentto assist the surgeon during total knee arthroplasty (TKA) toimprove the accuracy of implant alignment and soft tissuebalancing has been underway at the University of Tennessee.These smart sensors have been designed and tested in vitro asoutlined in this paper. First, piezo-resistive microcantileverswere used to measure macro forces on a surgical spacer block.An interface for the microcantilevers as well as the design ofread-out electronics to take the measurements is outlined in[5,6]. Additionally, a wireless communication system was alsotested, showing that it is possible to implement telemetry withthe microcantilever sensing system. Finally, custom capacitivemicroelectromechanical system (MEMS) arrays were alsodeveloped and tested for conformable [7] arrays to provideimproved spatial and axial force resolution on the curvedsurface of a provisional. Multiple sensor design iterations havebeen completed and verified [8], and custom high-capacity,high-resolution read-out electronics were also developed forinterpreting the signal from these capacitive sensors. For bothtypes of sensors, the microcantilevers and the capacitiveMEMS sensor array, identification of force magnitudes andlocation of those forces is possible in real-time for surgicalfeedback.

C. Post-Operative PhaseUtilizing fluoroscopy for three-dimensional (3D) in vivo

analysis is a powerful tool with widespread applications forpost-operative analysis of arthoscopic surgeries. It allowsjoints (both implanted and normal) to be studied in vivo underweight bearing conditions through normal day-to-day activitiessuch as walking, leg bending, arm raising, etc. Differentmethods have been created to extract 3D information (i.e.translation, rotation) from 2D fluoroscopic data. Our robust3D-to-2D registration method allows 3D data to be extractedmore accurately than similar techniques which employ anadded preliminary segmentation step [9]. After extracting the3D data, many different joint-specific 3D analyses can beperformed. This includes automatic calculation of hipseparation, loci tracking of different joints (e.g. hip, vertebrae,shoulder) [10], automatic tracking of ligament attachments, etc.We have also extended our 3D-to-2D registration method toautomatically fit a large number of closely spaced images (e.g.200-500), which we term "video fitting"[11 ]. This providessmoother motion of the implanted models through a givenactivity. Finally, after the relative transformation of the femur

and polyethylene components is known, we are able tocalculate the contact area between the two models. Thisprovides contact area information over a range of activities formore advanced analysis of the knee joint which could be usedas input into a finite element model. By continuing to expandour tools for 3D in vivo analysis, we wish to enhance the overallunderstanding of orthopedic kinetics and kinematics throughadvanced post-operative procedures.

II. METHODOLOGY

A. Pre-Operative PhaseThe process our biplanar reconstruction method follows is

outlined in Figure 1. Image processing is completed on the twox-ray images as pre-processing and enhancement steps.Automated morphometric measurements are performed on thebiplanar images to estimate the shape and size ofthe bone. Thisincreases the speed and accuracy of the registration process. Acalibration targets attached to patient femur and tibia figure arethen used to extract intrinsic and extrinsic camera parameters[12].These parameters are then used to reconstruct the 3D sceneas shown in figure 2.The average bone from the sex-specific atlas is placed in a 3Dscene with two or more X-ray images in an initial pose for 3Drigid alignment. The average bone and initial pose are used tocreate a normally distributed initial population of bone models(represented in the atlas) centered around the average bone atthe initial pose. The shape parameters, translation, and rotationcombine for 11 degrees of freedom (DOF) optimizationproblem. The genetic algorithm is then used, along with ournovel 3D-2D scoring metric and the morphometricmeasurements, to optimize both the shape and alignment of thepopulation to the X-ray images.The algorithm to register the bone of interest to both images is

posed as an optimization problem of both predicted modelshape and orientation. The objective function consists of thesum of a matching metric calculated for each image. Thus theobjective function is the combination of matching scorescomputed for individual images. To compute the matchingscore for each image, a hypothesized pose and shape of thebone model are projected onto the virtual image plane for each

XwrayrnlffageSi g Chbmage

Gene...

Algor.ihw.

Bonee Atlasa2|

Ditettion Set

3D Mane

Figure 1 - Biplanar reconstruction process

978-1-4244-2695-9/08/$25.00 ©)2008 IEEEProceedings of the 2008 IEEE, CIBEC'08

Page 3: [IEEE 2008 Cairo International Biomedical Engineering Conference (CIBEC) - Cairo, Egypt (2008.12.18-2008.12.20)] 2008 Cairo International Biomedical Engineering Conference - Operating

3

Automated cutting planes are created by projecting thepatient-specific bone models created through the biplanarreconstruction process onto our bone atlas.

Using the male and female statistical atlases, a set oflandmarks and relevant surgical axes are automatically

Figburebn calculated [3]. Figure 4 shows some of the relevant surgicalTArgets.axes used for implant placement including the transepicondylar

axis (TEA), mechanical axis and posterior condylar axis. Theseautomated landmarks are used to select the proper implant sizefor the patient and perform precise implant placement based onthe selected surgical technique as shown in figure 5.

Figure 2 - Biplanr x-rays for patient with calibration targetattached

image. This process is outlined in Figure 3. The predicted 2Dimage is matched to each of the x-ray images [1]. Highlights ofthe method include incorporating region and edge scoringmetrics into the overall matching score. The 2D projection ofthe bone model is separated into two images: the region image(binary with all pixels in the silhouette of the model white) andedge image (Prewitt edge detection on the silhouette andspreading of the resulting edge using a Gaussian smoothingfilter). A complete explanation can be found in [9].

The initial pose is used to constrain the bounds of translationand rotation during optimization. An initial optimization isperformed only on the translation and rotation of the basemodel. This allows higher constraints when varying both theshape parameters and position/orientation. This firstoptimization problem is only 6 DOF while the final bonemorphing can range anywhere from 10 to 15.When the genetic algorithm has converged to a rigid

alignment, the best member of the population is used as a Figure 4- Automatic calculated surgical axes on both femur andgenetic dopant in a second optimization step. The second tibiagenetic algorithm operates on the same objective function asthe first, but now is given additional DOF associated with shapeparameters from the statistical atlas. Depending on bone type,the shape parameters for the statistical atlas range from 5principal components (PCs) capturing 98.5% of the variationfor full femurs, to 4 PCs which capture a similar percentage ofcumulative variation for the tibia. In this second run, the geneticalgorithm still optimizes the 6 DOF for rigid alignment, but alsoadds the prominent shape parameters as more DOF. The firstoptimization allows the second to be further constrained, so thatthe search space of this high DOF problem can be reduced.Final optimization step involves fine tuning shape and pose 3i4to8 Siparameters using direction set method.

Figure 5 - Automated implant sizing and placement

Iput FinalIm e P hdi t d Te4 Pse .ItrOpatv Phs................................................................................................................................... ...............................................................................

Wireless technology and ultrasound technology are- Optimize _combined to provide intra-operative guidance to the surgeon

3 ~~~~~~~~~~~duringorthopedic surgeries. The overall system in which theFigure 3 - Optimization algorithm outline wireless or UWB (also labeled radio-frequency or RF) system

and ultrasound (US) system are components is outlined inFigure 6. It is clear from Figure 6 that these two systems are

978-1-4244-2695-9/08/$25.00 c)2008 IEEEProceedings of the 2008 IEEE, CIBEC'08

Page 4: [IEEE 2008 Cairo International Biomedical Engineering Conference (CIBEC) - Cairo, Egypt (2008.12.18-2008.12.20)] 2008 Cairo International Biomedical Engineering Conference - Operating

4

These system level constraints must be considered whiledesigning the base station and tag components of the system.

Figure 8 - Conceptual design of RF tracking system in OR withbase stations mounted to the ceiling

Figure 6 - Overall surgical navigation flowchart delineatingpre-operative and intra-operative phases

inextricably linked together and have to be developed intangent. The ultrasound system is used to noninvasivelyregister the bones (for example, the tibia and femur in a kneesurgery) while the RF tracking system is used to registerinstruments, ultrasound probes, ultrasound braces, etc. to theglobal reference frame. Figure 7 shows how the RF andultrasound systems can be integrated to track tools and bonesduring orthopedic surgeries. The instruments are tracked by theRF tracking system while the ultrasound braces track the bone3D positions and are tracked themselves by the UWB system.

Figure 9 shows the RF probe which uses UWB wirelesstechnology to pick points in the OR. The use of this proberequires the setup shown in Figure 8. Previous work in UWB isfound in [13-18] for both research and commercial systemswhile previous work done by our team can be found in [ 19-20],respectively.

Figure 9 - RF probe for point picking in OR

A GPS-like scheme is utilized along with TDOA to locate2-D and 3-D transmitting tag positions, as shown in Fig. 10.The average output power spectral density for indoor systemshas an upper bound of -41.3 dBm/MHz [21] as specified by theFCC. One typical UWB localization method is the use ofImpulse Radio (IR) UWB, where a baseband UWB pulse is

BaeStation ABase Station8

X il.l..1Figure 7 - RF and ultrasound bone tracking system

The complete RF tracking system, which uses UWBtechnology, is shown in Figure 8. As shown in Figure 8, the RFtracking system consists of four or more base stations (orreceivers) mounted to the ceiling, a control station whichgraphically displays the tracked instruments, and tags mountedto individual instruments which emit signals that are receivedby the base stations. On the system level, attention has to begiven to base station synchronization, multi-tagcommunication, tag and probe triangulation, and data display.

*GF s B Tag

Em Base Station C

Satellite C Satellite D BAse Station

(a) (b)Figure 10 - GPS system analogy: (a) GPS system, (b) UWB indoor

positioning system

978-1-4244-2695-9/08/$25.00 c)2008 IEEEProceedings of the 2008 IEEE, CIBEC'08

Page 5: [IEEE 2008 Cairo International Biomedical Engineering Conference (CIBEC) - Cairo, Egypt (2008.12.18-2008.12.20)] 2008 Cairo International Biomedical Engineering Conference - Operating

5

filtered by a UWB antenna [22]. However, with IR-UWB thereceived signals are noisy due to the complex transmittedwaveform and added multipath signals, which makes it difficultto accurately locate the position of the received LOS signal. Inour system, we modulate an UWB pulse with an 8 GHz carriersignal which resides at the upper end of the 3.1 - 10.6 GHzband. The use of this band reduces the size of the wideband RFcomponents in the transmitter and receiver and also bypassesmany of the interfering frequency bands that exist at the lowerend of the 3.1 - 10.6 GHz band.

The complete experimental setup of the developed system isshown in Fig. 11. In the developed system, we transmit amodulated narrow Gaussian pulse with a carrier frequency and

\ \d2\ \., BS2

\dn .

RRSnr

passband ofDC-5 GHz to suppress the 8 GHz carrier signal, theI/Q signals are sub-sampled using an UWB sub-sampling mixer,extending them to a larger time scale (i.e. jts range) whilemaintaining the same pulse shape. The sub-sampling mixeruses extended time techniques to achieve equivalent samplingrates in excess of 100 GS/s, which yields mm-range samplespacing and provides our peak detection algorithm with ampledata. Finally, the extended I/Q signals are processed by aconventional analog to digital converter (ADC) and standardField Programmable Gate Array (FPGA) unit.

Steps for bone registration with the ultrasound system areoutlined in Figure 13. Previous ultrasound work is found in[25-28]. The ultrasound brace is then fixed firmly aroundpatient femur and tibia. A freehand ultrasound probe is used toacquire points on both femur and tibia. Once point acquisitionis finished, the system registers the preoperative bone with theintraoperative patient position and patient leg fixation isremoved. After this initial registration and alignment step, theultrasound brace starts tracking the motion ofthe tracked pointsrelative to the initial points and the system realigns the femur ortibia in case of bone motion relative to the brace.

Figure 11 - Block diagram of indoor localization system showingone tag and three base stations which feed into the main system

controller

demodulate it at the receiver side. The source of our UWBpositioning system is a step-recovery diode (SRD) based pulsegenerator with a pulse width of 300 ps and bandwidth of greaterthan 3 GHz, shown in Fig. 12. A detailed discussion can befound in [4,23]. At the input of the traditional delay-line typepulse generator [24], a novel input matching network has beenintroduced to prevent pulse echoing, minimize pulse widthbroadening, and suppress any significant pulse distortion. Themodulated Gaussian pulse is then transmitted through anomni-directional UWB antenna. Multiple base stations arelocated at distinct positions in an indoor environment to receivethe modulated pulse signal. The received modulated Gaussianpulse at each base station first goes through a directionalVivaldi receiving antenna and then is amplified through a lownoise amplifier (LNA) and demodulated to obtain the I/Qsignals. After going through a low pass filter (LPF) with a

Em

0~

C:

E2nen

a)

e3ne:n

Time in ns

-25

-30

-35

-40 10 dB bandwidth > 3Hz-45

-50

-55

-600 1 2 3 4 5 6

Frequency in GHz

(a) (b)Figure 12 - Gaussian pulse which serves as system UWB source(a) time domain exhibiting 300 ps pulse width, (b) frequency

domain highlighting bandwidth in excess of 3 GHz

Figure 13 - Ultrasound registration steps

The brace's main function is to hold ultrasound transducersrigid attached relative to each other, and to provide good andfirm contact between transducers and patient skin. Figure 7provides a conceptual model of how these ultrasound trackingbraces will be mounted to the leg, detect the bones of interest,and then transmit the orientation data through RF circuitry tothe main controller of the surgical navigation system.

Incorporating sensors on standard surgical instrumentationduring orthopaedic surgery is not a new goal for improvingsurgical outcomes, but it remains elusive in implementationwith sufficient information gathering (sensing) and relevant,timely display to the surgeon.

Passive force in the unloaded knee is a key component toproper mechanics after Total Knee Arthroplasty (TKA). Theforce must be enough to keep the knee stable but not too muchto induce large pre-stress at the joint surface. Both tibial insertheight and ligament tightness contribute to this passive force.Collateral ligament release is a key step to obtaining a suitablepassive force across the knee and balancing the mediolateralcompartmental pressures in both flexion and extension.Unfortunately, ligament release is qualitatively judged basedon surgeon preference and experience. A breakthrough inintraoperative feedback combines cutting-edge technologiescreated by the ongoing revolution in microelectromechanicalsystems (MEMS), materials science, microelectronics, andwireless technologies. The smart provisional extends thisrevolution in sensing technology to provide intraoperativefeedback to optimize soft tissue interaction with orthopaedic

978-1-4244-2695-9/08/$25.00 ©)2008 IEEE

ri X-

Proceedings of the 2008 IEEE, CIBEC'08

TtaWhimposMm

Page 6: [IEEE 2008 Cairo International Biomedical Engineering Conference (CIBEC) - Cairo, Egypt (2008.12.18-2008.12.20)] 2008 Cairo International Biomedical Engineering Conference - Operating

6

joint replacements. The smart provisional utilizes capacitiveMEMS technology to give real-time feedback on theintercompartmental pressures within the passive intraoperativeknee.

Measuring the forces and contact areas in vivo is extremelyvaluable to implant designers, surgeons, and patients with thecommon goal of improved clinical outcomes. Measuring thesevalues as soft tissue release is performed allows for theoptimization of compartmental forces and patient-specificprediction of implant performance. Arrays of resistivemicrocantilevers and capacitive microsensors have beendesigned to measure axial strain profiles and contact areasbetween the femoral component and the tibial insert.Thousands of embedded microsensors forming anintraoperative load sensing array have been proven toaccurately read pressure distributions using ultra-sensitivecustom microelectronics.

Conceptual drawing of sensing spacer is shown in figure14-a. The fabricated microcantilever pressure sensor system isshown in Figures 14-b. A custom Application SpecificIntegrated Circuit (ASIC) is used to read signals from themicrocantilevers. A transmitter which operates at 433 MHz (auniversal band open for telemetry applications) and an antennaare also visible in Figure 14-b. Figure 15-a shows wafer withfabricated test array of pressure sensor with different sizes andspacing. Smart provisional with fabricated capacitive sensor isshown in figure 15-b. Figure 16 shows the system with readoutand the communication circuit layout.

(a) (b)Figure 14- Sensing spacer

(a) (b)Figure 15 - (a) Wafer with multiple size test arrays (b) Smart

provisional with two different sensor size and spacing

i0 -Reauout anu communication circuit layout.

C. Post-Operative PhaseFluoroscopy has emerged as a powerful tool for analyzing

implanted joints in vivo. The low power of the radiation usedcompared with conventional x-ray techniques allows data to becaptured over an extended period of time. This combined withthe freedom of motion allowed by single plane fluoroscopyallows many different weight bearing activities to be analyzed.Our group has used fluoroscopy to perform many distinct 3Danalyses including calculation of in vivo loading of the kneejoint [29], calculation of hip separation after total hiparthroplasty [30], and performance comparison of posteriorstabilized versus posterior cruciate-retaining total kneearthroplasty (TKA) [31] . Work has also been done in using thissame 3D-to-2D registration technique to analyze normal bones.This includes analysis of normal knees through fluoroscopy[32] as well as in vivo analysis of patients with and withoutanterior cruciate ligaments [33].

The robust 3D-to-2D registration technique utilized in all ofthis fluoroscopic analysis is outlined in [9]. As discussed in[34], this direct registration method outperforms similarmethods which employ an added segmentation step to extract acontour around the implant from the original fluororoscopicimage. Once the 3D information has been extracted fromfluoroscopy data, a full 3D analysis of the joint is performed.This includes calculation of hip separation, loci tracking ofdifferent joints (e.g. hip, vertebrae, shoulder)[ 10], andautomatic tracking of ligament attachments. Recently anothertool has been added, which we term "video fitting"[ 11].Instead of analyzing a small number of images for a givenactivity (e.g. capturing a fluoro image every 10° for a deep kneebend (DKB)), the activity is analyzed using a large number ofclosely spaced images. As outlined in Section 3.3, this makesfitting the entire image sequence almost completely automatedand also provides significantly more data points in analyzingthe activity.

The techniques employed in our different analytical tools forpost-operative analysis are tailored to distinct applications andtherefore largely vary from one to another. The 3D-to-2D

978-1-4244-2695-9/08/$25.00 c)2008 IEEEProceedings of the 2008 IEEE, CIBEC'08

Page 7: [IEEE 2008 Cairo International Biomedical Engineering Conference (CIBEC) - Cairo, Egypt (2008.12.18-2008.12.20)] 2008 Cairo International Biomedical Engineering Conference - Operating

7

registration method was developed previously [9,34]. A briefoverview is provided here as background information for theextension of this method to video fitting. The basic setup usedin 3D reconstruction of the fluoroscopy scene is shown inFigure 16. Details of this method can be found in [9], includingoptimization techniques used, sensitivity to manual bias etc.

Figure Reconstruction of 3Dcamera

Figure 17-3D implant registered with patient fluoroscopy

Figure 17 shows 3D femoral and tibial implants registered withpatient fluoroscopic image.Two powerful new tools have been introduced that build on ourfundamental 3D-to-2D registration method. Video fittingrepresents the next generation of 3D fluoroscopic analysis and

will serve as a useful technique for more advanced 3D analysistechniques of fluoroscopy in the future. Calculation of contactarea as shown in figure 18, combined with finite elementmethods [35], opens up new areas of analysis on both wear andthe internal stress/strain characteristics of polyethylenecomponents. Tools built on fluoroscopy provide powerfulpost-operative capabilities.

III. CONCLUSION

New technologies are constantly changing the operatingroom for orthopedic surgeons. This trend will continue andpromises to revolutionize orthopedic surgical navigation as thefuture unfolds.

In the area of pre-operative planning, techniques such asbiplanar reconstruction of 3D patient specific bone modelsholds promise in reducing pre-operative costs associated withorthopedic surgical navigation since no CT scan is needed andis replaced by two static X-ray images. This 3D patient-specificmodel can then be fed to another layer of pre-operativeplanning software to automatically find patient-specific cuttingplanes which can assist the surgeon in sizing the patient andalso in making the cuts during the orthopedic surgery.

Intra-operatively, new tracking technologies, which includewireless tracking of surgical instruments via UWB technologyand wireless bone tracking through a combination of A-modeultrasound and RF tracking, promise to bring wirelessnavigation in the OR into the 21St century by removing thewired systems currently used for this application. A wirelesspressure sensing system for use in ligament balancing willrevolutionize this stage of the orthopedic surgery by providingmuch more quantitative feedback to the surgeon in real-timewhich largely removes the subjectivity associated with howligament balancing is carried out by the majority of orthopedicsurgeons today. Finally, the automation of the cutting planesand sizing, which is done in the pre-operative step, incorporatesanother piece into the intra-operative portion of the surgery.The combination of all of these technologies promises to createa whole new and exciting experience for surgeons and patientsalike in orthopedic surgical navigation.

Finally, new advances in the post-operative phase, whichinclude reconstruction of patient-specific joint motion duringweight bearing activities via advanced computer visiontechniques applied to fluoroscopy as well as new advances inchemical sensing to help catch possible infection of the joint ata much earlier stage, promise to complete the picture in termsof revolutionizing how arthoscopic surgeries take place. Thecombination of all these technologies paints a picture of whatthe operating room of the future will look like in orthopedics.Only by carefully considering and dissecting each step of thislarge and complex process can one begin to piece together howthe addition of new technologies is possible and whatimplications these new technologies may have on the process asa whole.

IV. REFERENCES

[1] Mahfouz M. R., ElDakhakhni H. A., Abdel Fatah E. E.,Tadross R., Komistek R. D., Three-Dimensional BoneCreation And Landmarking Using Two Still X-Rays. 54th

978-1-4244-2695-9/08/$25.00 ©)2008 IEEEProceedings of the 2008 IEEE, CIBEC'08

Page 8: [IEEE 2008 Cairo International Biomedical Engineering Conference (CIBEC) - Cairo, Egypt (2008.12.18-2008.12.20)] 2008 Cairo International Biomedical Engineering Conference - Operating

8

Annual Meeting of Orthopaedic Research Socities, SanFrancisco, USA, 2008.

[2] Mahfouz M. R., Booth R. E., Argenson J. N., Merkl B. C.,Kuhn J. M., Abdel Fatah E. E., Utilization of BiplanarX-Ray Images in 3D Reconstruction of Patient-SpecificBones and Automatic Morphometric Measurements. 7thInternational Symposium on Computer Methods inBiomechanics and Biomedical Engineering, Antibes, Cotede Azur, France, 2006.

[3] Mahfouz M. R., Merkl B. C., Abdel Fatah E. E., Booth R.E., Argenson J. N., Automatic methods forcharacterization of sexual dimorphism of adult femora:distal femur. Computer Methods in Biomechanics andBiomedical Engineering, vol. 10, 2007 ,pp. 447 - 456.

[4] Mahfouz M. R., Zhang C., Kuhn M. J., Merkl B. C., FathyA., "Millimeter Accuracy Indoor 3D Positioning RadarUsing UWB Technology," Microwave Theory andTechniques IEEE Trans, 2007.

[5] Qu W., Islam S., To G., Mahfouz M. R. Design of AnalogSignal Processing Integrated Circuit for Multi-ChannelBiomedical Strain Measurement Instrument. Biodevices (1)2008: 256-259

[6] Gary To, M.S., Development of the TelemetricalIntraoperative Soft Tissue Tension Monitoring System inTotal Knee Replacement (TKR) with MEMS and ASICTechnologies. In Department of Mechanical, Aerospace,and Biomedical Engineering. Knoxville: University ofTennessee, 2007.

[7] Pritchard E., Mahfouz M. R., Evans B., Eliza S., Islam S.,.Flexible Capacitive MEMS Pressure Sensors. IEEESensors Conference, Lecce, Italy, 2008.

[8] Evans III B., PhD, MEMS Capacitive Strain SensingElements for Integrated Total Knee ArthroplastyProsthesis Monitoring. In Department of Mechanical,Aerospace, and Biomedical Engineering. Knoxville:University of Tennessee, 2007.

[9] Mahfouz M., Hoff W., Komistek R., Dennis D., 2003, Arobust method for registration of three-dimensional kneeimplant models to two-dimensional fluoroscopy images.Med. Imag., IEEE Trans., 22(12), pp. 1561-1574.

[10] Mahfouz M. R., Dennis D., Komistek R. D., et al., In VivoDetermination of Hip Separation in Subjects having EitherAlumina-on-Alumina or Alumina-on-Polyethylene TotalHip Arthroplasty. SICOT, San Diego, California, USA,August, 2002.

[11] Kuhn M. J., Merkl B. C., Mahfouz M. R., Komistek R. D.,3D-to-2D Registration of Implant Models to X-ray Imageswith Special Emphasis on Error Analysis, SegmentationEffects, and Polyethylene Wear. 18th Annual Symposiumofthe International Society for Technology in Arthroplasty,Kyoto, Japan, 2005.

[12]ElDakhakhni H., Mahfouz M. R., Abdel Fatah E. E.,Pritchard E., An Automatic Calibration Method ForBiPlanar 3D Reconstruction. 3rd Cairo InternationalBiomedical Engineering Conference, Cairo, Egypt, 2006.

[13] R. J. Fontana, "Recent system applications of short-pulseultra-wideband (UWB) technology," IEEE Trans.Microwave Theory and Tech., Volume 52, Issue 9, Sept.2004, pp.2087 -2104.

[14] Sapphire DART (RTLS) Product Data Sheet, MultispectralSolutions Inc., Germantown, MD, 2007,http://www.multispectral.com/pdf/Sapphire_DART.pdf

[15] Hardware Datasheet, Ubisense, Cambridge, UK, 2006,http://www.ubisense.net/SITE/UPLOAD/Document/TechDocs/Ubisense_hardware_datasheet_May_2006.pdf.

[16] Low Z. N., Cheong J. H., Law C. L., Ng W. T., Lee Y. J.,Pulse detection algorithm for line-of-sight (LOS) UWBranging applications. IEEE Ant. and Wireless Prop.Letters, vol. 4, 2005, pp. 63 - 67.

[17]Zetik R., Sachs J., Thoma R., UWB localization - activeand passive approach. in Proceedings of the 21st IEEEIMTC, vol. 2, 2004, pp. 1005-1009.

[18]Meier C., Terzis A., Lindenmeier S., A robust 3D highprecision radio location system. in IEEE MTT-SInternational Microwave Symposium, 2007, pp. 397 - 400.

[19] Zhang C., Kuhn M., Merkl B., Fathy A. E., and MahfouzM. R., Accurate UWB indoor localization system utilizingtime difference of arrival approach. in IEEE Radio andWireless Symposium, 2006, pp. 515-518.

[20] Zhang C., Kuhn M., Merkl B., Mahfouz M., and Fathy A.E., Development of an UWB indoor 3-D positioning radarwith millimeter accuracy. IEEE MTT-S InternationalMicrowave Symposium, 2006, pp. 106-109.

[21] Federal Communication Commission, The first report andorder regarding ultra-wideband transmission systems. FCC02-48, ET Docket No. 98-153, 2002.

[22] Low Z.N., Cheong J.H., Law C.L., Ng W.T., Lee Y.J.,Pulse detection algorithm for line-of-sight (LOS) UWBranging applications. IEEE Ant. and Wireless Prop.Letters, vol. 4, 2005, pp. 63 - 67.

[23] Zhang C., Fathy A. E., Reconfigurable pico-pulsegenerator for UWB applications. in IEEE MTT-SInternational Microwave Symposium, 2006, pp. 407 - 410.

[24] Lee J. S. and Nguyen C., Uniplanar picosecond pulsegenerator using step-recovery diode. IEE ElectronicsLetters, vol. 37, pp. 504 - 506, Apr 2001.

[25] Barratt D., Penney G., Chan C., Slomczykowski M., CarterT., Edwards P., and Hawkes D. Self-calibrating3D-ultrasound-based bone registration for minimallyinvasive orthopedic surgery. IEEE Trans Med Imag.Mar;25(3):312-23, 2006.

[26] Beasley R., Stefansic J., Herline A., Guttierez L.,Galloway Jr R., Registration of Ultrasound Images. Proc.SPIE Med Imag. Vol. 3658: 125-132, 1999.

[27] Brendel B., Winter S., Rick A., Stockheim M., and ErmertH., Registration of 3D CT and Ultrasound Datasets of theSpine Using Bone Structures. Comp Aided Surg.7:146-155, 2002.

[28] Brendel B., Winter S., Rick A., Stockheim M., and EmertH., Bone registration with 3D CT and Ultrasound Datasets.Intl Congress Series. 1256: 426-432, 2003.

[29] Komistek R. D., Kane T. R., Mahfouz M. R., Ochoa J. A.and Dennis D. A., Knee mechanics: a review of past andpresent techniques to determine in vivo loads, J. Biomech.,2005, Vol. 2, 215-28.

[30] Komistek R. D., Dennis D. A., Ochoa J. A., Haas B. D. andHammill C., In vivo comparison of hip separation aftermetal-on-metal or metal-on-polyethylene total hiparthroplasty, J. Bone Joint Surg., 2002, Vol. 10, 1836-41.

978-1-4244-2695-9/08/$25.00 ©)2008 IEEEProceedings of the 2008 IEEE, CIBEC'08

Page 9: [IEEE 2008 Cairo International Biomedical Engineering Conference (CIBEC) - Cairo, Egypt (2008.12.18-2008.12.20)] 2008 Cairo International Biomedical Engineering Conference - Operating

9

[31]Komistek R. D., Scott R. D., Dennis D. A., Yasgur D.,Anderson D. T. and Hajner M. E., In vivo comparison offemorotibial contact positions for press-fit posteriorstabilized and posterior cruciate-retaining total kneearthroplasties, J. Arthro., 2002, Vol. 2, 209-16.

[32] Komistek R. D., Dennis D. A. and Mahfouz M. R., In vivofluoroscopic analysis of the normal human knee. Clin.Orthop., 2003, Vol. 410, 69-81.

[33] Komistek R. D., Allain J., Anderson D. T., Dennis D. A.and Goutallier D., In vivo kinematics for subjects with andwithout an anterior cruciate ligament, Clin. Orthop., 2002,Vol. 404,315-25.

[34] Mahfouz M. R., Hoff W. A., Komistek R. D. and DennisD. A., Effect of segmentation errors on 3D-to-2Dregistration of implant models in X-ray images, J.Biomech., 2005, Vol. 2,229-39.

[35] Sharma, A., Komistek, R.D., Ranawat, C.S., Dennis, D.A.,Mahfouz, M.R., 2007. In vivo contact pressures in totalknee arthroplasty.Journal of Arthroplasty 22, 404-416.

978-1-4244-2695-9/08/$25.00 ©)2008 IEEEProceedings of the 2008 IEEE, CIBEC'08