Design of a haptic simulator for osteosynthesis screw insertions

Design of a haptic simulator for osteosynthesis screw insertions
of 4
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  Design of a Haptic Simulator for Osteosynthesis Screw Insertion Ann Majewicz ∗ Jason Glasser Rosemary Bauer Stephen M. Belkoff † Simon C. Mears Allison M. Okamura A BSTRACT Osteosynthesis procedures require the insertion of self-tappingorthopaedic screws through bone plates and into one or two layersof cortical bone, in order stabilize the fractured bones. Applyingtoo much torque can cause the screw to strip and lose purchase,reducing the effective hold of the screw against the plate. Expe-rienced surgeons have developed a “feel” for the torque-rotationrelationship that reduces the risk of stripping. A haptic simulatorfor orthopaedic screw insertion procedures is desirable for trainingsurgeons to learn this relationship in a virtual reality practice sce-nario. A one-degree-of-freedom friction-based haptic device wasdesigned and used to present bone screw insertion scenarios to ex-pert surgeon users. Evaluation of the system indicated that realismmay be improved through vertical motion of the virtual screw, anddevice calibration based on measurements in human bone. Index Terms: H.5.2 [Information Interfaces and Presentation]:User Interfaces—Haptics I/O, I.6.5 [Computing Methodologies]:Simulation and Modeling—Model Development 1 I NTRODUCTION Osteosynthesis is the reduction and stabilization of fracturedbones using mechanical connectors, commonly plates that are heldby bone screws. Successful insertion of a bone screw requires sig-nificant training, as the proper technique varies across patient pop-ulations, the type and condition of the bone, and the type of thescrew. Traditionally, trainees practice orthopaedic screw insertionfor osteosynthesis on cadaver bones or actual patients under thesupervision of an experienced surgeon. Synthetic bone analoguesmay also be used, but they are costly, can only simulate a specificbone density, and must be replaced after a few uses. A simulatorcould be less costly in the long term, allowing surgeons to use thesame tool to practice insertion on multiple bone types. Several sys-tems have been developed for computer-assisted bone screw inser-tion (e.g. [1, 2]), simulation of bone drilling (e.g. [3, 4]), and sim-ulation of bone screw insertion without haptic feedback (e.g. [5]),but to the authors’ knowledge, there is no previous work on hapticsimulation of the osteosynthesis bone screw insertion procedure.There are three major phases of screw insertion (Figure 1): inser-tion, tightening and stripping. During the insertion phase, cuttingand friction forces resist the user’s applied torque. In the tighten-ing phase, screw loading causes an increase in the resisting torque. ∗ A. Majewicz, J. Glasser, R. Bauer, and A. M. Okamura are with theDepartment of Mechanical Engineering and Laboratory for ComputationalSensing and Robotics (LCSR), Johns Hopkins University, 3400 N. CharlesSt., Baltimore, MD USA. † S. M. Belkoff and S. C. Mears are with the Department of OrthopaedicSurgery, Johns Hopkins Bayview Medical Center, 4940 Eastern AvenueBaltimore, MD 21224 USA.   InsertionTightening Stripping Figure 1: Three possible phases of orthopaedic screw insertion. Dur-ing the insertion phase, cutting and friction forces cause some resis-tance torque. Loading the screw during the tightening phase createsmuch larger resistance torques. During and after thread damagein the stripping phase, the magnitude of the resistance torques de-creases dramatically. Finally, in the stripping phase, the bone threads are stripped whenthe damage threshold of the material is exceeded. This leads to asharp decrease in the resistance torque, as the material causing theresistance is removed with each additional turn of the screw.During bone screw insertion, it is difficult determine how muchtorque should be applied to reach an optimal tightness withoutstripping the bone. Experienced surgeons develop an intuition forproper insertion torque over time; however, it is very difficult toteach to novice surgeons. To understand the torque at which strip-ping will occur, anovice must tighten the screwuntil thread damageoccurs, which is infeasible in live patient training. Therefore, it ischallenging to train a surgeon, or evaluate his proficiency, with con-temporary methods.The success of an osteosynthesis procedure depends on the sur-geon tightening the orthopaedic screw until an optimal torque isreached. This is commonly referred to as getting good purchasewith the screw. Achieving the maximum fixation force, the com-pressive force generated between the plate and bone by the screw,without damaging the bone threads results in maximum efficiencyfor the bone screw [6]. Damage to the bone thread can reduce thepullout strength of the screw by forty percent or more [7]. This inturn decreases the fixation force, reducing the effectiveness of theprocedure. Even in healthy bone, identifying the optimal torquecan be a challenging task for the surgeon. But in osteoporotic bone,screw tightening is especially difficult. This is in part because thetorque necessary to damage the bone threads, the stripping torque,is roughly proportional to the radiological density of the bone [6].Osteoporosis and other diseases can significantly reduce the radio-logical density of bone and lower the stripping torque.In this paper, we explain the properties of bone screw insertionthat led to the design of a new haptic device, the implementation of the haptic simulator, and its performance. 2 M OTIVATION FOR H APTIC D EVICE D ESIGN We postulate that experienced surgeons use the relationship be-tween angular displacement and torque applied to the screw as theprimary mechanism for accurately perceiving their proximity to the 497 IEEE Haptics Symposium 201025 - 26 March, Waltham, Massachusetts, USA978-1-4244-6820-1/10/$26.00 ©2010 IEEE  012345678900.511.5 User position (revolutions)    T  o  r  q  u  e   (   N    −   m   ) Normal Bone, Single Cortical Wall 05101520253000.511.522.53 User position (revolutions)    T  o  r  q  u  e   (   N    −   m   ) Normal Bone, Double Cortical Wall 05101500. User position (revolutions)    T  o  r  q  u  e   (   N    −   m   ) Osteoporotic Bone, Single Cortical Wall Figure 2: Torque profiles acquired during orthopaedic screw inser-tion for three cases: single cortical wall (normal bone), double cor-tical wall (normal bone), and single cortical wall (osteoporotic bonesimulation foam). Reproduced from [8]. stripping torque [6]. Thus, a simulator that takes input rotationsfrom the user and outputs the appropriate resistance torque shouldbe sufficient for training purposes.Thomas et al. [8] measured the torque profiles for orthopaedicscrew insertion in several types of bone. They discovered that thetorque profiles for orthopaedic screw insertion consist of three ma- jor phases. The first phase, insertion, is a gradual increase in resis-tance torque over several rotations caused by the growing frictionbetween the bone and the screw. The second phase, tightening, isa relatively rapid ramping up of torque over less than one rotation,as the head of the screw comes into contact with the bone plateand the threads of the screw are forced against the newly formedthreads in the bone. The third phase, stripping, is a rapid decreasein resistance torque after the screw threads tear through the bonethreads. The general shape of these profiles is independent of bonehealth or thickness. Figure 2 shows three torque profiles measuredby Thomas et al. [8].The primary source of torque variation during bone screw in-sertion is friction, which is challenging to simulate (especially forlarge torques) with conventional impedance-controlled haptic de-vices. The typical alternative, an admittance-controlled haptic de-vice, requires expensive force (or torque) sensing. Thus, we pro-pose a device that uses friction to apply the resistive torque. In ourdesign, the user turns a shaft that is acted on by a mechanical brakewith a stiffness and a coefficient of friction (Figure 3). The com-pression of this brake is regulated by a linear actuator. This allowsfor a straightforward relationship between the number of steps thelinear actuator moves and the torque required to rotate the shaft.Taking the number of rotations applied by the user as the input andthe motor position as the output, a simple model can be determinedfrom the physical properties of the system.Let x equal the motor position and k  equal the spring constant. F  = kx , (1)where F  equals the force applied normal to the shaft. The geometryimplies: τ  resmax = µ  Fr  (2)where τ  resmax is the maximum resistance torque, µ  is the coefficientof friction, and r  is the radius of the shaft at the point of contact. x PLUNGERSPRINGLINEARACTUATORDRUM R IJ USER   IJ RES   F fr  F Figure 3: Concept of a friction-based haptic device to display torque. Substituting(1)into(2), thereisadirectrelationshipbetweenmotorposition and resistive torque: τ  resmax = µ  kxr  . (3)Let τ  user  be the input torque from the user and τ  res be the torque feltby the user. τ  res is then determined by: τ  res = { τ  resmax τ  user  ≥ τ  resmax τ  user  τ  user  < τ  resmax (4)The physical properties of the system handle the conditional na-ture of the output by using only frictional forces to resist the user’sinput torque. Thomas et al. [8] provided data for constant velocityinsertion of an orthopaedic screw. However, inserting a screw byhand is done in discrete steps of nearly constant velocity, with pe-riods of zero velocity when the joint limit of the wrist is reached.Cordey et al. [6] determined that the discretization of the insertiondoes not significantly affect the torque-rotation curve. 3 S YSTEM D ESIGN In order to recreate the torques that occur during bone screw in-sertion, we designed a friction drive system to apply appropriateresistive torques based on rotation. To achieve this, we attached atypical orthopaedic screw to an aluminum shaft as shown in Figure4a. A standard screwdriver from a small fragment fracture repairset is used to turn the orthopaedic screw (Figure 4b). This cor-responds to the tools used to acquire the data in [8]. A 0.8 mmthick and 1 cm wide piece of silicone rubber, with a coefficient of friction of 1.2, is wrapped around the shaft in order to generate ap-propriate friction forces. The shaft, which has a maximum radiusof 4.45 cm, is mounted on an acrylic frame with sleeve and thrustbearings to reduce the unwanted friction between the shaft and theacrylic. Angular position data is obtained from an encoder attachedto the shaft.The friction on the shaft was modulated by a plunger with sil-icone rubber on the end. The plunger is constrained by rollers toone degree of freedom of motion, perpendicular to the user’s inputshaft. The plunger is pressed against the shaft by a stiff compres-sion spring (272 N/cm). The compressed length of the spring ismodulated by a linear actuator. The linear actuator is a stepper mo-tor with a 0.01 cm step size to enable small increments in torque.The linear actuator is grounded relative to the user input shaft.The control system of the device is presented in the block dia-gram in Figure 5. The device reads user rotations and outputs atorque appropriate for insertion into a particular bone type. Thethree types of bone simulations possible with our current systemare: (1) normal single cortical wall bone, (2) normal double corti-cal wall bone, and (3) osteoporotic single cortical wall bone.The user is able to select the desired bone simulation. Thecomputer accesses the appropriate torque versus rotation data file 498  Simulation SelectionLinear ActuatorPowerFriction ShaftScrewPlungerSafety/Calibration Limit SwitchInterface to DAQ andMotor Controller (a) Haptic device (b) User interaction Figure 4: Apparatus for orthopaedic screw insertion simulation. (a)The haptic device and controller. (b) A user interacts with the devicethrough a orthopaedic screwdriver. SimulationSelectionLinearActuatorMotorControllerDAQ InEncoderQuadBoardComputerDAQ OutLimitSwitch USER POSITIONINPUTDESIRED TORQUEOUTPUT Figure 5: Control diagram for the simulator. and, after initializing the device, measures the user rotation withthe shaft encoder. The program selects the resistance torque cor-responding to the rotation. The software determines how manysteps the actuator needs to take to display the desired torque to theuser and then sends the appropriate commands to the stepper motordriver. The stepper motor remains stationary when the end of thedata file is reached or if a safety limit switch is activated. Informa-tion such as bone simulation type, desired torque display, numberof rotations, limit switch state, and whether the stripping torque hasbeen reached can be displayed via a graphical user interface.There are three major electrical components in the system: alinear actuator with a stepper motor driver, two simulation selec-tion switches, and a pushbutton limit switch. The stepper motordriver is a logic-controlled driver that uses a quad Darlington ar-ray to provide enough current to energize the motor windings. AC++ program controls the direction and the distance traveled byfull stepping the motor windings.The user selects the desired simulation through a set of toggleswitches, which are inputs to the DAQ. The program selects theappropriate files for reading based on the switches. The pushbut-ton limit switch serves as an emergency stop in case the actuatorretracts too far as well as a recalibration mechanism. Before eachinsertion simulation, the device calibrates itself by retracting themotor until the limit switch is triggered. The motor then extends apre-calibrated distance to place the end of the spring block in con-tact with the friction wheel. This ensures that the proper force willbe applied to the friction wheel. 4 E XPERIMENTS Two sets of experiments were used to obtain both quantitativeand qualitative data regarding the performance of the system. Thequantitative data was found with a torque sensor and the qualita-tive data was obtained from subjective feedback from experiencedorthopaedic surgeons. 4.1 Torque Data An ATI MINI-45 force/torque sensor attached to the user inputshaft measured the torques applied to a user during the simula-tion. Only the torque data from three rotations before and duringthe stripping phase was obtained because too many rotations woulddamage the force sensor cable. By measuring the torque data, wefound that a scaling factor between the motor steps and the out-put torque was needed to calibrate the device to the expected peak torques. Initially, there was a constant scaling factor between bothnormal and osteoporotic bone, however, over time the necessaryscaling factor changed and was different for normal bone (a scal-ing factor of 6 to 7) and osteoporotic bone (a scaling factor of 5).This is likely due to a discrepancy between the actual coefficientof friction for the silicone rubber and that stated by the manufac-turer, wear in the rubber over time, and small changes in the initialplunger position due to friction in the spring block.The torque and rotation data were acquired after calibrating thedevice to the expected peak torques. Only single cortical wall bonewas tested. The adhesion strength of attachment between the sili-conerubberandtheshaftwaslessthanthemaximumtorqueappliedfor double cortical wall bone, so that test case was eliminated. 4.2 Surgeon Data We performed a human subject experiment in which five expe-rienced orthopaedic surgeons (ages 36-58) used the simulator. Theprotocol was approved by the Johns Hopkins Institutional ReviewBoard, and subjects gave informed consent. Only one subject hadprevious experience with a haptic virtual environment. As with thetorque experiment, only three rotations prior to and during the strip-ping phase were tested. The surgeons’ time constraints preventeda longer experiment. The surgeons were asked to imagine that thescrew had been inserted most of the way, and the screw would stripin a few rotations. The surgeons were asked to rotate the screw-driver mounted on the user input shaft and stop when they felt likethey were about the strip the screw. They were then told if they hadstripped the screw or not, and showed their location on the torque-rotation profile from [8]. They were asked to continue rotating thescrewdriver to strip the screw and focus on the sensations felt. Thesurgeons filled out a short survey to rate the effectiveness of the de-vice. A section for comments and suggestions was included, anda short discussion followed the survey to identify potential mecha-nisms to add realism to the simulation. 5 R ESULTS AND D ISCUSSION 5.1 Torque Data The torque-rotation profiles comparing our haptic simulatortorque-rotation profiles to those of real bone for both normal andosteoporotic bone are shown in Figure 6.A significant difference between the simulated and real bonetorque-rotation profiles is the noise in the measurements from thesimulator. This is an artifact of the data acquisition method. Thesimulator data is acquired during human application of torques tothe system, and when the user stops momentarily or varies the turn-ing speed, dips in the torque data occur. These dips are not presentin the real bone data, which was acquired from an automated screw-turning device. Another difference, noticeable in the normal bone 499  0123456789024681012Userositionrevolutions)    T  o  r  q  u  e   (   N    −   m   ) Haptic simulatorReal bone (a) Normal bone 051015012345678Userposition(revolutions)    T  o  r  q  u  e   (   N    −   m   ) Haptic simulatorBone phantom (b) Osteoporotic bone Figure 6: Comparison of measured torque profiles of the haptic sim-ulator and real bone [8] for (a) osteoporotic bone, single cortical walland (b) normal bone, single cortical wall. simulation data, is the limited ability of the simulator to generatesufficient torque to reach the peak seen in the normal bone realdata. In testing, we found that the peak torque for our simulatoris approximately 0.92 N-m, and the peak torque from real normalbone is approximately 1.3 N-m. This mismatch can be explainedby a difference between the expected and actual coefficients of fric-tion of the rubber as well as possible changes in the coefficient of friction due to changes in user velocity during rotation. 5.2 Surgeon Data All five surgeon subjects commented on their experience withthe system and provided suggestions for improvement, and fourprovided a numerical rating of the device on a scale of 1 to 5, inwhich 1 indicated that the simulation felt nothing like real bone and5 indicated that the simulator was indistinguishable from real bone.The average rating for the normal bone simulation was 2.75 and therating for the osteoporotic bone simulation was 3.0. The subjectsstated that both simulations were similar to real bone; however, afew major sensations were missing.All subjects said that the torques were considerably higher thanexpected. Visual feedback, particularly of the distance between thescrewheadandtheboneplate, isacriticalindicatorforwhentostopturning the screw. This may have increased perceived stiffness, asthe screw’s lack of vertical translation may have caused a haptic il-lusion similar to that described in [9]. Additionally, the “real” bonedata used in these experiments were from a sheep tibia and a phan-tom bone [8]. It is possible that both the sheep tibia and phantombone are harder than human bone, but currently there is no dataavailable to test this hypothesis. Data should be collected duringscrew insertion into human bone for a more realistic simulation.Other potential causes of discrepancy between expected and ac-tual torque include sound and sudden onset of torque. The exper-iment did not mask the noise from the stepper motor, which couldhave increased the perceived stiffness of the screw [10]. Addition-ally, the surgeons were only tested during the final portion of thetightening phase and the stripping phase. The suddenness of theonset of the tightening phase could have made the simulation feelstiffer than expected. Including the insertion phase may improveperception of the torque-rotation profile. 6 C ONCLUSION AND F UTURE W ORK We have designed and evaluated a one-degree-of-freedom hap-tic simulator for orthopaedic screw insertion during osteosynthe-sis procedures. We measured the performance of the device bothquantitatively, with torque data, and quantitatively, with feedback from experienced surgeons. Both experiments indicate that ourdevice provides a reasonably realistic bone screw insertion expe-rience, when compared with real bone data and personal experi-ences. However, improvements are needed in both the device andexperiment design. The relationship between angular displacementand applied torque is not the only important feedback mechanismfor accurately perceiving proximity to the stripping torque; verticalmotion of the screw head is also important. This may be achievedthrough a graphical display or the addition of a translational degreeof freedom. The materials of the device should also be modified toallow for the display of higher torques. In addition, the experimentshould involve more rotations of the screw driver and the simulatorshould be calibrated and tested against human bone data. We alsoaim to develop a general model to relate screw and bone parame-ters to the characteristic torque versus angular displacement curve.This would allow trainees and instructors to simulate screw inser-tions from the full spectrum of osteosynthesis procedures. A CKNOWLEDGEMENTS The authors wish to thank Dr. Ralph Thomas and Ali Uneri.This work was supported in part by Johns Hopkins University anda National Science Foundation Fellowship. R EFERENCES [1] D. Gygax, et al. , “Computer-assisted surgery for screw insertion intothe distal sesamoid bone in horses: An in vitro study,” VeterinarySurgery , vol. 35, no. 7, pp. 626–633, 2006.[2] P. Peters, “Computer assisted screw insertion into real 3d rapid pro-totyping pelvis models,” Clinical Biomechanics , vol. 17, no. 5, pp.376–382, 2002.[3] D. Morris, et al. , “A collaborative virtual environment for the simu-lation of temporal bone surgery,” in Medical Image Computing and Computer-Assisted Intervention . Springer, 2004, pp. 319–327.[4] ——, “An interactive simulation environment for craniofacial surgi-cal procedures,” in Medicine Meets Virtual Reality, Studies in HealthTechnology and Informatics . IOS Press, 2005, pp. 334–341.[5] R. Rush, et al. , “Beyond the operating room: A simulator for sacroil-iac screw insertion,” Surgical Innovation , vol. 15, no. 4, pp. 321–323,2008.[6] J. Cordey, et al. , “Human torque control in the use of bone screws,”in Uhthoff HK, ed. Current Concepts of Internal Fixation Fractures .Springer-Verlag, 1980, pp. 235–43.[7] K. Lawson et al. , “Effect of insertion torque on bone screw pulloutstrength,” Orthopedics , vol. 24, no. 5, pp. 451–454, jun 2001.[8] R. Thomas, et al. , “Automated surgical screwdriver: Automated screwplacement,” Proceedings of the Institution of Mechanical Engineers ,vol. 222, no. 5, pp. 451–454, jun 2008.[9] M. A. Srinivasan, et al. , “The impact of visual information on hapticperception of stiffness in virtual environments,” in Proc ASME Dy-namics Systems and Control Division , 1996, pp. 555–559.[10] F. J. Canadas-Quesada et al. , “Improvement of perceived stiffness us-ing auditory stimuli in haptic virtual reality,” in Proc. IEEE Mediter-ranean Electrotechnical Conference MELECON 2006  , 16–19 May2006, pp. 462–465. 500
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks