Int. J. Oral Maxillofac. Surg. 2010; 39: 75–82doi:10.1016/j.ijom.2009.10.023, available online at http://www.sciencedirect.com

Technical Note

Orthognathic Surgery

Computer-assisted three-dimensional surgical planning:3D virtual articulator: technicalnote§

S. Ghanai, R. Marmulla, J. Wiechnik, J. Muhling, B. Kotrikova: Computer-assistedthree-dimensional surgical planning: 3D virtual articulator: technical note. Int. J.Oral Maxillofac. Surg. 2010; 39: 75–82. # 2009 International Association of Oraland Maxillofacial Surgeons. Published by Elsevier Ltd. All rights reserved.

§ Grants: Deutsche Forschungsge-meinschaft (German Research Foundation)within the collaborative research center 414‘‘Information Technology in MedicalScience. Computer- and Sensor-SupportedSurgery’’.

0901-5027/01075 + 08 $36.00/0 # 2009 Inte

rnational Association of Oral and Maxillofacial Surge

S. Ghanai, R. Marmulla,J. Wiechnik, J. Muhling,B. KotrikovaUniversity Hospital Heidelberg, Department ofMaxillo- and Craniofacial Surgery, ImNeuenheimer Feld 400, D-69120 Heidelberg,Germany

Abstract. This study presents a computer-assisted planning system for dysgnathiatreatment. It describes the process of information gathering using a virtualarticulator and how the splints are constructed for orthognathic surgery. Thedeviation of the virtually planned splints is shown in six cases on the basis ofconventionally planned cases. In all cases the plaster models were prepared andscanned using a 3D laser scanner. Successive lateral and posterior-anteriorcephalometric images were used for reconstruction before surgery. By identifyingspecific points on the X-rays and marking them on the virtual models, it waspossible to enhance the 2D images to create a realistic 3D environment and toperform virtual repositioning of the jaw. A hexapod was used to transfer the virtualplanning to the real splints. Preliminary results showed that conventionalrepositioning could be replicated using the virtual articulator.

Keywords: dysgnathia; cephalometry; registra-tion; articulator; virtual reality; planning; 3D.

Accepted for publication 27 October 2009Available online 14 December 2009

In many fields, computer-assisted plan-ning has enabled surgeons to performcomplex interventions3,5. New techniques,such as navigated surgery 1,4,6 have beenestablished in craniofacial surgery.

To improve treatment and results, sys-tems such as Gaertner’s2 virtual articu-lator have been used to simulate static ordynamic occlusion. This project used

natural reference points acquired bymeans of a pointer device. The relationbetween the models is established bymeans of a registration imprint andmovement is recorded by a jaw-motion-analyzer (Zebris Medical GmbH,Isny, Germany), but the results cannot beretransferred to reality.

In another project7, plaster modelswere set up in the articulator; they wereconventional in respect to the calibrateddouble-base method – muensteranermodel operation system (KD-MMS).The models were later placed into thehexapod and transformed by an operationplanning tool to find the optimum posi-

tion. The pneumatically fixed modelswere fixed into their final position byapplying plaster. The surgeon can refer-ence the real cement models and thedisplayed transformation data in theprogram.

Most comparable projects are restrictedto soft-tissue simulation based on CT datawith the corresponding exposure dose.Even if the jaw is reposition, the mainaim of these projects is improved post-operative aesthetic appearance and func-tion of the patient’s stomathognaticsystem.

The project presented here requires onlytwo routine cephalometric images, signif-

ons. Published by Elsevier Ltd. All rights reserved.

76 Ghanai et al.

icantly minimizing the X-ray and radia-tion dose to the patient, and non-invasivefixation of the patient during exposure.Using the two images, the authors devel-oped a virtual articulator including anato-mical registration and controlledrepositioning of the jaws. The transforma-tion ends with the creation of a real-worldsplint for intra-operative use. The aim is toimprove the accuracy of orthognathic sur-gery by advancing 2D images with theaddition of the third dimension, and usingthe enhanced 3D environment in the com-puter-assisted planning system.

Material and methods

Six patients whose treatment had beenplanned conventionally were divided intotwo groups of three patients; one grouprequiring reposition of a single jaw and theother repositioning both jaws. The radi-ological images described below weretaken from all six patients, and two pairsof plaster models of the jaws were built.One model pair was used for conventionalplanning using a semi-adjustable articula-tor and the other pair for the virtual plan-ning method. The implemented virtualarticulator orientates on the selected rota-tion points of the caput mandibulae.

In the conventional method, a waximprint of the patient was created and,to achieve an exact anatomical constella-tion, a face bow (KaVo, Biberbach, Ger-many) was used.

Cephalometric X-ray

Patients were examined routinely bymeans of a digital radiological device(ORTHOPHOS XGPlus, Sirona, Ben-sheim, Germany) with a cephalometricattachment using mainly Sidex version5.54 (Sirona, Bensheim, Germany). Thisattachment has a three-point fixation unit(left/right porion and glabella) which isfixed to a rotation table at the top of thecephalometric attachment giving a vir-tual rotation axis through the patient.With the help of this unit the patientis always positioned at the same distancefrom the image-plane and in the field ofview of the X-ray. For this study, onedigital image was taken from the poster-ior-anterior and one from the lateraldirection.

Positioner and initialization object

In order to be able to transfer the virtuallyplanned splints into reality, the plastermodels have to have a well-defined posi-

tion and orientation, which has to bereproducible in the virtual environment.

The X1med3D positioner (Med3D, Hei-delberg, Germany; Fig. 1) has six armsenabling the user to set up every move-ment (including rotations, tilting, moving)with an accuracy of 1/10 mm (manufac-turer’s information). On the arms, there isa scale for setting up defined positions ofthe table. The table consists of anembedded Adesso-split (Baumann DentalGmbH, Keltern-Ellmendingen, Germany)and three linear, independent landmarksthat simplify the referencing of the centerand orientation of the table. Above thetable, there is a stamp-like attachment(Fig. 1) with an Adesso-split.

For specifying a fixed working area andto fix the orientation of the stamp, aninitializing object was created (Fig. 1).It consists of a cylinder with a height of10 cm with a fixed Adesso-split plate oneither side. The split plates are mirrored onan axis going through the center of thecylindrical object. By positioning thisobject on the table, setting all arms ofthe hexapod to the same length and push-ing the stamp onto the object, the workingarea was defined and was fixed by thedistance holder (Fig. 1). The adjustablerange is between 7.5 cm and 14 cm.

3D laser scanner

Scanning of the defined models was per-formed by a DigiScope (3D Alliance Inc.,Bischoffen, Germany), which is a laser-based 3D coordinate-measuring devicewith a measuring volume of 200� 200 � 200 to 600 � 600 � 600 mmwith 3 (maximum 6) movement axes.The digital optical sensors have a recur-rence accuracy of 8 mm.

Microscribe G2x

For measurement of the deviation betweenconventional and computer-assistedplanned splints, a MicroScribe G2X(Immersion Corporation, San Jose,USA) was used. This is a high-perfor-mance sensor for tracking the positionand orientation of the stylus tip. Its work-ing space is a sphere of 1.27 m and it hasan accuracy of 0.23 mm measured on 100points ANSI sphere (manufacturer’s infor-mation).

Procedure

The scheme (Fig. 2) shows the differentstages, such as data acquisition, prepara-tion of registration, reposition and returnof the planned splints into the real world.

Data acquisition and preparation of themodels are described above.

Planning was carried out by a surgeonand a technician. The surgeon identifiedthe points on the models and images andundertook the planning. The technicianused the same segmented points and couldrefer to the real models and the conven-tionally planned splints.

Preparation of the models

After initializing the X1med3D (Fig. 1),the mandible and the maxilla were cemen-ted to the Adesso assembly disc using thewax imprint or, in two cases, the conven-tionally planned starting splint for fixingthe relation. The stamp was fixed in thisposition, resulting in a horizontal-onlyvolume expansion while the cement dries.

Through this procedure, the authorsheld the stamp at a set distance and fixedthe models at a specific distance anddefined unambiguous orientation to theplate.

Transfer of the models into the virtualword

Before the scanning procedure, the plateof the hexapod carrying the cement modelwas mounted onto the table of the scanner.Starting with the three landmarks on theplate, each model was scanned from dif-ferent sides with a small advancement ofthe scanner and a high resolution using theScanOS (3D Alliance Inc., Bischoffen,Germany) software.

For performance reasons, the number ofpoints was reduced from 2.75 million to amaximum of 362,000. This was achievedby removing doubled points, unneededpoints and by a substantial reduction ofpoints at less important sites, such as thepalate. The software used was Rapid-form2006 (INUS Technology, Seoul,Korea). The number of points was reducedand the surface smoothened, without geo-metric loss.

Transformation of the 2D images into a 3Denvironment

In order to be able to reference radiologi-cal images to each other at the time thepatient was scanned, information on thefixed geometric metrics of the ORTHO-FLOW XGPLUS and on the images’ sizeand resolution was required. With theinformation on the size and resolution,which can be found in the Sidex files,the images can be aligned (Fig. 3). Forfinding the rotation axis of the images, theuser has to pick the porions in each of the

Computer-assisted three-dimensional surgical planning: 3D virtual articulator: technical note 77

Fig. 1. The X1med3D with the modified plate and stamp using the Adesso mounting plate while initializing the distance with the initializationobject. Also shown are the fixations for downward distance. (1) Fixation of translation in the z-direction, (2) fixation to hold the distance on the z-axis, (3) lever to adjust the distance in the z-direction, (4) fixation of the stamp, (5) stamp with Adesso-split plate, (6) the modified plate withAdesso-split plate, (7) initialization object with two Adesso-split discs, (8) mm scale and (9) cm scale of the X1med3D.

images, which can be accomplished in aseparate dialog showing the appropriateimage in full resolution. Reconstructingthe path of the X-ray, beginning at thecenter of the line connecting the left andright porions of each image and the virtualsource of the X-ray, at a distance of230 mm (manufacturer’s information)the intersection with the virtual rotationaxis can be found. By bringing thesepoints onto one point and rotating oneof the images by 908 the 3D environmentis established (Fig. 3). The scene nowshows the situation at the time the patientwas scanned, including the rotation thepatient performed before the second imagewas taken.

The 3D points of the selected porionsare the intersection points of the recon-struction of the corresponding X-ray pathsof each image.

The center of the left and right caputmandibulae (Fig. 3) are selected to con-struct the rotational axis of the virtualarticulator. In the case of planning toreposition the maxilla, the spina nasalisanterior has to be selected, building theCamper’s line for a possible transforma-tion in the posterior-anterior direction.

Another group of three points have tobe picked, including the incisal and buc-cal hunches of the teeth 16 and 26, forplacing the maxilla models into theenvironment.

The origin of the coordinate system wastransferred to the center of the line con-necting the selected 3D points of the cen-ter of the caput mandibulae. The x-axiswas rotated into this line intersecting thetwo points and creating the rotation axisfor dynamic closure of the mandible bymeans of the virtual articulator.

Registration

In order to reproduce the real initializedposition of the models in the 3D scenario,the virtual models were matched with theorigin by using the automatically selectedlandmarks on the virtual table (Fig. 4). Theupper model was rotated around the center

78 Ghanai et al.

Fig. 2. The four stages of the method and its main steps.

Computer-assisted three-dimensional surgical planning: 3D virtual articulator: technical note 79

Fig. 5. The transformation mode with four different perspectives for visual control of the applied transformation.

Fig. 3. The initialized view of the software shows the two cephalometric images and their relation in the second stage before (a) and after (b)registration corresponding with the size of the real cement models.

Fig. 4. Registration process. (1) Registered to the table of the virtual X1med3D, (2) initial registration reproduced and (3) matched into the 3Dscenario with the registered image planes.

80 Ghanai et al.

Fig. 6. The validation set up with the two plates and the magnets.

of the distance arising from the arm lengthsof the X1med3D at the initial height. Thesituation when the patient was imprintinghis teeth into the wax was reconstructed.

Fig. 7. Deviation between the conventional and

The incisal point and the buccal hunchof teeth 16 and 26 were selected in thesame order as in the images before. Themaxilla was matched with the 3D coordi-

virtually planned splints (a) entrance, (b) tempora

nates of the images, carrying the mandiblealong.

Planning the reposition

In the case of repositioning the maxilla,the origin is changed to the orientation ofthe Camper’s line to perform the transfor-mation with respect to the occlusion planeof the jaw (Fig. 5).

In the case where the transformation ofthe maxilla was finished, the mandible wasclosed by rotating a copy of the mandibleuntil it intersected with the upper jaw.Afterwards, the position of the closedmandible could be manipulated in the trans-formation mode. The calculated deviationbetween the jaws results in a coloured mapprojection to one of the jaws, representingthe intersecting points in red. This helps toassess the quality of occlusion.

Creation of the splint

To create the splint, the closed mandible isopened by a rotation equivalent to 1 mm.Applying the transformation between theinitialized and the closed position of the

ry, and (c) end splint planned by the surgeon

.

Computer-assisted three-dimensional surgical planning: 3D virtual articulator: technical note 81

Fig. 8. A patient case planned conventionally (a) preoperative and b) postoperative and virtually (c) preoperative and (d) postoperative.

mandible to a virtual model of theX1med3D gives the set up of the armlengths of the positioner.

After reinitialization and set up of thearm lengths of the real X1Med3D carryingthe models the splints were created bymeans of pressing the maxilla into themandible model covered by an ultravioletscattered-light-sensitive material.

Validation and results

The deviation of the conventionally andvirtually planned splint was measuredusing a MicroScribe G2X (ImmersionCorporation, San Jose, USA). For this,the plate of the X1med3D was fixed ontoa table. Before and after the measurementsof the splints, the landmarks on the platewere measured. To quantify the splint, itwas placed between the models of maxilla(on the upper table) and mandible (on thelower table) (Fig. 6). Measuring the bor-ders of magnets of the upper plate resultsin translation and rotation in respect of thelandmarks of the plate.

Fig. 7 shows the sum of all errors of themethod compared with the conventionalplanning method, showing the minimum,maximum, standard and absolute devia-tion. Fig. 8 gives preoperative and post-operative views of a patient case planned

conventionally. The results of the tempor-ary splints are not as satisfactory as theothers. One of the patients was not appro-priately fixed in the X-ray device andmoved, because of this, a correct relationbetween the images could not be estab-lished and the results were hampered sig-nificantly. The results of the end splintswere good. Omitting this patient, the max-imum error for temporary splints would beabout 1.8 mm.

The results show that conventionalplanning by the technician can be repro-duced as accurately as listed above. Thesurgeons’ results showed that an inexper-ienced physician can achieve good results.

Discussion

The success of orthognathic surgerydepends on accurate planning. In dys-gnathia, the planning of the reposition ofthe jaw has to be precise.

Gaertner and Kordass2 is one of the fewstudies dealing with repositioning of thejaws, with the aim of evaluating the staticand dynamic occlusion of the jaws basedon the real movement of the lower jaw,which was recorded for reproduction.Their study showed that executing a func-tional and occlusion analysis was possible.Visualization of movement and occlusion

was realized in a 3D environment, inwhich the user can zoom into a regionof interest. This system does not allowrepositioning the jaws and retransformingthe information gained into the real world.The benefit of the system is limited to thequality of visualization.

Vollmer et al.7 presented a system usinga hexapod for repositioning the jaws. Afterinitialization of the jaws in a conventionalarticulator, they were fixed to the upper ringof the hexapod. Through movement of thering, the physician can determine the bestposition of the jaw by evaluating the valuesprinted on the screen and the occlusion ofthe plastered models. The software alsogives information on whether the positionis within the defined limits. Once an opti-mized position is found, the jaw models arefixed by plaster. When setting up the mod-els within the hexapod, the view of themodels becomes restricted through the legsof the device. Visualization of the jaws islacking in this project.

In the present work, a closed method forcomputer-assisted planning is presented.As in the work of Vollmer et al.7, a high-precision hexapod was use, and like Gaert-ner2, 3D visualization was integrated intothe system. This visualization techniquewas most useful when planning the repo-sitioning of the jaws and evaluating occlu-

82 Ghanai et al.

sion. Initially, information is taken fromthe patient, virtual planning follows, andfinally the individualized splints for therepositioning of the patient’s jaw are pro-duced. The patient receives routine expo-sure only, and no face bow is required.Documentary requirements were fulfilledthrough the digital storage of all data.

The results show that the system canreproduce the conventionally plannedsplints. In the case where the patientwas moving in the X-ray device, theend-splint was still good. This leads tothe assumption that such errors would onlyaffect the temporary splint, which needs tobe evaluated. Other causes of defectscould be the difficulty of marking thepoints on the images, especially the selec-tion of the center of the caput mandibulae,and the patient having chondrifications.Another cause of defects in the measure-ments is that, due to documentation, dif-ferent plaster models were used forcreating the splints in the conventionaland virtual planning method. This resultsin the conventionally planned splints hav-ing a slide error fitting to the second pair ofplaster models.

The main benefit of the system is theomission of the face bow and the conven-tional articulations, the possibility of easyand fast repositioning of the jaw without

using any materials and minimizing theuse of X-rays during repositioning.

In the future, selecting points on the X-ray scans will be improved by using mar-kers fixed to the patient during the scan-ning procedure. This should reduce theerror due to selection of bad points andshould improve the accuracy of registeringthe models. The error in planning thetemporary splints would be minimized.

Acknowledgements. This work was fundedby the Deutsche Forschungsgemeinschaft(German Research Foundation) within thecollaborative research center 414 ‘‘Infor-mation Technology in Medical Science.Computer- and Sensor-Supported Sur-gery’’. It also was supported by the 3D-Alliance GmbH, who made the laser scan-ner available, and by INUS Technology.

This paper is dedicated to the memoryof Prof. Joachim Muehling.

References

1. Ewers R, Schicho K, Undt G,Wanschitz F, Truppe M, Seemann R,Wagner A. Basic research and 12 years ofclinical experience in computer-assistednavigation technology. International Jour-nal of Oral and Maxillofacial Surgery2005: 34: 1–8.

2. Gaertner C, Kordass B. The virtualarticulator: development and evaluation.International Journal of ComputerizedDentistry 2003: 6: 11–24.

3. Hassfeld S, Muhling J. Computerassisted oral and maxillofacial surgery –a review and an assessment of technology.International Journal of Oral and Maxillo-facial Surgery 2001: 30: 2–13.

4. Korb W, Marmulla R, Raczkowsky J,Muhling J, Hassfeld S. Robots in theoperating theatre – chances and challenges.International Journal of Oral and Maxillo-facial Surgery 2004: 33: 721–732.

5. Marmulla R, Hoppe H, Muhling J,Eggers G. An augmented reality systemfor image-guided surgery. InternationalJournal of Oral and Maxillofacial Surgery2005: 34: 594–596.

6. Meehan M, Teschner M, Girod S.Three-dimensional simulation and predic-tion of craniofacial surgery. Orthodontics& craniofacial research 2003: 6: 102–107.

7. Vollmer D, Ehmer U, Bourauel C,Linss G. Planning orthodontic surgerywith the Hexapod system. Biomedizi-nische Technik 2001: 46: 63–68.

Address.Sassan GhanaiBergheimer Str. 101a69115 Heidelberg / GermanyTel.: +49 6221 7963304fax: +49 6221 7963305.E-mail: sassan.ghanai@web.de