Registration of 3-Dimensional Facial Photographs for Clinical Use

g

l

b

F

b

F

R

b

F

c

J Oral Maxillofac Surg68:2391-2401, 2010

Registration of 3-Dimensional FacialPhotographs for Clinical Use

Thomas J.J. Maal, MSc,* Bram van Loon, MD,†

Joanneke M. Plooij, MD,‡ Frits Rangel, DDS,§

Anke M. Ettema, MD, PhD,¶ Wilfred A. Borstlap, MD, DMD, PhD,�and Stefaan J. Bergé, MD, DMD, PhD#

Purpose: To objectively evaluate treatment outcomes in oral and maxillofacial surgery, pre- andpost-treatment 3-dimensional (3D) photographs of the patient’s face can be registered. For clinical use,it is of great importance that this registration process is accurate (� 1 mm). The purpose of this studywas to determine the accuracy of different registration procedures.

Materials and Methods: Fifteen volunteers (7 males, 8 females; mean age, 23.6 years; range, 21 to 26 years)were invited to participate in this study. Three-dimensional photographs were captured at 3 different times:baseline (T0), after 1 minute (T1), and 3 weeks later (T2). Furthermore, a 3D photograph of the volunteer laughing(TL) was acquired to investigate the effect of facial expression. Two different registration methods were used toregister the photographs acquired at all different times: surface-based registration and reference-based registration.Within the surface-based registration, 2 different software packages (Maxilim [Medicim NV, Mechelen, Belgium]and 3dMD Patient [3dMD LLC, Atlanta, GA]) were used to register the 3D photographs acquired at the varioustimes. The surface-based registration process was repeated with the preprocessed photographs. Reference-basedregistration (Maxilim) was performed twice by 2 observers investigating the inter- and intraobserver error.

Results: The mean registration errors are small for the 3D photographs at rest (0.39 mm for T0-T1 and 0.52 mmfor T0-T2). The mean registration error increased to 1.2 mm for the registration between the 3D photographsacquired at T0 and TL. The mean registration error for the reference-based method was 1.0 mm for T0-T1, 1.1 mmfor T0-T2, and 1.5 mm for T0 and TL. The mean registration errors for the preprocessed photographs were evensmaller (0.30 mm for T0-T1, 0.42 mm for T0-T2, and 1.2 mm for T0 and TL). Furthermore, a strong correlationbetween the results of both software packages used for surface-based registration was found. The intra- andinterobserver error for the reference-based registration method was found to be 1.2 and 1.0 mm, respectively.

Conclusion: Surface-based registration is an accurate method to compare 3D photographs of the sameindividual at different times. When performing the registration procedure with the preprocessedphotographs, the registration error decreases. No significant difference could be found between bothsoftware packages that were used to perform surface-based registration.© 2010 American Association of Oral and Maxillofacial Surgeons
J Oral Maxillofac Surg 68:2391-2401, 2010
*Scientific Researcher, Department of Oral and Maxillofacial Sur-

ery, Radboud University Nijmegen Medical Centre, The Nether-

ands, and Facial Imaging Research Group, Nijmegen, Bruges.

†Resident, Department of Oral and Maxillofacial Surgery, Rad-

oud University Nijmegen Medical Centre, The Netherlands, and

acial Imaging Research Group, Nijmegen, Bruges.

‡Resident, Department of Oral and Maxillofacial Surgery, Rad-



§Orthodontist, Department of Orthodontics and Oral Biology,

adboud University Nijmegen Medical Centre, The Netherlands.

¶Resident, Department of Oral and Maxillofacial Surgery, Rad-



�Oral and Maxillofacial Surgeon, Department of Oral and Maxillofa-

ial Surgery, Radboud University Nijmegen Medical Centre, The Neth-

erlands, and Facial Imaging Research Group, Nijmegen, Bruges.

#Professor and Oral and Maxillofacial Surgeon, Department of

Oral and Maxillofacial Surgery, Radboud University Nijmegen

Medical Centre, The Netherlands, and Facial Imaging Research

Group, Nijmegen, Bruges.

Presented orally at the following meetings: Deutsche Gesellschaft

fur Mund, Kiefer und Gesichtschirurgie, June 2009, Vienna, Austria.

This work was supported by a grant from the Dutch Technology

Foundation (10315).

Address correspondence and reprint requests to Dr Maal:

Department of Oral and Maxillofacial Surgery, Radboud Univer-

sity Nijmegen Medical Center, PO Box 9101, 6500 HB Nijmegen,

Postal number 590, The Netherlands; e-mail: [email protected]

© 2010 American Association of Oral and Maxillofacial Surgeons

0278-2391/10/6810-0005$36.00/0

doi:10.1016/j.joms.2009.10.017

2391

mailto:[email protected]

Ihsthwtgsi

iemcttdnt

capgfp

cptrt

r

p

tbgs

Foi

M

2392 REGISTRATION OF FACIAL 3D PHOTOGRAPHS FOR CLINICAL USE

n the past few years, 3-dimensional (3D) technologyas evolved with high speed in oral and maxillofacialurgery. The evolution of the cone beam computedomography scanner made accurate imaging of theard tissues or bony structures of the face possibleith a relatively low radiation dose. Apart from hard-

issue imaging, 3D photo cameras (3D stereophoto-rammetry) can be used to capture the soft tissueurface of the face with correct geometry and texturenformation.

To evaluate the results of surgical interventionsn oral and maxillofacial surgery, pre- and postop-rative 3D photographs of the patient’s face can beatched.1 In medical imaging, this matching pro-

edure is referred to as registration. After registra-ion of the 3D photographs, the differences be-ween them can be visualized by a color scale oristance map. In this way, results of surgical2-7 andonsurgical treatment8-10 can be evaluated quanti-atively and objectively. Other useful applications

IGURE 1. Clinical application of registering different 3D photogf masseter hypertrophia using Botulinum toxin. The images on thnvestigate the amount of soft tissue loss.

aal et al. Registration of Facial 3D Photographs for Clinical Use. J Or

omparing different 3D photographs are the evalu-tion of different types of swelling over time (eg,hlegmon, abscess, tumor growth), cross-sectionalrowth changes,11 and establishment of databasesor normative populations.12 Various clinical exam-les are illustrated in Figure 1.Because the imaging systems are affected by

hanges in muscle tone, facial expression, and headosture, reliability and reproducibility of these sys-ems have to be validated. Most reliable studies haveeferred to linear measurements to validate their sys-ems.

For clinical use, it is of great importance that theegistration process of the pre- and postoperative 3D

hotographs is accurate (� 1 mm).13 This is impor-

ant because any changes in facial morphology coulde due to inherent errors of the technique or to actualrowth or treatment changes. With the validity of thecanning system already evaluated, the purpose of

The images on the top row illustrate the evaluation of a treatmentom row illustrate the use of mirroring and surface registration to

raphs.e bott

al Maxillofac Surg 2010.

ttd

M

fwvgMlarve

lgr

ff

gpBww

Ci

utaatttw

awrpd

wtNc

spbitf

Fafar

M

MAAL ET AL 2393

his study was to quantify the reproducibility of ob-aining 3D photographs over time and to evaluate theifferent registration procedures.

aterials and Methods

DATA ACQUISITION

In this prospective study 15 volunteers (7 males, 8emales; mean age, 23.6 years; range, 21-26 years)ere invited to participate. 3D photographs of the

olunteers were captured using a 3D stereophoto-rammetric camera setup and the software programodular System (3dMDfaceSystem; 3dMD LLC, At-

anta, GA). During acquisition, the volunteers weresked to bite in maximum intercuspidation, swallow,elax their lips, and keep their eyes open. For everyolunteer, a 3D photograph was acquired on 3 differ-nt points of time:

● T0 (the first 3D photograph)● T1 (1 minute after the first 3D photograph)● T2 (3 weeks after the first 3D photograph)

Furthermore, a 3D photograph of the volunteersaughing (TL) was acquired shortly after T0 to investi-ate whether different facial expression affects theegistration error.

Two different types of registration were per-ormed: surface-based registration and referencerame-based registration.

SURFACE-BASED REGISTRATION

Surface-based registration of 2 different 3D photo-raphs was performed using 2 different softwareackages: Maxilim v.2.2.1 (Medicim NV, Mechelen,elgium) and 3dMD Patient v3.0 (3dMDpatient Soft-are Platform; 3dMD LLC, Atlanta, GA). Both soft-are packages use an adapted version of the Iterative

IGURE 2. Illustration of the different registration procedures. Twocquired at all different time points: surface-based registration (illustrarame). Within the surface-based registration, 2 different software packcquired at the various time points. The surface-based registration pegistration (Maxilim) was performed twice by 2 observers by investi

aal et al. Registration of Facial 3D Photographs for Clinical Use. J Oral M

losest Point algorithm to perform surface-based reg-stration.1,14

First, the original 3D photographs were registeredsing both software packages. A surface-based regis-ration was performed between the 3D photographcquired at T0 and T1. Second, the 3D photographscquired at time T0 and T2 were registered. Finally,he 3D photographs of time T0 and TL were registeredo investigate the influence of facial expression. Allhese registrations result in several distance maps thatere used to analyze the accuracy of the registration.After registration of the original 3D photographs,second registration procedure was performed inhich the 3D photographs were preprocessed. By

emoving the hair and neck regions from the 3Dhotographs, obvious error regions were excludeduring registration (Fig 2).The registration procedures and preprocessingere performed on a Dell Precision M70 (Intel Pen-

ium IV 2.62-GHz processor speed, 2.0 Gbp RAM,VIDIA Quadro Fx Go 1400, 256 Mbp, graphicsard).

REFERENCE FRAME-BASED REGISTRATION

After the surface-based registration procedure, aecond, but different registration procedure waserformed. This method is called reference frame-ased registration. The 3D software program Max-

lim allows a step-by-step setup of a frame based onhe Cartesian coordinate system.15 It involves theollowing consecutive steps (Fig 3):

1. Selecting the photograph the frame would besetup on

2. Selecting the right and left exocanthion (this is apart of the vertical plane setup)

nt registration methods were used to register the 3D photographse orange frame) and reference-based registration (illustrated in cyanaxilim and 3dMD Patient) were used to register the 3D photographs

was repeated with the preprocessed photographs. Reference-basedhe inter- and intraobserver error.

differeted in thages (Mrocessgating t

axillofac Surg 2010.

tf

pcpattrfm

ssl

twttb

ttftriwv

Fdra

M se. J Or


3. Defining the line crossing the right exocanthionand superaural

4. Indicating the pupil reconstructed point land-mark. This landmark is defined as the point inthe middle of the line through the pupils. This ispart of the median plane setup.

After finalizing these steps, the software computeshe position of each plane, which forms the referencerame (Fig 3).

After setting the reference frames for all the 3Dhotographs (T0, T1, T2, and TL), the photographsould be superimposed by overlying all referencelanes. In this way the 3D photograph acquired at T0

nd T1, T0 and T2, and finally T0 and TL were regis-ered. After performing the reference-based registra-ion, several distance maps were computed and theegistration errors were analyzed. Two observers per-ormed the setup of the reference frame for this

IGURE 3. 3D photograph-based reference frame. The horizontaegrees below the cantion-superaurale line, along the horizontaleconstructed point (the center point between the left and right eyeslong the horizontal direction of the natural head position. The med

aal et al. Registration of Facial 3D Photographs for Clinical U

ethod because placement of the landmarks is ob- b

erver-dependant. In this way an intra- and interob-erver analysis of the reproducibility could be calcu-ated.

VALIDATION

To validate the accuracy of the different methods,he difference between the corresponding surfacesas calculated as a distance or error map. This dis-

ance map computes the difference (Euclidean dis-ance) between the 3D photographs on a large num-er of points (�20,000).The 50th, 90th, and 95th percentile of the regis-

ration error were computed in millimeters. Usinghese measurements, the differences between sur-ace-based registration and reference-based registra-ion could be evaluated. Within the surface-basedegistration procedure, the influence of preprocess-ng the 3D photographs could be investigated as

ell as the use of different software systems. Toisualize these differences, a graph of the mean,

of the reference frame is automatically computed as a plane 6.6n of the natural head position. The plane goes through the pupilertical plane is a plane perpendicular to the horizontal plane andne is a plane perpendicular to the horizontal and vertical planes.15


l planedirectio). The vian pla

ox plots, and correlation graphs were computed.

Tboie

R

pmrtTrm

bpm

tprvT(

trt1

ii

D

ruA

ow

gepnai

tittgE

amtclamssi

lp9mwratsv0

b

3M3MRR

M se. J Or

MAAL ET AL 2395

he intra- and interobserver error for the surface-ased registration method was evaluated in previ-us studies.1 In the current study, the inter- and

ntraobserver error for registration using the refer-nce-based method was investigated.

esults

The mean registration errors are small for the 3Dhotographs at rest at different points of time (0.39m for T0-T1 and 0.52 mm for T0-T2). The mean

egistration error increases to 1.2 mm for the registra-ion between the 3D photographs acquired at T0 and

L (Table 1). The mean registration error for theeference-based method was 1.0 mm for T0-T1, 1.1m for T0-T2, and 1.5 mm for T0 and TL.To visualize not only the mean registration error

ut also the variance in the registration error, boxlots were computed for the different time points andethods (Fig 4).Within the surface-based registration method, a dis-

inction between the original and preprocessed 3Dhotographs can be made. The mean registration er-ors for the preprocessed photographs were less ob-ious than for the original photographs (0.30 mm for

0-T1, 0.42 mm for T0-T2, and 1.2 mm for T0 and TL)Fig 5).

The registration error for the surface-based registra-ion is smaller if compared with the reference-basedegistration (Fig 6). Within the reference-based regis-ration, a mean error of 1.2 mm (interobserver) and.0 mm (intraobserver), respectively, was found.Concerning the used software for the surface reg-

stration, both software packages showed a very sim-lar registration error (Fig 7).

iscussion

In this prospective study, the accuracy of 3D ste-eophotogrammetry was evaluated in a clinical settingsing photographs of 15 individuals at different times.

Table 1. MEAN AND STANDARD DEVIATION OF THE RE

In mm

T0-T1

Mean SD

dMD (original) 0.3977 0.5228axilim (original) 0.3913 0.4871

dMD (preprocessed) 0.2855 0.3391axilim (preprocessed) 0.3291 0.3496eference-based (Obs1) 0,9844 0.7997eference-based (Obs2) 1.0069 0.7570


fter registration of the different 3D photographs s

f the same individuals, different registration errorsere analyzed to investigate the accuracy.In 1838, the first portfolio of stereoscopic photo-

raphs was created.16,17 Since that time, stereographyvolved from the crude dual camera systems of theast to the modern 3D digital photographic systems,evertheless still using the same principle: offset im-ges are merged together to create a stereoscopicmage.

With the advent of digital technology, digital pho-ography has become an increasingly important tooln facial surgery.17 Nowadays many commercial sys-ems are available on the market (eg, 3dMDface Sys-em [3dMD LLC], Di3D [Dimensional Imaging, Glas-ow, UK], 3D-Sensoren FaceSCAN [3D Shape GmbH,rlangen, Germany]).Earlier studies were performed to investigate the

ccuracy of stereophotogrammetry. These studiesostly focused on reliably measuring distances be-

ween typical anthropometric points on the 3D re-onstructed images against corresponding points onive subjects or phantom models (eg, plaster casts) as

form of validation.2,18-24 Some other studies useore complex methods to obtain and analyze 3D

hapes.25-27 Also, the accuracy of other surface acqui-ition systems, eg, laser scanning, has been evaluatedn several studies.28-35

Kau et al investigated the accuracy of capturingaser scans of the same individual at different timeoints using a commercially available Minolta Vivid00 laser scanner system.36 The results showed aean registration error of below 0.4 mm; the erroras within a range of 0.85 mm for 90% of the

egistration. Ma et al18 investigated the accuracy ofstructured light technique to capture the geome-

ry of the face and investigated the accuracy oftructured light scans by capturing the same indi-idual at different moments in time. A reliability of.2 mm was found in this study.To the best of our knowledge, no studies have

een performed to investigate the accuracy of 3D

ATION ERROR

T0-T2 T0-TL

Mean SD Mean SD

0.5304 0.6181 1.2438 1.25160.5122 0.6170 1.1818 1.24330.4212 0.4155 1.3923 1.29120.4222 0.4231 1.3399 1.32121.2690 0.9766 1.6664 1.32731.2453 0.9690 1.6271 1.2884


GISTR

tereophotogrammetry (3dMD) in a clinical setting.

Fr

M


IGURE 4. Box plots illustrating the median, 25th, and 75th percentiles of the registration error for all surface- and reference-basedegistration methods.

aal et al. Registration of Facial 3D Photographs for Clinical Use. J Oral Maxillofac Surg 2010.

Aatsai

wtmat

M se. J Or

MAAL ET AL 2397

ll 3D photographs used in the current study werecquired in a similar way. The 3D photographs ofhe patients are acquired in daily practice. Theame specialized photographer was responsible forcquiring the 3D photographs. To investigate the

FIGURE 5. Original 3D photog


nfluence of time more profoundly, 3D photographs fi

ere acquired in a way that short-term as well as long-erm varieties could be evaluated. The differences in

ean registration error between T0-T1 (0.39 mm)nd T0-T2 were diminutive, which illustrated thathe photograph acquired 1 minute (T1) after the

vs preprocessed photographs.


raphs

rst photograph (T0) could be reproduced 3 weeks

loit

lrp

Fr

M se. J Or


ater (T2) with high accuracy. The 3D photographf the patient laughing was captured to study the

nfluence of different facial expression. After regis-

IGURE 6. Illustration of the differences in the registrationegistration.


ration of the 3D photograph of the individual r

aughing with a 3D photograph of the individual inest (T0), the expected large registration error, es-ecially in the mouth region, was confirmed by the

(mm) between surface-based registration and reference-based


error

esults.

tefe

frtB

Fb

M se. J Or

MAAL ET AL 2399

The mean registration error is partially caused byhe system error of the acquisition system. Thisrror was described by Boehnen and Flynn andound to be �0.1 mm.37 Apart from the system

IGURE 7. Comparison between the 2 different software packagetween both registration algorithms.


rror and registration error, there are several other t

actors that might influence the accuracy of theegistration. One of those is the ability to capturehe face in the same facial expression every time.ecause a 3D photograph is a static picture, cap-

xilim and 3dMD Patient). The results showed a large correlation


es (Ma

ured only in 1 moment, facial expression some-

tTffmTptwaf

obstoprtseblhwcsr

ttabtbrtolml

icattAa

s3Kwr

rp

itfettertrcw

R

1

1

1

1

1

1


imes can be of great influence to reproducibility.herefore, much attention was paid to acquire the

ace in the same facial expression.26,27,38 Apartrom facial expression, positioning of the individualight also influence the accuracy of registration.o minimize this influence, the head was carefullyositioned in the natural head position every singleime.39-41 Other influencing factors could be loss ofeight of the individual, tiredness of the facial skin,

nd maybe even the monthly hormonal cycle foremales included in a study.42

When comparing the results of registering theriginal and preprocessed 3D photographs, it coulde noticed that the mean registration error wasignificantly reduced (P � .0097, 1-way ANOVAest) for the preprocessed photographs (Fig 5). Thisverall effect was expected; however, for the 3Dhotographs of the individuals laughing (T0-TL), theegistration error increased for the registration ofhe preprocessed 3D photograph. In the first in-tance, this result seemed to be unexpected. Nev-rtheless, the registration error was already estima-le in the original 3D photograph of the individual

aughing. The regions removed by preprocessingad a smaller registration error than the regions inhich obvious differences in facial expression oc-

urred. The ratio between these regions becamemaller, which finally resulted in a larger overallegistration error.

Furthermore, the analysis of the results betweenhe surface-based and reference-based method illus-rated that surface-based registrations produced moreccurate results. A significant difference betweenoth methods was found (P � .00, 1-way ANOVAest). Surface-based registration was ideal in this studyecause 3D photographs of identical individuals wereegistered. On the contrary, reference-based registra-ion is expected to give better results for registrationf different individuals. Also, other complex methods

ike Procrustes Analysis or Active Appearance Modelsight be useful for this type of registration prob-

ems.43,44

Finally, the results of both the surface-based reg-stration methods (Maxilim and 3dMD Patient)ould be compared. Both software packages use andapted version of the Iterative Closest Point regis-ration. No significant difference was found be-ween both registration algorithms (P � .86, 1-wayNOVA test), which illustrates that both algorithmsre robust and give correct results.

When comparing the results from the presenttudy with similar studies, the accuracy of thedMD Face system is equal. The reliability found byau et al33 (� 0.4 mm) and Ma et al18 (� 0.2 mm)as performed on scans without the neck and hair

egion and therefore can be compared with the1

esults of the preprocessed 3D photographs of theresent study (0.39 mm).It can be concluded that surface-based registration

s an accurate method to compare 3D photographs ofhe same individual at different time points. There-ore, 3D stereophotogrammetry is an accurate tool tovaluate facial changes (surgical or nonsurgical) overime. The results from the reference-based regis-ration method showed a larger registrationrror compared with the results of the surface-basedegistration method. When performing the registra-ion procedure with preprocessed photographs, theegistration error decreased. No significant differenceould be found between both software packages thatere used to perform surface-based registration.

eferences1. Maal TJ, Plooij JM, Rangel FA, et al: The accuracy of matching

three-dimensional photographs with skin surfaces derived fromcone-beam computed tomography. Int J Oral Maxillofac Surg37:641, 2008

2. Ayoub AF, Siebert P, Moos KF, et al: A vision-based three-dimensional capture system for maxillofacial assessment andsurgical planning. Br J Oral Maxillofac Surg 36:353, 1998

3. Ayoub AF, Wray D, Moos KF, et al: Three-dimensional model-ing for modern diagnosis and planning in maxillofacial sur-gery. Int J Adult Orthodon Orthognath Surg 11:225, 1996

4. Ji Y, Zhang F, Schwartz J: Assessment of facial tissue expansionwith three-dimensional digitizer scanning. J Craniofac Surg13:687, 2002

5. McCance AM, Moss JP, Fright WR, et al: Three-dimensionalanalysis techniques—Part 3: Color-coded system for three-di-mensional measurement of bone and ratio of soft tissue tobone: The analysis. Cleft Palate Craniofac J 34:52, 1997

6. McCance AM, Moss JP, Fright WR, et al: A three dimensionalanalysis of soft and hard tissue changes following bimaxillaryorthognathic surgery in skeletal III patients. Br J Oral MaxillofacSurg 30:305, 1992

7. Marmulla R, Hassfeld S, Luth T, et al: Laser-scan-based naviga-tion in cranio-maxillofacial surgery. J Craniomaxillofac Surg31:267, 2003

8. Ismail SF, Moss JP, Hennessy R: Three-dimensional assessmentof the effects of extraction and nonextraction orthodontictreatment on the face. Am J Orthod Dentofac Orthop 121:244,2002

9. Moss JP, Ismail SF, Hennessy RJ: Three-dimensional assessmentof treatment outcomes on the face. Orthod Craniofac Res6:126, 2003; discussion, 179 (suppl 1)

0. McDonagh S, Moss JP, Goodwin P, et al: A prospective opticalsurface scanning and cephalometric assessment of the effect offunctional appliances on the soft tissues. Eur J Orthod 23:115,2001

1. Nute SJ, Moss JP: Three-dimensional facial growth studied byoptical surface scanning. J Orthod 27:31, 2000

2. Yamada T, Mori Y, Minami K, et al: Three-dimensional analysisof facial morphology in normal Japanese children as controldata for cleft surgery. Cleft Palate Craniofac J 39:517, 2002

3. Metzger MC, Hohlweg-Majert B, Schön R, et al: Verification ofclinical precision after computer-aided reconstruction incraniomaxillofacial surgery. Oral Surg Oral Med Oral PatholOral Radiol Endod 104:e1, 2007

4. Besl PJ, McKay HD: A method for registration of 3D shapes.Patterns Anal Machine Intell IEEE Trans 14:239, 1992

5. Plooij JM, Swennen GRJ, Rangel FA, et al: Evaluation of repro-ducibility and reliability of 3D soft tissue analysis using 3Dstereophotogrammetry. Int J Oral Maxillofac Surg 38:267, 2009

6. Wheatstone G: Note on a new portable reflecting stereoscope.J Franklin Inst 26:205, 1838

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

3

3

4

4

4

4

4

MAAL ET AL 2401

7. Lee S: Three-dimensional photography and its application tofacial plastic surgery. Arch Facial Plast Surg 6:410, 2004

8. Ma L, Xu T, Lin J: Validation of a three-dimensional facialscanning system based on structured light techniques. ComputMethods Programs Biomed 94:290, 2009

9. Aldridge K, Boyadjiev SA, Capone GT, et al: Precision and errorof three-dimensional phenotypic measures acquired from3dMD photogrammetric images. Am J Med Genet A 138A:247,2005

0. Kimoto K, Garrett NR: Evaluation of a 3D digital photographicimaging system of the human face. J Oral Rehabil 34:201, 2007

1. Weinberg SM, Kolar JC: Three-dimensional surface imaging:Limitations and considerations from the anthropometric per-spective. J Craniofac Surg 165:847, 2005

2. Krimmel M, Kluba S, Dietz K, et al: [Assessment of Precisionand Accuracy of Digital Surface Photogrammetry with the dsp400 System]. Biomed Tech (Berl) 50:45, 2005

3. Krimmel M, Kluba S, Bacher M, et al: Digital surface photo-grammetry for anthropometric analysis of the cleft infant face.Cleft Palate Craniofac J 43:350, 2006

4. Weinberg SM, Naidoo S, Govier DP, et al: Anthropometricprecision and accuracy of digital three-dimensional photogram-metry: Comparing the Genex and 3dMD imaging systems withone another and with direct anthropometry. J Craniofac Surg17:477, 2006

5. Coombes AM, Moss JP, Linney AD, et al: A mathematicalmethod for the comparison of three-dimensional changes inthe facial surface. Eur J Orthod 13:95, 1991

6. Johnston DJ, Millett DT, Ayoub AF, et al: Are facial expressionsreproducible? Cleft Palate Craniofac J 40:291, 2003

7. Okada E: Three-dimensional facial simulations and measure-ments: Changes of facial contour and units associated withfacial expression. J Craniofac Surg 12:167, 2001

8. Weinberg SM, Scott NM, Neiswanger K, et al: Digital three-dimensional photogrammetry: Evaluation of anthropometricprecision and accuracy using a Genex 3D camera system. CleftPalate Craniofac J 41:507, 2004

9. Aung SC, Ngim RC, Lee ST: Evaluation of the laser scanner as asurface measuring tool and its accuracy compared with directfacial anthropometric measurements. Br J Plast Surg 48:551,1995

0. Baik H-S, Jeon J-M, Lee H-J: Facial soft-tissue analysis of Koreanadults with normal occlusion using a 3-dimensional laser scan-
ner. Am J Orthod Dentofac Orthop 131:759, 2007
1. Bush K, Antonyshyn O: Three-dimensional facial anthropome-try using a laser surface scanner: Validation of the technique.Plast Reconstr Surg 98:226, 1996

2. Coward TJ, Watson RM, Scott BJJ: Laser scanning for theidentification of repeatable landmarks of the ears and face. Br JPlast Surg 50:308, 1997

3. Kau CH, Zhurov A, Scheer R, et al: The feasibility of measuringthree-dimensional facial morphology in children. OrthodCraniofac Res 7:198, 2004

4. Kovacs L, Zimmermann A, Brockmann G, et al: Accuracy andprecision of the three-dimensional assessment of the facialsurface using a 3-D laser scanner. IEEE Trans Med Imaging25:742, 2006

5. Kusnoto B, Evans CA: Reliability of a 3D surface laser scannerfor orthodontic applications. Am J Orthod Dentofac Orthop122:342, 2002

6. Kau CH, Richmond S, Zhurov AI, et al: Reliability of measuringfacial morphology with a 3-dimensional laser scanning system.Am J Orthod Dentofac Orthop 128:424, 2005

7. Boehnen C, Flynn P: Accuracy of 3D scanning technologies ina face scanning scenario. Progress 5th Int Conf on 3-D DigitalImaging and Modeling 2007

8. Yin L, Wei X, Sun Y, et al: A 3D facial expression database forfacial behavior research. Proceedings of the 7th InternationalConference on Automatic Face and Gesture Recognition. 2006,211

9. Chiu CS, Clark RK: Reproducibility of natural head position. JDent 19:130, 1991

0. Lundstrom A, Lundstrom F, Lebret LM, et al: Natural headposition and natural head orientation: Basic considerations incephalometric analysis and research. Eur J Orthod 17:111,1995

1. Solow B, Tallgren A: Natural head position in standing subjects.Acta Odontol Scand 29:591, 1971

2. Ensign JE, Rowe J, Kowalski K: Premenstrual Syndrome: Etiol-ogy and Treatment Possibilities. Vol 47. AORN, 1988, p 962,964, 966-967, 969-971

3. Ayala-Raggi SE, Altamirano-Robles L, Cruz-Enriquez J: To-wards an illumination-based 3D active appearance model forfast face alignment. Lecture Notes Comput Sci 5197:568,2008

4. Hu Y, Zhang Z, Xu X, et al: Building large scale 3d face database
for face analysis. Lecture Notes Comput Sci 4577:343, 2007

Registration of 3-Dimensional Facial Photographs for Clinical Use

Documents

Transcript of Registration of 3-Dimensional Facial Photographs for Clinical Use