CRANIOMAXILLOFACIAL DEFORMITIES/COSMETIC SURGERY

Three-Dimensional Facial Simulation

in Bilateral Sagittal Split Osteotomy:
A Validation Study of 100 Patients
Jeroen H. F. Liebregts, DDS,* Maarten Timmermans, MD, DMD,y
Martien J. J. De Koning, MD, DMD,z Stefaan J. Berge, MD, DMD, PhD,x
and Thomas J. J. Maal, PhDk
Purpose: Three-dimensional (3D) virtual planning of orthognathic surgery in combination with 3D soft
tissue simulation allows the surgeon and the patient to assess the 3D soft tissue simulation. This study was
conducted to validate the predictability of the mass tensor model soft tissue simulation algorithm
combined with cone-beam computed tomographic (CBCT) imaging for patients who underwent mandib-
ular advancement using a bilateral sagittal split osteotomy (BSSO).
Materials and Methods: One hundred patients were treated with a BSSO according to the Hunsuck
modification. The pre- and postoperative CBCT scans were matched and the mandible was segmented
and aligned. The 3D distance maps and 3D cephalometric analyses were used to calculate the differences
between the soft tissue simulation and the actual postoperative results. Other study variables were age,
gender, and amount of mandibular advancement or rotation.
Results: For the entire face, the mean absolute error was 0.9  0.3 mm, the mean absolute 90th percen-
tile was 1.9 mm, and for all 100 patients the absolute mean error was less than or equal to 2 mm. The
subarea with the least accuracy was the lower lip area, with a mean absolute error of 1.2  0.5 mm. No
correlation could be found between the error of prediction and the amount of advancement or rotation
of the mandible or age or gender of the patient.
Conclusion: Overall, the soft tissue prediction algorithm combined with CBCT imaging is an accurate
model for predicting soft tissue changes after mandibular advancement. Future studies will focus on
validating the mass tensor model soft tissue algorithm for bimaxillary surgery.
Ó 2014 American Association of Oral and Maxillofacial Surgeons
J Oral Maxillofac Surg -:1-10, 2014

The predictability of soft tissue simulation has become
one of the most important research subjects concern-
ing orthognathic surgery since surgeons and orthodon-
tists shifted their attention from occlusion-based
planning to soft tissue-based planning.1,2 The recent
introduction of 3-dimensional (3D) virtual planning of
orthognathic surgery in combination with 3D soft
tissue simulation has been a major step forward toward
real-time planning. For the surgeon, orthodontist, and
patient, 3D soft tissue simulation allows an assessment
of 3D changes in a real-time environment, allowing
quick adaptations to treatment planning when pre-
sented with an unfavorable soft tissue simulation.
Changes in the frontal view owing to surgical correc-
tions of the mandible and maxilla in the midline can
be an integrated part of the decision-making process.
Previously, no long-term validation study on soft
tissue simulation predictability using cone-beam

Received from the Department of Oral and Maxillofacial Surgery, Address correspondence and reprint requests to Dr Maal: Depart-
Radboud University Medical Centre, Nijmegen, The Netherlands; ment of Oral and Maxillofacial Surgery, PO Box 9101, Radboud Uni-
and the Facial Imaging Research Group, Nijmegen and Bruges, versity Medical Centre, 6500 HB Nijmegen, Postal Number 590,
Belgium. Netherlands; e-mail: t.maal@mka.umcn.nl
*Resident. Received January 23 2014
yResident. Accepted November 4 2014
zOMF Surgeon. Ó 2014 American Association of Oral and Maxillofacial Surgeons
xProfessor, Head Department OMF. 0278-2391/14/01711-X
kHead Department 3D Lab. http://dx.doi.org/10.1016/j.joms.2014.11.006
Drs Liebregts and Timmermans contributed equally to this study.

1
2 3D FACIAL SIMULATION PREDICTABILITY IN BSSO
computed tomographic (CBCT) imaging has been
published with a large number of patients undergoing
a single type of orthognathic operation.
The aim of this study was to assess the accuracy of
the mass tensor model algorithm used in Maxilim
2.2.2.1 (Medicim NV, Mechelen, Belgium) for the simu-
lation of surgery on CBCT images. The influence of a
patient’s age and gender and the amount of mandib-
ular advancement and rotation were independently
assessed to explain possible discrepancies.
Materials and Methods
PATIENTS
One hundred patients (35 male and 65 female) were
included in this study. All patients underwent mandib-
ular advancement using a bilateral sagittal split osteot-
omy (BSSO) according to the Hunsuck modification3
from January 2007 to December 2010 at the Depart-
ment of Oral and Maxillofacial Surgery, Radboud Uni-
versity Medical Centre (Nijmegen, The Netherlands).
This study was conducted under ethics committee
approval (study protocol 181/2005). All patients
were older than 13 years, with an average age of
31.6 years at the time of surgery (range, 13 to 68 yr).
Two weeks before surgery, a CBCT scan was acquired.
Furthermore, for all patients, a postoperative CBCT
scan was acquired at least 6 months after surgical
correction (mean, 13.4 months after surgery); thus,
skeletal relapse was not a factor.
Inclusion criteria were a nonsyndromic mandibular
hypoplasia, patients undergoing orthognathic surgery
at the authors’ center, a signed informed consent, and
the availability of an extended-height CBCT scan
before and at least 6 months after surgery.
The exclusion criteria were a chin support used
during CBCT scanning, previous orthognathic surgery,
other orthognathic procedures including chin osteot-
omy performed simultaneously, presence of orthodon-
tic appliances when the postoperative CBCT scan was
obtained, patients who were edentulous, the absence
of upper incisors or lower incisors, or extensive restor-
ative dental work after surgery.
IMAGE ACQUISITION
The extended-height CBCT scan was acquired using
the i-CAT 3D Imaging System (Imaging Sciences Inter-
national, Hatfield, PA). All patients were scanned while
seated with the head in a natural position using the
same i-CAT machine. They were asked to swallow,
relax their lips, and keep their eyes open. All patients
were scanned with a wax bite to ensure proper
condylar seating.
The acquired data from the CBCT scans were
exported in Digital Imaging and Communications in
Medicine (DICOM) format to Maxilim. In Maxilim,
the skull and skin tissue of the pre- and postoperative
scans were segmented using a threshold technique,
resulting in an accurate 3D reconstruction.
VOXEL-BASED REGISTRATION
After reconstruction, pre- and postoperative CBCT
scans were matched using voxel-based registrations
on an unaltered subvolume.4 This subvolume con-
sisted of the cranial base, forehead, and zygomatic
arches. After matching the images, the mandible was
segmented, as described by Swennen et al,5 on the
preoperative scan and the bone cuts were made for
the BSSO on the preoperative scan according to the
postoperative result.
SIMULATION OF SOFT TISSUES
After registration of the preoperative and postoper-
ative CBCT scans, the simulation of the soft tissues
could be computed. A mass tensor model soft tissue
simulation algorithm was applied to the preoperative
soft tissue surface. The preoperative mandible was
optimally aligned with the postoperative mandible,
representing exactly the transformation reached
during surgery using the surface-based registration
method described by Besl and McKay.6 Based on the
transformation of the mandible, a simulation of the
soft tissues was computed. Thus, the result of the
soft tissue simulation could be compared with the
actual postoperative soft tissue result.
VALIDATION OF SOFT TISSUE SIMULATION
Validation of the soft tissue simulation was per-
formed using cephalometric analysis (method A) and
a 3D volume measurement using distance mapping
(method B).
3D Cephalometric Analysis (Method A)
To investigate the differences between the actual
postoperative result and the soft tissue simulation, a
cephalometric soft tissue analysis was performed. The
first step was to set up a hard tissue reference frame
(Fig 1A). The pre- and postoperative 3D reconstruc-
tions were marked with cephalometric points for the
subnasale, labiale superius, stomion, labiale inferius,
sub-labiale, and soft tissue pogonion. To compute the
actual hard tissue movements, additional hard tissue
landmarks were indicated for the upper incisor land-
mark, lower incisor landmark, the mesial buccal cusps
of the first molars in the upper jaw, and the pogonion.
After the cephalometric analysis was performed, the
euclidean distances were computed for all correspond-
ing cephalometric landmarks between the preopera-
tive simulation and the actual postoperative result.
This euclidean distance represented the degree of
LIEBREGTS ET AL 3

FIGURE 1. A, Setting up a hard tissue reference frame and B-E, selecting the regions of interest (upper lip area, lower lip area, and chin area)
limited to specific regions based on specific anatomic landmarks. (Fig 1 continued on next page.)
Liebregts et al. 3D Facial Simulation Predictability in BSSO. J Oral Maxillofac Surg 2014.
4 3D FACIAL SIMULATION PREDICTABILITY IN BSSO

FIGURE 1 (cont’d).
Liebregts et al. 3D Facial Simulation Predictability in BSSO. J Oral Maxillofac Surg 2014.

deviation and provides an indication of the accuracy of lip area and the chin area. Distance maps for these
the simulation. Changes of the soft and hard tissue specific areas were used to calculate the differences
points according to the surgical movement between the simulation and the actual postoperative
were measured. result (absolute mean difference) in the upper lip
area, lower lip area, and chin area.
3D Distance Map (Method B)
In this second method, the soft tissue simulation
was validated using a distance map. This distance STATISTICAL ANALYSIS
map was calculated using the inter-surface distance Statistical data analyses were performed with SPSS
algorithm for the surface-based registration between 20.0.0.1 (IBM SPSS, Inc, Chicago, IL). Cephalometric
the soft tissue surface simulation and the actual post- analysis was used to compute the mean amount of
operative soft tissue surface. Differences can be visual- advancement, rotation, and translation of the
ized using a color scale image indicating the amount of mandible as a result of the BSSO. The mean euclidean
difference. From this color scale image, an absolute distance and standard deviation between the soft
mean difference could be calculated. tissue landmarks (subnasale, labiale superius, stomion,
To improve the validation, scattered radiation arti- labiale inferius, sub-labiale, and soft tissue pogonion)
facts were omitted. Furthermore, the soft tissue located
in the neck area and surrounding the nasal tip were
omitted, because the neck area is strongly influenced Table 1. PATIENTS’ CHARACTERISTICS
by head position and segmentation for the nose region Mean Minimum Maximum
is poor because the nasal bone is very thin. After
computing these results for the complete face, the re- Men/women 35/65
gion of interest was limited to specific regions based Age (yr) 31.6  13.4 13 68
on specific anatomic landmarks (Fig 1B-E). The mental Follow-up (mo) 13.4  4.7 6 28
foramina were indicated and represented the left and Previous SARPE, n 14
right borders parallel to the vertical plane of the refer- Brackets per 98
ence frame. These landmarks were defined as the most operation, n
anterior part of the mental foramen. Advancement of 4.2  2.1 0.3 10.2
The subnasal landmark was indicated and a horizon- pogonion (mm)
tal plane through this landmark represented the upper Rotation of 1.4  1.1 0.0 5.8
pogonion (mm)
border of the upper lip subregion. A plane through the
stomion parallel to the horizontal plane was used to Abbreviation: SARPE, surgically assisted rapid palatal expan-
separate the upper and lower lips. A plane parallel to sion.
the horizontal plane through the soft tissue point Liebregts et al. 3D Facial Simulation Predictability in BSSO. J Oral
sub-labiale indicated the separation between the lower Maxillofac Surg 2014.
LIEBREGTS ET AL 5

Table 2. DIFFERENCES FOR INDIVIDUAL LANDMARKS AT 3-DIMENSIONAL CEPHALOMETRIC ANALYSIS

Landmark Dmean (mm) Dmin (mm) Dmax (mm) D95% (mm) DIQR (mm) D+95% (mm)

Subnasale 1.1  0.5 0.0 3.2 1.1 0.7 1.3

Labiale superius 1.5  0.7 0.1 4.1 1.4 0.8 1.7
Labiale inferius 2.0  1.0 0.4 5.7 1.8 1.3 2.2
Sub-labiale 1.7  1.1 0.1 5.4 1.5 1.2 2.0
Pogonion 1.5  0.9 0.1 4.1 1.4 1.0 1.7

Abbreviations: D+95%, +95% confidence interval for difference; D95%, 95% confidence interval for difference; DIQR, difference
in interquartile range; Dmax, maximum absolute difference; Dmean, mean absolute difference; Dmin, minimum absolute difference.
Liebregts et al. 3D Facial Simulation Predictability in BSSO. J Oral Maxillofac Surg 2014.

were calculated. For the 3D volume measurements, There were more women (n = 65) enrolled in the
the mean absolute error, standard deviation, and study than men (n = 35). The mean age at surgery
90th percentile were computed using measurements was 31.6 years (range, 13 to 68 yr) and the mean
derived from the distance maps for each patient and follow-up time after surgery was 13.4 months (range,
for the sample as a whole. To compare these results 6 to 28 months). The average mandibular advance-
with previous validation studies, the percentage ratio ment and rotation were 4.2 mm (range, 0.3 to 10.2
between the number of simulations with a high degree mm) and 1.4 mm (range, 0.0 to 5.8 mm), respectively.
of accuracy (error level, #2 mm) and the total number
of scans was calculated. 3D CEPHALOMETRIC RESULTS
The Pearson product-moment correlation coeffi- Cephalometric analysis (Table 2) of the simulation
cient was calculated between the absolute mean versus the postoperative results showed mean abso-
distance and the dependent variables, namely age, lute errors of 1.1  0.5 mm at the subnasale, 1.5 
amount of advancement, amount of rotation, and 0.7 mm at the labiale superius, 2.0  1.0 mm at the
amount of translation for detecting statistical signifi- labiale inferius, 1.7  1.1 mm at the sub-labiale, and
cance at the .05 level (P # .05). Using the Student 1.5  0.9 mm at the soft tissue pogonion.
t test, the influence of gender on the simulation accu-
racy was calculated.
3D VOLUMETRIC RESULTS
Table 3 presents the 3D volumetric measurements
Results comparing the simulation with the postoperative re-
One hundred patients were enrolled in this study. sults. The mean absolute error was 0.9  0.3 mm for
No patients were omitted and no data were missing the entire face and 100% of cases had an average abso-
for analysis. A summary of the patients’ characteristics lute error level less or equal to 2 mm. Furthermore,
at surgery is presented in Table 1. 78% of patients had an average absolute error level
less or equal to 1 mm. For the limited regions of inter-
est, the mean absolute errors were 0.9  0.5, 1.2  0.5,
Table 3. ABSOLUTE DIFFERENCES FOR THE ENTIRE and 0.8  0.5 mm for the upper lip area, lower lip area,
FACE AND 3 SUBAREAS USING 3-DIMENSIONAL and chin area, respectively. Ninety-eight percent of
DISTANCE MAPS
cases had an average absolute error level less or equal
Tmean T#1 mm T#2 mm T90% to 2 mm for the upper lip area, 94% for the lower lip
Area (mm) (%) (%) (mm) area, and 97% for the chin area. The mean absolute
90th percentile values were 1.9 mm for the entire
Entire face 0.9  0.3 78 100 1.9 face, 1.8 mm for the upper lip area, 2.5 mm for the
Upper lip area 0.9  0.5 65 98 1.8 lower lip area, and 1.6 mm for the chin area (Fig 2).
Lower lip area 1.2  0.5 45 94 2.5
Chin area 0.8  0.5 72 97 1.6
CORRELATION ANALYSIS
Abbreviations: T#1 mm, percentage of simulations within No meaningful correlation was found between the
1-mm absolute minimal level of difference; T#2 mm, percent-
mean absolute error for the entire face or the limited
age of simulations within 2-mm absolute minimal level of
difference; T90%, mean absolute 90th percentile; Tmean, regions of interest and the independent variables
mean absolute difference. age, amount of advancement of the mandible, or
Liebregts et al. 3D Facial Simulation Predictability in BSSO. J Oral amount of rotation of the mandible (Table 4). Gender
Maxillofac Surg 2014. did not significantly affect the mean errors of the
6 3D FACIAL SIMULATION PREDICTABILITY IN BSSO

FIGURE 2. Error distribution for the entire face and the limited regions of interest.
Liebregts et al. 3D Facial Simulation Predictability in BSSO. J Oral Maxillofac Surg 2014.

entire face, upper lip, lower lip, and chin area (P = .73, mining soft tissue and hard tissue landmarks with high
.06, .21, and .71, respectively, by t test). accuracy because of superimposition of anatomic
structures in a 2-dimensional environment. The intro-
duction and widespread availability of CBCT scanning
Discussion made 3D cephalometric hard tissue and soft tissue
analysis possible with a high degree of accuracy and
The goal of this study was to assess the accuracy of
reproducibility based on a CT-based reference frame.8
the 3D soft tissue prediction produced by Maxillim
Recently, quantifying soft tissue profile changes
software with CBCT imaging in 100 consecutive
using 3D evaluation has become more popular.9 There
patients who underwent a BSSO. Swennen et al7 stated
are different types of soft tissue simulation algorithms
that conventional planning with lateral cephalograms
available to match the viscoelastic behavior of soft
is prone to analysis bias owing to the difficulty of deter-
tissues. These models are based on the biomechanical
properties of living tissue described by Fung.10 The
Table 4. PEARSON CORRELATIONS BETWEEN
most frequently used computing algorithms for soft
ABSOLUTE MEAN DISTANCE (ACCURACY OF SOFT tissue simulations are the mass spring model,11,12 the
TISSUE SIMULATION) AND DEPENDENT VARIABLES finite-element model,13-15 and the mass tensor
(AGE, MANDIBULAR ADVANCEMENT, AND
MANDIBULAR ROTATION)
model.11,16,17 These algorithms have been applied by
several commercial software applications for 3D soft
Entire Upper Lower tissue simulation. Until recently, a limited number of
Correlation Face Lip Lip Chin validation studies had been conducted on the
accuracy of 3D soft tissue simulation of these
Age (yr) 0.005 0.382 0.256 0.045 software packages.
Mandibular 0.184 0.094 0.271 0.381 Bianchi et al18 and Marchetti et al13 reported a
advancement (mm) reliable simulation outcome for the entire face, with
Mandibular rotation 0.016 0.012 0.098 0.035 average absolute errors of 0.94 and 0.75 mm, respec-
(mm)
tively. The 2 groups reported a high accuracy with
Note: Advancement and rotation were measured at the level an error lower than the 2-mm tolerance level (86.8%
of the pogonion. and 91%, respectively). The 2 studies used SurgiCase-
Liebregts et al. 3D Facial Simulation Predictability in BSSO. J Oral CMF software (Materialise, Leuven, Belgium) using
Maxillofac Surg 2014. CBCT scanning or multislice CT imaging with a
LIEBREGTS ET AL
FIGURE 3. Example of a good match between the postoperative soft tissue and the soft tissue simulation. 3D, 3-dimensional.
Liebregts et al. 3D Facial Simulation Predictability in BSSO. J Oral Maxillofac Surg 2014.
volumetric finite-element model for soft tissue simula-
tion according to Sarti et al.19 It used modified linear
elasticity equations and each region (bone, soft tissue,
and embedded material) received its own equation to
obtain a simulation algorithm. In the 2 studies, the
research population was small (N = 10) and consisted
of patients with mixed dentoskeletal deformities
receiving different types of surgical correction,
including genioplasty. Mollemans et al11 validated the
mass tensor model algorithm in a 10-patient series
with mixed dentoskeletal deformities using CT data
processed by Maxillim software. They reported an
average median distance measuring 0.60 mm and the
90th percentile was smaller than 1.5 mm. They
noticed that the areas with the greatest deviation
were the area of the lips and the chin area. It was sug-
gested that separating the lips from each other during
the simulation, adding sliding contacts between the
teeth and the lips, and preventing mental straining
during imaging would largely improve the simulation.
Schendel et al12 also performed a study to investigate
the accuracy of soft tissue simulation in orthognathic
patients. They used Vultus software (3dMD, Atlanta,
GA) in 23 patients (monomaxillary and bimaxillary
cases) and found an overall accuracy of 0.27 mm. In
agreement with the study by Mollemans et al,11 they
found the largest deviation in the perioral region,
which includes the dynamic parts of the face.
ACCURACY OF DATA ACQUISITION
The accuracy of data acquisition (ie, systemic error)
is negatively influenced by CBCT scanning, segmenta-
tion of tissues using a threshold technique, the voxel-
7
based and surface-based registration method, and the
reproducibility of 3D cephalometric analysis.
With CBCT scanning, an image of a patient’s bone
and soft tissue can be acquired with an accuracy of
0.28 mm compared with the gold standard, laser
surface scanning.20 CBCT imaging has advantages and
disadvantages compared with multislice CT imaging.
Advantages include a lower radiation dose, lower costs,
and the possibility of scanning while the patient is
seated. The direct availability and scanning while the
patient is seated allow for direct instructions from the
maxillofacial surgeon to ensure a correct natural head
position and relaxed lips with the proper occlusion.
The limitations of these types of studies relate mainly
to the position of the patient at the time of scanning.
A natural head position is very important, because an
excessive head tilt or flexion can distort the tissue.
Most important is that patients relax the lips to avoid
the muscle hyperfunction of grimacing during scan-
ning. A standardized scanning protocol is very impor-
tant to avoid variables. Other disadvantages include
radiation artifacts from the cone radiation beam, the
absence of comparable Hounsfield units, and different
soft tissue densities. The poor boundary conditions
for the threshold technique makes adding specific
embedded boundary conditions for skin and muscles
with only CBCT imaging computationally demanding;
this negatively influences the systemic error during
segmentation.21 In this study, it was necessary to manu-
ally remove the radiation artifacts at the edges of the
images. Also, because of the low radiation dose of the
CBCT scan, the nasal tip had to be excluded from the
simulation owing to poor boundary conditions.
8 3D FACIAL SIMULATION PREDICTABILITY IN BSSO
FIGURE 4. Example of a poor match between the postoperative soft tissue and the soft tissue simulation. 3D, 3-dimensional.
Liebregts et al. 3D Facial Simulation Predictability in BSSO. J Oral Maxillofac Surg 2014.
Other influential factors are the accuracy of the
voxel-based registration algorithm with an accuracy
of 0.5 times the largest voxel size of the largest data-
set22 (ie, 0.2 mm for the present setup) and the
surface-based matching method using the iterative
closest-point algorithm.23 Calculating the mean abso-
lute error with the iterative closest-point algorithm
leads to an underestimation of the real error because
the distance between the simulation and the iterative
closest point on the postoperative skin surface is
measured rather than the actual distance between
corresponding points.
Plooij et al8 reported an average error of reproduc-
ibility of 0.97 mm for soft tissue points in the facial
midline using 3D cephalometry. Comparable results
were found in the present study. The subnasale was
predicted with a mean absolute difference of 1.1
mm. Because the position of the subnasale is most
likely not influenced by mandibular advancement or
the removal of orthodontic appliances, this error rep-
resents the reproducibility of the 3D cephalometric
soft tissue points marked on the preoperative and
postoperative scans.
Other factors influencing the accuracy of the predic-
tion are the reproducibility of a patient’s position dur-
ing pre- and postoperative scanning (mental straining,
condylar seating, and eye position), weight gain or
loss during the evaluation period,24,25 soft tissue
swelling after surgery,26 postoperative orthodontic
changes, and the removal of orthodontic appli-
ances.27,28 To limit these influences, all postoperative
scans were obtained after the postoperative swelling
had resolved and after the removal of orthodontic
appliances to compare the postoperative scans among
individual patients.
PREDICTABILITY OF 3D SOFT TISSUE SIMULATION
If the minimum level of difference, which would be
clinically important and negate the effect of systemic
errors, is arbitrarily set at 2 mm,29 then the accuracy
of the prediction for the face as a whole and for the
subareas of the upper lip, lower lip, and chin is high
(100%, 98%, 94%, and 97%, respectively). These re-
sults are higher compared with results from previous
studies. The differences are likely caused by the rela-
tively small overall facial changes caused solely by a
bilateral sagittal split mandibular advancement
compared with facial changes as a result of bimaxillary
surgery with or without a genioplasty used in the
studies cited earlier. Also, the upright scanning
method in combination with direct instructions dur-
ing data acquisition could be contributable factors to
the more favorable results.
The mean absolute error found in the present study
for the face as a whole (0.9 mm) is comparable to that
in previous studies. This means that the predicted skin
surface for the entire face on average differs by less
than 1 mm from the actual postoperative skin surface.
The mean 90th percentile is an important factor for
evaluating the maximum absolute error for the simula-
tion. The mean absolute 90th percentile for the entire
face is 1.9 mm and this is comparable to the 1.6 mm
found by Mollemans et al.11 When calculating the
distances of every triangle corner in the entire face to
evaluate the accuracy of the simulation against the
LIEBREGTS ET AL
postoperative result, the error level is lower compared
with the calculation in a limited region of interest. This
is caused by the large amount of unchanged facial tis-
sue when comparing the face as a whole, contributing
to a lower mean difference between the (unchanged)
simulation and the postoperative result. To minimize
this effect, limited regions of interest for the upper
lip, lower lip, and chin area were set between fixed
boundaries, as described by Maal et al.30 The subarea
with the least accuracy was the lower lip area
(Figs 3, 4). The mean absolute error was 1.2 mm and
95% had an accuracy with an error lower than the 2-
mm tolerance level. The mean absolute 90th percentile
of 2.5 for the lower lip area also reflects this. To further
differentiate the differences found using the distances
between the simulation and the postoperative skin sur-
faces, a 3D cephalometric analysis was performed. For
all soft tissue points, the subnasale was predicted with
the most accuracy (mean absolute difference, 1.1 mm).
The labiale inferius was predicted with the least accu-
racy (median absolute difference, 2.0 mm). Multiple
studies have shown that the lips are the most difficult
to predict11,26,31,32 and that the simulation is most
accurate when the surgical movements are moderate
and patients have competent lips with little eversion.33
The presence of orthodontic appliances during the
preoperative scan also could have an effect on the
more ventrally simulated upper and lower lips
compared with the postoperative positions. Abed
et al9 reported no lip posture differences directly after
the removal of orthodontic appliances using conven-
tional profile photographs. The late effects of removal
of these labial appliances on the position of the upper
and lower lips are unknown. Other possible influ-
encing factors previously described are variation in
the thickness of the soft tissues,34,35 soft tissue
swelling after surgery,26,36 weight gain or loss in the
postoperative period, the presence of rigid
orthodontic appliances preoperatively, and dental
changes owing to orthodontic treatment after surgery.
There was no correlation found between the error
of simulation and age, gender, amount of mandibular
advancement, or amount of mandibular rotation for
the 3D distance maps and the 3D cephalometric anal-
ysis. This is most likely due to the fact that the overall
error of prediction is small in combination with the
various possible influencing factors.
CLINICAL RELEVANCE
The 3D virtual planning of orthognathic surgery in
combination with 3D soft tissue simulation makes
assessment of the 3D soft tissue simulation possible
for the surgeon and the patient.
Sarver et al37 found that 89% of patients thought the
simulations were realistic, 83% of patients said the
9
simulation helped them make a treatment decision,
and 72% believed that the simulation allowed them
to be an integral part of the treatment process. For
use in a clinical setting, the mean absolute 90th
percentile is an important factor to contemplate. It
represents the maximum error for 90% of measure-
ments and reflects the possible discrepancies of the
simulation. Only the lower lip area had a mean abso-
lute 90th percentile above the arbitrarily set minimal
level of difference of 2.0 mm. On visual inspection,
the largest errors were caused by patients with a differ-
ence in lip posturing of the lower lip between the pre-
operative and postoperative scans. As a result, the
lower lip was simulated more ventrally compared
with the actual postoperative result. Correct posi-
tioning of the lower lip during the preoperative scan
is the key to a predictable simulation.
The 3D soft tissue simulation predictability for facial
changes after advancement of the mandible using the
mass tensor model algorithm with CBCT imaging was
found to be high. The accuracy of the prediction
(average absolute error level, #2 mm) for the face as
a whole and for the subareas of the upper lip, lower
lip, and chin were 100%, 98%, 94%, and 97%, respec-
tively. For the face as a whole, the mean absolute
90th percentile remained smaller than 2 mm. The sub-
area with the least accuracy was the lower lip, with
mean absolute errors of 1.2 mm for the 3D distance
map analysis and 2.0 mm for the 3D cephalometric
analysis. No correlation could be found between the
error of prediction and the amount of advancement
or rotation of the mandible or age and gender of the pa-
tient. This is most likely due to different patient factors
influencing the reproducibility for the position of the
lower lip and the small mean absolute errors found
in this study. Visual inspection of the differences
between the simulations and the actual postoperative
results showed better results when the patient was
scanned with relaxed lips with little eversion. Overall,
the mass tensor model algorithm combined with
CBCT imaging is an accurate method for predicting
soft tissue changes after mandibular advancement
and can readily be used to predict postoperative tissue
changes for patients, surgeons, and orthodontists.
Future studies will focus on validating the mass tensor
model algorithm for bimaxillary surgery.
References
1. Holdaway RA: A soft-tissue cephalometric analysis and its use in
orthodontic treatment planning. Part I. Am J Orthod 84:1, 1983
2. Alves PV, Mazucheli J, Vogel CJ, et al: How the lower face soft
tissue changes after mandibular advancement or setback.
J Craniofac Surg 19:593, 2008
3. Hunsuck EE: A modified intraoral sagittal splitting technic for
correction of mandibular prognathism. J Oral Surg 26:250, 1968
4. Nada RM, Maal TJ, Breuning KH, et al: Accuracy and reproduc-
ibility of voxel based superimposition of cone beam computed
10
tomography models on the anterior cranial base and the zygo-
matic arches. PLoS One 6:e16520, 2011
5. Swennen GR, Mollemans W, Schutyser F: Three-dimensional
treatment planning of orthognathic surgery in the era of virtual
imaging. J Oral Maxillofac Surg 67:2080, 2009
6. Besl PJ, McKay ND: A method for registration of 3-D shapes. IEEE
Trans Pattern Anal Mach Intell 14:239, 1992
7. Swennen GR, Schutyser F, Barth EL, et al: A new method of 3-D
cephalometry. Part I: The anatomic Cartesian 3-D reference
system. J Craniofac Surg 17:314, 2006
8. Plooij JM, Swennen GR, Rangel FA, et al: Evaluation of reproduc-
ibility and reliability of 3D soft tissue analysis using 3D stereo-
photogrammetry. Int J Oral Maxillofac Surg 38:267, 2009
9. Abed Y, Har-Zion G, Redlich M: Lip posture following debonding
of labial appliances based on conventional profile photographs.
Angle Orthod 79:235, 2009
10. Fung YC: Biomechanics: Mechanical Properties of Living Tissues
(ed 2). New York, NY, Springer-Verlag, 1993
11. Mollemans W, Schutyser F, Nadjmi N, et al: Predicting soft tissue
deformations for a maxillofacial surgery planning system: From
computational strategies to a complete clinical validation. Med
Image Anal 11:282, 2007
12. Schendel SA, Jacobson R, Khalessi S: 3-Dimensional facial simu-
lation in orthognathic surgery: Is it accurate? J Oral Maxillofac
Surg 71:1406, 2013
13. Marchetti C, Bianchi A, Muyldermans L, et al: Validation of new
soft tissue software in orthognathic surgery planning. Int J Oral
Maxillofac Surg 40:26, 2011
14. Marchetti C, Bianchi A, Bassi M, et al: Mathematical modeling
and numerical simulation in maxillofacial virtual surgery.
J Craniofac Surg 18:826, 2007
15. Marchetti C, Bianchi A, Bassi M, et al: Mathematical modeling
and numerical simulation in maxillo-facial virtual surgery
(VISU). J Craniofac Surg 17:661, 2006
16. Shafi MI, Ayoub A, Ju X, et al: The accuracy of three-
dimensional prediction planning for the surgical correction of
facial deformities using Maxilim. Int J Oral Maxillofac Surg
42:801, 2013
17. Nadjmi N, Tehranchi A, Azami N, et al: Comparison of soft-tissue
profiles in Le Fort I osteotomy patients with Dolphin and Maxi-
lim softwares. Am J Orthod Dentofacial Orthop 144:654, 2013
18. Bianchi A, Muyldermans L, Di Martino M, et al: Facial soft tissue
esthetic predictions: Validation in craniomaxillofacial surgery
with cone beam computed tomography data. J Oral Maxillofac
Surg 68:1471, 2010
19. Sarti A, Gori R, Lamberti C: A physically based model to simulate
maxillo-facial surgery from 3D CT images. Future Gener Comput
Syst 15:217, 1999
20. Liang X, Lambrichts I, Sun Y, et al: A comparative evaluation of
cone beam computed tomography (CBCT) and multi-slice CT
(MSCT). Part II: On 3D model accuracy. Eur J Radiol 75:270, 2010
21. Loubele M, Maes F, Schutyser F, et al: Assessment of bone
segmentation quality of cone-beam CT versus multislice spiral
CT: A pilot study. Oral Surg Oral Med Oral Pathol Oral Radiol
Endod 102:225, 2006
3D FACIAL SIMULATION PREDICTABILITY IN BSSO
22. Maes F, Collignon A, Vandermeulen D, et al: Multimodality image
registration by maximization of mutual information. IEEE Trans
Med Imaging 16:187, 1997
23. Chabanas M, Payan Y, Marecaux C, et al: Comparison of
linear and non-linear soft tissue models with post-operative
CT scan in maxillofacial surgery, in Cotin S, Metaxas D
(eds): Medical Simulation. Heidelberg, Germany, Springer,
2004, pp 19–27
24. Kuvat SV, Guven E, Hocaoglu E, et al: Body fat composition and
weight changes after double-jaw osteotomy. J Craniofac Surg 21:
1516, 2010
25. Worrall SF: Changes in weight and body composition after or-
thognathic surgery and jaw fractures: A comparison of mini-
plates and intermaxillary fixation. Br J Oral Maxillofac Surg 32:
289, 1994
26. Kau CH, Cronin AJ, Richmond S: A three-dimensional evaluation
of postoperative swelling following orthognathic surgery at 6
months. Plast Reconstr Surg 119:2192, 2007
27. Chew MT: Soft and hard tissue changes after bimaxillary surgery
in Chinese Class III patients. Angle Orthod 75:959, 2005
28. Joss CU, Joss-Vassalli IM, Kiliaridis S, et al: Soft tissue profile
changes after bilateral sagittal split osteotomy for mandibular
advancement: A systematic review. J Oral Maxillofac Surg 68:
1260, 2010
29. Proffit WR, Phillips C, Turvey TA: Stability following superior re-
positioning of the maxilla by LeFort I osteotomy. Am J Orthod
Dentofacial Orthop 92:151, 1987
30. Maal TJ, de Koning MJ, Plooij JM, et al: One year postoperative
hard and soft tissue volumetric changes after a BSSO mandibular
advancement. Int J Oral Maxillofac Surg 41:1137, 2012
31. Jones RM, Khambay BS, McHugh S, et al: The validity of a
computer-assisted simulation system for orthognathic surgery
(CASSOS) for planning the surgical correction of class III skeletal
deformities: Single-jaw versus bimaxillary surgery. Int J Oral
Maxillofac Surg 36:900, 2007
32. Smith JD, Thomas PM, Proffit WR: A comparison of current
prediction imaging programs. Am J Orthod Dentofacial Orthop
125:527, 2004
33. Hernandez-Orsini R, Jacobson A, Sarver DM, et al: Short-term
and long-term soft tissue profile changes after mandibular
advancements using rigid fixation techniques. Int J Adult
Orthodon Orthognath Surg 4:209, 1989
34. Eckhardt CE, Cunningham SJ: How predictable is orthognathic
surgery? Eur J Orthod 26:303, 2004
35. Mobarak KA, Espeland L, Krogstad O, et al: Soft tissue profile
changes following mandibular advancement surgery: Predict-
ability and long-term outcome. Am J Orthod Dentofacial Orthop
119:353, 2001
36. Dolce C, Hatch JP, Van Sickels JE, et al: Five-year outcome and
predictability of soft tissue profiles when wire or rigid fixation
is used in mandibular advancement surgery. Am J Orthod Dento-
facial Orthop 124:249, 2003
37. Sarver DM, Johnston MW, Matukas VJ: Video imaging for plan-
ning and counseling in orthognathic surgery. J Oral Maxillofac
Surg 46:939, 1988