journal of prosthodontic research 63 (2019) 313–320
Journal of Prosthodontic Research
Original article
Simplifying the digital workflow of facial prostheses manufacturing
using a three-dimensional (3D) database: setup, development, and
aspects of virtual data validation for reproduction
a,b, c a c
Alexey Unkovskiy *, Ariadne Roehler , Fabian Huettig , Juergen Geis-Gerstorfer ,
d e c
Joern Brom , Constanze Keutel , Sebastian Spintzyk
a
Department of Prosthodontics at the Centre of Dentistry, Oral Medicine, and Maxillofacial Surgery with Dental School, Tuebingen University Hospital,
Tuebingen, Germany
b
Department of Dental Surgery, Sechenov First Moscow State Medical University, Moscow, Russia
c
Section Medical Materials and Science, Tuebingen University Hospital, Tuebingen, Germany
d
Brom Epithetik, Heidelberg, Germany
e
Department of Oral and Maxillofacial Surgery, and Head of Radiology Department at the Centre of Dentistry, Oral Medicine and Maxillofacial Surgery with
Dental School, Tuebingen University Hospital, Tübingen, Germany
A R T I C L E I N F O A B S T R A C T
Article history: Purpose: To set up the digital database (DDB) of various anatomical parts, skin details and retention
Received 24 September 2018
elements in order to simplify the digital workflow of facial prostheses manufacturing; and to quantify the
Received in revised form 8 January 2019
reproduction of skin wrinkles on the prostheses prototypes with stereolithography (SLA) and direct light
Accepted 17 January 2019
processing (DLP) methods.
Available online 18 February 2019
Methods: Two structured light scanners were used to obtain the nasal and auricle forms of 50 probands.
Furthermore, the ala nasi and scapha areas were captured with the digital single lens reflex camera and
Keywords:
saved in jpeg format. The four magnetic retention elements were remodeled in computer aided design
Nasal prostheses
(CAD) software. The 14 test blocks with embossed wrinkles of 0.05–0.8 mm were printed with SLA and
Auricular prostheses
DLP methods and afterwards analyzed by means of profilometry and confocal microscopy.
Silicone prostheses
Anaplastology Results: The introduced DDB allows for production of customized facial prosthesis and makes it possible
Additive manufacturing to consider the integration of concrete retention elements on the CAD stage, which makes the prosthesis
modelling more predictable and efficient. The obtained skin structures can be applied onto the prosthesis
surface for customization. The reproduction of wrinkles from 0.1 to 0.8 mm in depth may be associated
with the loss of 4.5%–11% of its profile with SLA or DLP respectively. Besides, the reproduction of 0.05 mm
wrinkles may be met with up to 40% profile increasement.
Conclusions: The utilization of DDB may simplify the digital workflow of facial prostheses manufacturing.
The transfer of digitally applied skin wrinkles till the prostheses’ prototypes may be associated with
deviations from 11 to 40%.
© 2019 Japan Prosthodontic Society. Published by Elsevier Ltd. All rights reserved.
1. Introduction to enhance the workflow of facial prostheses manufacturing in
terms of efficiency, ease of production and patients burdens [2].
Patients afflicted by facial defects experience immense Various protocols of AM utilization have been introduced,
psychological pressure and usually undergo a long lasting featuring direct mold making (DMM) [3], indirect mold making
treatment process in order to camouflage their mutilation by (IMM) [4], and also rapid manufacturing (RM) with direct silicone
means of a prosthetic appliance [1]. The utilization of computer printing of the final prostheses [5,6]. The last mentioned approach
aided design (CAD) and additive manufacturing (AM) was reported remains to be a potentially valid option for the future, but can be
considered only for provisional solutions, today [7]. The approach
of IMM was shown to achieve a more predictable rehabilitation
outcome in terms of overall esthetics [8].
* Corresponding author at: Department of Prosthodontics at the Centre of
Whichever AM approach is chosen, a virtual prosthesis has to be
Dentistry, Oral Medicine, and Maxillofacial Surgery with Dental School, Tubingen
constructed first and commonly in standard tessellation language
University Hospital, Osianderstr. 2-8, 72076 Tübingen, Germany.
E-mail address: [email protected] (A. Unkovskiy). (STL) format. In case of unilateral loss of an auricle the native
https://doi.org/10.1016/j.jpor.2019.01.004
1883-1958/© 2019 Japan Prosthodontic Society. Published by Elsevier Ltd. All rights reserved.
314 A. Unkovskiy et al. / journal of prosthodontic research 63 (2019) 313–320
anatomy can be captured with a surface scanner and adopted from consideration such categories as administration, functionality,
the opposite side with the use of a “mirror-imaging technique” [9]. query, storage, safety and easiness of use.
However, a bilateral auricle absence and also facial defects, which The open access software package XAMPP (Version 5.2.7. for MS
cross the facial midline, would restrict such an approach. Windows) developed by a non-profit project Apache Friends [17]
Thus, the three dimensional (3D) design calls for experience was chosen. It is SQL-based and imitates server functionality on a
with complex software solutions for free-form sculpting. This local computer. This way it provides an opportunity to load local
requires not only technical skills and artistic vision, but also plenty saved data in a web browser through the administration service
of time. with the use of phpMyAdmin [18].
That is why an alternative data source i.e. a digital database
(DDB) of facial parts with organ-specified anatomy would ease the 2.1.1. Integrated anatomical forms
CAD process. Ciocca et al. highlighted potential benefits of a DDB In order to preliminarily populate the DDB with nasal and
utilization [10]. auricular anatomy, 50 probands aged from 19 to 62 years gave
Such DDB for facial prosthetics should not only provide their informed consent to undergo the light scanning pro-
“custom” facial parts, but also customized supportive objects cedure and the usage of their 3D data free of charge. The ethical
and soft tissue surface textures to improve function and esthetics. committee of the Tuebingen University Hospital voted affirma-
As far as prosthesis function is concerned, the virtual tively (037/2018BO1).
implementation of retention elements is crucial in a digital For the auricular forms, mastoid area was isolated from hair
workflow of facial prostheses manufacturing [11]. Following the with a medical cap and the left auricle of each proband was
bone availability, the transdermal implant position might be defatted with 80% ethanol. In order to capture the complex ear
suboptimal for the prosthetic supply. In order to predict and cope anatomy with its numerous undercuts and adjacent tissue portion
with shortage of space, a virtual set of commonly utilized retention the portable Artec Spider scanner (Artec 3D, Luxembourg,
elements (i.e. magnet or bars) would allow the best choice during Luxembourg) was used. The gathered raw data underwent the
the prosthetic CAD design. post-processing stage with Artec Studio software (ver. 11, Artec
With regard to the esthetical outcome and camouflage, skin 3D). The output file in STL-format was assigned with an ID: “O-DB-
structure details must be applied onto the prosthesis surface [12]. ###” and sent to Zbrush software (Pixologic, Inc., Angeles, CA,
The facial skin wrinkles have been measured to range between USA) for the further CAD preparation. Therefore, each virtual
0.06 to 0.94 mm in depth, and their full reproduction by means of a model was cut with a digital knife tool leaving 1 cm of adjacent
digital workflow might be hardly feasible [13,14]. Neither is the tissue around the auricle. The areas between helix and scapha,
resolution of CT with its voxel size of 0.2–0.5 mm, nor the light helix and antihelix, tragus and antitragus as well as ear canal were
scanners with an accuracy of 0.02 mm are able to capture and revisited, whether they were captured distinctively. Further
describe wrinkles of 0.1 mm or less in depth [15]. Thus, it may be artefacts and surfaces depiction shortcomings were edited with
useful to apply the desired amount of skin details out from a DDB following digital brush tools listed in Table 1.
set during the CAD process. For the capturing of the nasal anatomy each proband was
In the topical literature only one DDB for nasal anatomy was scanned with pritimirror system (pritidenta GmbH, Leinfelden-
found [16]. Thus, the first aim of the present study was to generate Echterdingen, Germany) following the manufactures specifications
an extensive DDB for facial prosthetics including auricular and with support of pritiimaging software (pritidenta GmbH). The 3D
nose anatomy, exemplary custom retention elements as well as a image of the whole face was exported in a native OBJ format and
preliminary set of skin wrinkles and skin surface textures. imported to Zbrush software where the nose was cut out with 1 cm
These efforts should make a digital workflow of facial prosthesis of adjacent tissue portion. The nostrils were cut through the nose
manufacturing more applicable in a daily routine and more bulk and the inner surface was smoothed out to be ready for the
attractive for maxillofacial technicians. alignment to a defect side. Each virtual model was assigned with an
Furthermore, the set of skin wrinkles should be validated by a ID: “N-DB-###” and saved in STL-format. All gathered nose and
quantification of skin wrinkles reproduction with two 3D printing auricular 3D models were preliminarily labeled with age (years)
systems based on stereolithographie (SLA) and digital light and gender (female/male) as a later search criteria for the user.
processing (DLP) in context of IMM approach. It was hypothesized
that these AM machines with a printing resolution of 0.025 mm 2.1.2. Implementation of retention elements
both are able to reproduce faithfully the skin wrinkles starting As a first example for the database it was decided to implement
from 0.05 mm. the magnetic retention elements (Technovent Limited, Bringend,
South Wales, UK) that are frequently used in Department of Oral
2. Materials and methods and Maxillofacial Surgery of the University Clinic, Tuebingen. After
clearance with the patent owner, the original parts were measured
2.1. Populating the database with a digital caliper (DIGI-MET 0–25 mm, Helios-Preisser,
Gammertingen, Germany) and their dimensions were recon-
Considering the complexity of a database concept a suitable structed in NX software (NX 10.0, Siemens Product Lifecycle
database management system must be found, taking into Management Software Inc.) (Fig. 1). The imported STL-files were
Table 1. Digital tools in the Zbrush software and their preferences for processing of raw scan data of facial anatomy.
Brush code Draw size (units) Focal shift (units) Z intensity (units) Application
BMV 0–20 0 50 Deepen folds
BPO 20 À40 20 Smoothing
BTD 20 À50 40 Remove blobs
BCB 15–30 À50 30 Treating undercuts
BDS 17 40 23 Draw fine wrinkles
BIN 0–20 0 10 Closing holes
BNO 0–20 0 10 Applying texture
BST 1000 À100 25 Applying texture
A. Unkovskiy et al. / journal of prosthodontic research 63 (2019) 313–320 315
Fig. 1. Technovent magnets constructed in NX software to be implemented in a
database.
labeled in the database with information about width/diameter,
height, and the order number of the manufacturer.
2.1.3. Implementation of skin textures and wrinkles
In order to create a library of skin texture details, probands’ skin
surface was captured chairside with the use of digital single-lens
Fig. 2. (A) Overview to the virtual design of the 40 Â 20 mm test block with
reflex camera (DSLR) (Nikon D300s, Nikon Corp., Tokio Japan) with
wrinkles (lines) of 0.05, 0.1, 0.2, 0.4, 0.6, 0.8 mm in depth, designed with the BDS
a macro lens (AF-S Dx Micro Nikkor 85 mm 1:3,5G ED VR, Nikon tool. (B) Side aspect in magnification of the 0.4 mm wrinkle. Measuring of the
Corp.), tele converter (Kenko N-AF 2x Teleplus Pro 300, Nikon wrinkle depth in Zbrush software from the deepest point to the wrinkle margin
with a use of “transpose tool” (BTS).
Corp.) and flash unit (Mecablitz 15 MS-1, Metz, Zirndorf, Germany).
The ala nasi area was chosen to capture nose skin surface and the
scapha area was captured for auricles skin surface. Additionally,
2.2.2. Wrinkle profile measurements
the peel of a naval orange was photographed with the same camera
The wrinkle depth on each test block was examined in cross
in a light box. All these structure images were saved in JPG-format
section under microscope (M 400 Photomakroskop, Wild-Heer-
categorized with the respective area to the database.
brugg AG, Heerbrugg, Switzerland) with 64Â magnification and
captured with a camera (Canon Eos 700D, Canon Inc., Tokio, Japan).
2.2. Validation tests for the reproduction on example of wrinkles
The wrinkles depth was measured software-based on the screen
(Datinf Measure, Datinf GmbH, Tuebingen, Germany) (Fig. 3).
2.2.1. Preparing the test blocks
Three repeated measurements of each wrinkle were recorded.
In order to define the level of skin surface reproduction by
Further surface analysis was performed with profilometer
means of modern AM methods a test block 40 Â 20 Â 5 mm was
(Mahr S6E, Mahr GmbH, Goettingen, Germany). In order to ensure
designed in Zbrush software. Wrinkles ranging from 0.05 to
the reproducibility of all measurements a retainer was designed
0.8 mm in depth, according to the classification of Lemperle, were
and printed with fused deposition modelling (FDM) printer
embossed evenly on the test blocks [13]. Holding the “Shift” button
(Replicator, MakerBot Industries, Brooklyn, NY, USA) to fix the
ensured the horizontal position of the test block. A line was
test blocks on the profilometer testing area.
dragged downwards from the upper block border in 90 angle with
Due to the probes’ amplitude limitation of 0.25 mm, only
the Damian Standard Brush (BDS) — ensured by holding the “Shift”
wrinkles below that value – namely 0.05, 0.1 and 0.2 mm in depth –
button (Fig. 2A). The brush preferences listed in Table 1 were used
could be analyzed this way. The measuring probe drove
for the creation of various wrinkles depth. To ensure a precise
perpendicular to the wrinkle orientation. Three repeated profiles
depth of each wrinkle all digital tools were calibrated in mm first. A
for each test block were recorded. The acquired data was analyzed
cube form with defined dimension (1 Â1 Â1 mm) was designed in
in measuring software (MountainsMap universal 7.3, Digital Surf,
NX software and uploaded in Zbrush. The Transpose Brush (BTS)
Besançon, France) and converted into profile curves. The absolute
was used to measure the uploaded cube in standard for this
profile depth (Rv) value was obtained for each wrinkle (Fig. 4).
software “units” and so to define the “unit-to-mm” ratio. This
allowed calibrating transpose tool (BTS) scale in mm. The profile of
2.2.3. Statistical methods for quantification and comparison of the
each wrinkle was measured by dragging the “BTS” line from the
data
deepest wrinkle profile point up to the test block upper border,
All gathered data was analyzed with JMP 13.1 software package
which was the nil level for each wrinkle. Holding the “Shift” button
(SAS Corp., Heidelberg, Germany). The distributions of the
ensured the vertical position of the test block and the same vertical
measurements (wrinkle depth; dependent variable) were grouped
direction of BTS line. This also allowed stopping the BTS line right
by the AM machine (DLP and SLA; independent variable) and
on the nil level of the test block automatically (Fig. 2B).
The constructed digital test block with embossed wrinkle
profiles were printed 14 times each with streolithography (SLA)
(Form 2, Formlabs, Somerville, MA, USA) from “Grey Resin”
material (Formlabs) and direct light processing (DLP) (Solflex 650,
W2P, Klosterneuburg, Austria) from “V-print model” material
(VOCO, Cuxhaven, Germany) with 45 printing orientation to the
build platform. For both 3D printers the printing resolution was set
on 0.025 mm. The printed test blocks were post-processed
immediately as follows: 6 min storing in isopropanol 80%, drying
for 30 min in room air, and lightcuring with 2 Â 2000 flashes
(Otoflash G171, VOCO). In order to get rid of any printing artefacts
on the test block edges, 5 mm were cut off orthogonally with the
Fig. 3. Wrinkle depth measurement in Datinf software. Three repeated measure-
micro saw (Accustom-50, Strues, Willich, Germany) exhibiting the
ments were carried out from the deepest wrinkle points to the wrinkle margin and
wrinkles cross-section for the further analysis.
the average was used for calculation.
316 A. Unkovskiy et al. / journal of prosthodontic research 63 (2019) 313–320
auricle, was adopted from a DDB by entering patient’s sex and age
(Æ5 years) (Fig. 5C). The picked image in STL format was then
properly positioned to the defect side and its margins were aligned
to the adjacent soft tissue (Fig. 5D). The prosthesis form can be also
manipulated with a bunch of digital tools to align symmetry and
reach facial harmony in an overview picture, as described
elsewhere [14].
Fig. 4. Winkle depth profilometrical measurement in MountainsMap software. The
As seen here and found in most cases, a shortage of space may
green line represents the wrinkle depth from the deepest profile point to the
lead to “digital” perforation of the retention elements to the
wrinkle margin.
prosthesis outer surface (Fig. 6A). Hence, a virtual clay tool is used
for a selective build-up of the prosthesis bulk (Fig. 6B).
The uploaded virtual magnet construction serves like a
tested for normality of distribution by goodness of fit with placeholder (Fig. 6C), as it can be then cut out from the prosthesis
Shapiro–Wilk test [19] and homoscedasticity of the compared bulk with a Boolean function, providing thereby the socket of
distributions with the Barlett test, using p < 0.05 for both. corresponding geometry congruent to the prosthesis magnet
In case of a non-normal distribution within a set of chosen (Fig. 6D).
comparisons, the Wilcoxon rank sum test was used to evaluate Once the prosthesis is adjusted in terms of form and relation
statistical difference using alpha = 0.05; otherwise ANOVA Tukey– (Fig. 7A), the skin details can be applied. This is for example
Kramer post-hoc test was applied on the same alpha level of 0.05. possible in the ZBrush software by uploading the jpeg file from the
DDB, which can be embossed onto the prosthesis surface (Fig. 7B).
3. Results Interestingly, application of the naval orange peel yielded a
naturally looking skin surface.
3.1. Application of the database After finishing these stages the prosthesis is ready for
materialization (CAM). If not printed directly in silicone, a
The DDB called “Epi-Database” for the digital workflow of facial prosthesis prototype is printed in wax for a chairside try-in
prostheses manufacturing was set up on a 2TB HDD (WD Elements session. Such prototypes can be transferred to silicone prosthesis
portable, San Jose, CA, USA). After the CAD stage 21 post-processed using the conventional molding technique (Fig. 8A), including the
virtual auricles and 48 noses were implemented as labeled integration of magnets into prosthesis bulk, according to their
datasets (categorized by age and sex) within the database. position in the CAD (Fig. 8B).
Additionally, retention magnets of four sizes and various skin
details were generated. 3.1.2. Nasal prosthesis
The following pathway describes the utilization of the database Rehabilitation of a nasal defect and constructing a nasal
on two clinical examples. prosthesis in a digital workflow has some peculiarities. Thus,
not only the organ outer anatomy must be restored but also the
3.1.1. Auricular prosthesis inner one, considering the reestablishment of breathing entrances.
A virtual image of the deficient side with implant magnetic This case demonstrates schematically the utilization computed
attachments (VF-MC1-S, Mini Magnacap, Technovent) being tomography (CT) as the main virtual data source of the deficient
screwed onto the implants has been obtained with Artec Spider area. Once a CT has been obtained the bony and soft tissue
scanner and uploaded in Zbrush software. This way the magnetic structure can be easily segmented in a special software (Invesalius,
attachments act as scan bodies and allow for locating their position Renato Archer CTI, Campinas, Brasil) (Fig. 9A), which allows for
in the CAD software, the same way it is done for the intraoral translation of DICOM data in STL format. The extracted STL model
caption of implant position. For this purpose the three cylinders of of the facial (soft tissue) with a nasal defect and the skull anatomy
the same geometry and dimensions are put over the obtained were then uploaded into the Zbrush freeform software (Fig. 9B).
virtual images of the original magnetic attachments. (Fig. 5A). The suitable anatomical form of a female nose aged between 35
Three suitable prosthesis magnets in STL format, in this case ML-2S and 50 was appended and aligned to the adjacent soft tissue. The
magnets (Technovent), were appended onto the screwed implant nasal anatomy was checked to fit the overall facial proportions and
attachments (Fig. 5B). Afterwards a missing facial part, in this case shown to the patient (Fig. 9C). Once the position and form were
Fig. 5. CAD of auricle prosthesis. (A) Locating the implant position on the defect side; (B) placement of magnets of the chosen geometry onto the implants; (C) adoption of a
suitable auricle anatomy out from the database according to patients’ sex and age; (D) adjustment of auricle position and alignment of prosthesis margins.
A. Unkovskiy et al. / journal of prosthodontic research 63 (2019) 313–320 317
Fig. 6. CAD individualization of auricular prosthesis. (A) Perforation of the chosen
Fig. 9. (A) Segmentation bony structures and soft tissues using the CT data in the
magnets size through the surface. (B) Coverage of the magnets with a virtual build
Invesalius software; (B) the STL file of bony structures and soft tissue being exported
up; (C) cut out of the definitively positioned magnets from the prosthesis bulk with
in the Zbrush software; (C) applying of the nasal prosthesis from a DDB according to
a Boolean function; (D) finally the prosthesis bulk owes a placeholder box
the age and gender and its further alignment to the adjacent soft tissue; (D)
congruent to the chosen magnet profile.
blending of the soft tissue mask and virtual implant (gold bar) placement following
both the bone availability and the adjusted prosthesis position.
Fig. 7. Auricle prosthesis after being aligned virtually to the adjacent tissue (A) and
after application of skin details onto its surface (B).
Fig. 10. (A) The implant (gold bar) position and relation to prosthesis bulk was
approved; (B) the ML-2S magnet (Technovent) was appended over the implant and
fixed virtually in the prosthesis bulk; (C) the magnet was cut out with the Boolean
function, leaving thereby a socket of corresponding geometry.
function allowed for creation of a corresponding socket for a future
real magnet (Fig. 10C). As final stage of CAD the photographs of
orange peel and ala nasi were embossed onto the prosthesis
surface to add some details the same way it was done for the
auricular prosthesis.
3.2. Validation of the digital wrinkles after materialization
The mean values of the wrinkles depth reproduced with SLA
Fig. 8. (A) a finished auricle prosthesis manufactured with a use of the RP approach;
fi
(B) magnets (Technovent) being integrated into prosthesis bulk, according to their and DLP methods gathered by an optical and pro lometrical
position in the virtual design. analysis are shown in Table 2.
The Shapiro–Wilk test revealed non-normal distributions for
the data gathered with the stereomicroscopy. For this reason the
approved, the soft tissue mask could be blended, so that the Wilcoxon test was used to evaluate statistical difference using
relation of the prosthesis to the bony structures was clearly seen alpha = 0.05; for the data gather with profilometry a normal
(Fig. 9D). This allowed for the implant position planning, taking distribution was observed. Therefore the Tukey–Kramer post-hoc
into consideration both, bone availability and future prosthesis test was applied on the alpha level of 0.05.
position. As nowadays there is no digital database for the According to the optical analysis, the wrinkles from 0.1 to
commonly used extraoral implants, the cylinder of a corresponding 0.8 mm reproduced by both printers were not able to reach the
geometry with the same cross section was appended and acted as reference values. Interestingly, the wrinkles of 0.05 mm became
an implant. Once the implant was virtually placed (Fig. 10A), the even more pronounced (18–40% “deeper” than the reference
magnet (in this case ML-2S, Technovent) could be put upon it and values with both SLA and DLP). The wrinkles from 0.1 to 0.8 mm
checked once again, whether it fits in a prosthesis bulk (Fig. 10B). were reproduced more precisely with DLP, having the maximum
The virtual extraction of the magnet 3D model with a Boolean deviation of À4.5% to the reference value. In case of SLA deviations
318 A. Unkovskiy et al. / journal of prosthodontic research 63 (2019) 313–320
Table 2. Mean values of wrinkle depth for the test blocks made with stereolithography (SLA) and digital light processing (DLP). The first row represents the mean wrinkle
depth gathered by stereomicroscopy and profilometry, followed by standard deviation and relative deviation of the mean measurements to the reference value in percent (%).
Shapiro–Wilk test was used for the assessment of data normality and Barlett test for homoscedasticity. As in case of stereomicroscopy the Shapiro–Wilk test revealed a non-
normal data distribution, the non-parametric Wilcoxon analysis was used for statistical comparison. For profilometry due the normal data distribution the statistical
significance was analyzed with Tukey Kramer test.
Wrinkles depth in mm
0.05 0.1 0.2 0.4 0.6 0.8
Stereomicroscopy
Mean Ry SLA; SD (mm); relative deviation to 0.059; 0.008; 0.088; 0.010; 0.178; 0.008; 0.364; 0.014; 0.548; 0.024; 0.747; 0.032;
reference (%) +18% À8% À11% À9% À8.7% À6.4%
Shapiro–Wilk test W = 0.88 W = 0.89 W = 0.95 W = 0.96 W = 0.93 W = 0.93
p = 0.0004 p = 0.001 p = 0.056 p = 0.257 p = 0.013 p = 0.011
Mean Ry DLP; SD (mm); relative deviation to 0.070; 0.010; 0.10; 0.007; 0 0.191; 0.010; 0.390; 0.016; 0.595; 0.018; 0.797; 0.021;
reference (%) +40% À4.5% À2.5% À0.8% À0.4%
Shapiro–Wilk test W = 0.97 W = 0.96 W = 0.91 W = 0.96 W = 0.97 W = 0.9
p = 0.283 p = 0.247 p = 0.0027 p = 0.284 p = 0.387 p = 0.017
Comparison mean Ry SLA/DLP
Barlett-Test F = 2.63 F = 5.35 F = 1.97 F = 0.91 F = 3.26 F = 7.51
p = 0.11 p = 0.021 p = 0.159 p = 0.34 p = 0.071 p = 0.006
Wilcoxon rank sum test α = 0.05 Z = À4.85 Z = À5.11 Z = À5.74 Z = À6.33 Z = À7.21 Z = À6.64
p < 0.0001 p < 0.0001 p < 0.0001 p < 0.0001 p < 0.0001 p < 0.0001
Profilometry
Mean Ry SLA; SD (mm); relative deviation to 0.047; 0.005; 0.085; 0.009; 0.171; 0.007;
reference À6% À15% À14.5%
Shapiro–Wilk test W = 0.86 W = 0.86 W = 0.88
p = 0.067 p = 0.078 p = 0.132
Mean Ry DLP; SD (mm); relative deviation to 0.064; 0.011; 0.105; 0.012; +5% 0.195; 0.009;
reference +28% À2.5%
Shapiro–Wilk test W = 0.85 W = 0.94 W = 0.97
p = 0.64 p = 0.56 p = 0.88
Comparison mean Ry DLP/SLA
Barlett-Test F = 3.65 F = 0.7 F = 0.83
p = 0.56 p = 0.41 p = 0.36
Tukey Kramer α = 0.05 p < 0.0004 p < 0.0009 p < 0.0001
ranged between À6 and À11% not reaching the true value. The [2,3], some researches today have concentrated rather on the
results of the profilometry coincided with the optical analysis and expediency and easiness of its application [11,20]. The main
are presented in Table 2. limitations highlighted in the topical literature are the high costs
Furthermore, the deviations of relative differences for each off soft- and hardware solutions for data acquisition, CAD and
value gathered with stereomicroscopy were depicted in Fig. 11. It materialization. These technologies call not only for technical
can be seen again, that the wrinkles from 0.1 to 0.8 mm were experience, but also for digital sculpting skills to perform the CAD
generally reproduced better with the DLP and became rather stage and for a certain structured pathway to translate the CAD
smoothed. Whereas the wrinkles of 0.05 mm became even more data into a final prosthesis
pronounced and were reproduced nearly with the same accuracy
with both SLA and DLP. 4.1. Significance of a database
As the certain deviations in wrinkles reproduction were
revealed, the hypothesis of the present study can be rejected. As demonstrated in this study, a database for virtual facial parts
and skin details could be setup by open access solutions. Such a
4. Discussion database could support a free-form software, such as ZBrush,
during the CAD stage. Moreover, the content of this database may
Since the general feasibility of digital workflow for facial also allow the usage of patients’ CBCT datasets instead of STL-files
prostheses manufacturing has been successfully demonstrated delivered by costly and sophisticated surface scanners, such as
Artec Spider, which is priced about 20.000 USD plus 500 USD
annual software license [21,22]. Obviously, the CBCT with its voxel
size ranging from 0.2 to 0.4 mm is not able to describe the skin
surface structure faithfully, however may provide sufficient
information about the general facial proportion and anatomy of
the defect and donor-organ. Thus, the skin details can be adopted
from the database and embossed on a prosthesis pattern on a CAD
stage, as it was demonstrated in the second clinical case.
For basic CAD manipulations, i.e. adjusting the prosthesis
position and adjustment of prosthesis edges a row of free software
solutions such as MeshLab (http://www.meshlab.net/) can be
found in web and successfully used for virtual data editing in
maxillofacial prosthetics [23]. Surely the pre-saved anatomical
forms from the database need to be adapted to the patients’
Fig. 11. Deviations of mean relative differences for each value gathered with the
general facial proportions and to the soft tissues adjacent to the
stereomicroscopy. The black line represents the reference value of each wrinkles
defect. However, these refinements and adjustments may be not
depth (0.05, 0.1, 0.2, 0.4, 0.6, 0.8 mm). The blue spots represent the relative mean
that time-consuming, as sculpting of prosthesis from the very
differences of each value gathered by each single measurement of the DLP
reproduced wrinkles and the red spots — of SLA. beginning. The free-form sculpting of a nose, for instance, begins
A. Unkovskiy et al. / journal of prosthodontic research 63 (2019) 313–320 319
with the appendence of a cube as a basic pattern, which should be 4.3. Skin surface reproduction
then transformed with the virtual clay tool in to a nasal anatomy.
According to authors’ experience such manipulations may cost The wrinkles from 0.05 to 0.8 mm in depth were visible on the
time, especially the creation of nostrils and achieving of a smooth AM-manufactured test blocks. However, the present study
surface adjacent to the defect site, without affecting the prosthesis disclosed certain shortcomings in their reproduction by means
form and overall wall-thickness. The utilization of a database may of SLA and DLP printers. Generally, the SLA-reproduced wrinkles
allow for time-savings here, as demonstrated in a clinical case. showed greater deviations to original wrinkle depth values than
To the authors’ best knowledge, all existing databases today DLP, as they became less pronounced. This may be attributed to the
include only nasal anatomy [16]. The present study shows a pathway wise the layers are cured together. Thus, in case of SLA a laser beam
how to create and utilize the more extensive DDB, which covers also travels the layer surface curing the particles step by step in a
the attachments, auricle anatomy, and skin surface details. Anyhow, sequence defined by a printing software, whereas in case of DLP a
the categorization of the datasets with descriptors included to the whole layer is cured at once. The earlier studies showed that the
database has to be refined further. For attachments this is most easy wrinkles of 0.1 can be faithfully reproduced on a visible level with
due to its manufacturer, item number as well as their dimensions. the use of Thermojet printer and a 0.04 mm printing accuracy [15].
The selection via age and gender is a limitation that might be However the reproduction capability was not quantified for this
overcome with the implementation of height-length ratios as well as stage of digital workflow.
sophisticatedanatomical features.Such a setofdescriptorsshouldbe The present study allows for identifying the level of texture
condensed by experts (MFTs together with maxillofacial surgeons). details loss while transferring them from CAD to AM. Thus,
Same should be performed for skin structures and wrinkles. reproduction of wrinkles from 0.1 to 0.8 mm in depth may be
The least but not last aspect of CAD in maxillofacial prosthetics is associated with the loss of 4.5%–11% of its profile, dependent on a
the planning of the prosthesis anchoring [11]. The recent technical printer used — DLP or SLA respectively. Besides, the reproduction of
digital protocols reported, disregard this aspect, and this technical 0.05 mm wrinkles may be met with up to 40% profile incensement.
step is accomplished rather in analog way. The utilization of the It must be emphasized that reproduction of only skin wrinkles
digitalizedretentionelements mayallow finding anoptimal position does not cover fully the aspect of reproduction accuracy (both
in the limited space of a prosthesis pattern on the CAD stage prior to trueness and precision). The skin surface is more complex
the chairside try-in session. The use of CT also allows here for the structure, than just a sum of wrinkles and includes also fine
backwards-planning, which begins with the virtual placementof the furrows and pores. These anatomical structures, however, may be
implant. This manipulation is, however still very raw, as no database difficult to quantify and reproduce on a CAD stage in order to use
(including the created one) and software include the virtual images them as a referent value. For this reason in the present study the
of the commonly used implants in various dimensions, as it is done, wrinkles of a predetermined profile were chosen in order to assess
for instance, in the intraoral rehabilitation to perform a backwards- the reproduction capabilities of CAM with the chosen AM methods.
planning [24].Inthose cases,where the implantsare insertedpriorto The reproduction of more complex surface profiles may be
the facial scanning, the so-called scan bodies are required for a considered in further studies and calls for utilization of a more
consistent capturing of the implant position. accurate hardware for the surface analysis, such as laser
For this reason in the first clinical case the implants were microscopy, for instance.
scanned with the magnet attachments being screwed on, which
acted this way as scan bodies. As far as the structured light scanner 5. Conclusion
used was not able to capture the magnetics attachments faithfully,
the cylinders of the same dimensions and the same cross section The digital workflow in maxillofacial prosthetics still can not be
were aligned with their virtual images in the CAD software. This considered as state of the art in this brunch of rehabilitation. The
may lead to certain discrepancies between the real and virtual utilization of DDB may aid for its more spread application and a
implants positions. The same problem is true in the field of more consistent utilization by primarily decreasing the invest-
intraoral rehabilitation, but its discussion can be found in ments and simplifying the CAD stage.
corresponding researches [25]. Reproduction of digitally applied skin wrinkles may be
associated with deviations of 11–40% details depending on the
4.2. Significance of surface structures reproduction initial wrinkle depth. The virtual application of skin details have to
consider the concrete later AM machine and therefore a single set
It was shown that the reproduction of skin structure depends of skin details is not enough for a DBB.
on the AM method chosen. The clinical relevance of the encountered deviations of
The simple validation of only wrinkles’ reproduction in this wrinkles’ profile is difficult to assess on the stage of prosthesis
study may not represent the reproduction of the real skin surface. prototype, as it remains unclear which amount of skin details could
The findings of the present study suggest that more complex be transferred till the final silicone prosthesis. The quantification of
structures, such as pores or furrows, may be also affected skin details amount that could be seen from private and social
throughout the digital workflow. Furthermore, this study follows distances must be addressed in future researches.
the skin wrinkles reproduction till the prosthesis prototype in
context of IMM approach, but not till the final silicone prosthesis. Acknowledgement
This should be clarified in further studies, including the impact of
processing of prototypes (wax casting of a prototype replica and its The present study was financially supported by the “Research
transfer into medical silicone), which may impair the surface Foundation Dental e.V.”. Furthermore, the authors thank the W2P
structure in the final prosthesis. Engineering GmbH and VOCO GmbH for the hardware and
This implies two rationales. First, the users of the database have materials supply.
to validate the outcome of applied structures and wrinkles towards
their reproduction with the specific CAM. Second, and conse-
References
quently, these structures have to be available in graduated
embossments in CAD software to allow a case specific selection/
[1] Taylor TD. Clinical maxillofacial prosthetics. Chicago, London: Quintessence
application in the virtual model. Publishing Company; 2000.
320 A. Unkovskiy et al. / journal of prosthodontic research 63 (2019) 313–320
[2] Ariani N, Visser A, van Oort RP, Kusdhany L, Rahardjo TB, Krom BP, et al. Current texture reproduction of auricular prostheses replicas. J Prosthodont
state of craniofacial prosthetic rehabilitation. Int J Prosthodont 2013;26:57–67. 2017;00:1–9.
[3] Yadav S, Narayan AI, Choudhry A, Balakrishnan D. CAD/CAM-assisted auricular [15] Eggbeer D, Evans PL, Bibb R. A pilot study in the application of texture relief
prosthesis fabrication for a quick, precise, and more retentive outcome: a for digitally designed facial prostheses. Proc Inst Mech Eng H 2006;220:705–
clinical report. J Prosthodont 2017;26:616–21. 14.
[4] Watson J, Hatamleh MM. Complete integration of technology for improved [16] Reitemeier B, Götzel B, Schöne C, Stockmann F, Müller R, Lexmann J, et al.
reproduction of auricular prostheses. J Prosthet Dent 2014;111:430–6. Creation and utilization of a digital database for nasal prosthesis models.
[5] Jindal SK, Sherriff M, Waters MG, Coward TJ. Development of a 3D printable Onkologie 2013;36:7–11.
maxillofacial silicone: part I. Optimization of polydimethylsiloxane chains and [17] XAMPP for Windows (PHP 7.3.0). https://www.apachefriends.org. [Accessed
cross-linker concentration. J Prosthet Dent 2016;116:617–22. 24 September 2018].
[6] Jindal SK, Sherriff M, Waters MG, Smay JE, Coward TJ. Development of a 3D [18] PhpMyAdmin MySQL administration tools. https://www.phpmyadmin.net/.
printable maxillofacial silicone: part II. Optimization of moderator and [Accessed 24 September 2018].
thixotropic agent. J Prosthet Dent 2018;119:299–304. [19] Shapiro S, Wilk M. An analysis of variance test for normality (complete
[7] Unkovskiy A, Spintzyk S, Brom J, Huettig F, Keutel C. Direct 3D printing of samples). Biometrika 1965;52:591–611.
silicone facial prostheses: a preliminary experience in digital workflow. J [20] Eggbeer D, Bibb R, Evans P, Ji L. Evaluation of direct and indirect additive
Prosthet Dent 2018;120:303–8. manufacture of maxillofacial prostheses. Proc Inst Mech Eng H 2012;226:718–
[8] Unkovskiy A, Brom J, Huettig F, Keutel C. Auricular prostheses produced by 28.
means of conventional and digital workflows: a clinical report on esthetic [21] Coward TJ, Scott BJ, Watson RM, Richards R. A comparison of prosthetic
outcomes. Int J Prosthodont 2018;31:63–6. ear models created from data captured by computerized tomography,
[9] Nuseir A, Hatamleh M, Watson J, Al-Wahadni AM, Alzoubi F, Murad M. magnetic resonance imaging, and laser scanning. Int J Prosthodont
Improved construction of auricular prosthesis by digital technologies. J 2007;20:275–85.
Craniofac Surg 2015;26: e502-5. [22] Coward TJ, Watson RM, Richards R, Scott BJ. A comparison of three methods to
[10] Ciocca L, Fantini M, Marchetti C, Scotti R, Monaco C. Immediate facial evaluate the position of an artificial ear on the deficient side of the face from a
rehabilitation in cancer patients using CAD-CAM and rapid prototyping three-dimensional surface scan of patients with hemifacial microsomia. Int J
technology: a pilot study. Support Care Cancer 2010;18:723–8. Prosthodont 2012;25:160–5.
[11] Daniel S. Towards developing CAD/CAM solutions in the retention of extra-oral [23] Salazar-Gamarra R, Seelaus R, da Silva JV, da Silva AM, Dib LL. Monoscopic
facial prosthetics. 2014 [Ph.D. thesis]. p146-70. photogrammetry to obtain 3D models by a mobile device: a method for
[12] Eggbeer D. The computer aided design and fabrication of facial prostheses (Ph. making facial prostheses. J Otolaryngol Head Neck Surg 2016;45:33.
D. thesis). Cardiff: University of Wales Institute; 2008. [24] Pozzi A, Arcuri L, Moy PK. The smiling scan technique: facially driven guided
[13] Lemperle G, Holmes RE, Cohen SR, Lemperle SM. A classification of facial surgery and prosthetics. J Prosthodont Res 2018;62:514–7.
wrinkles. Plast Reconstr Surg 2001;108:1735–50 discussion 1751-2. [25] Andriessen FS, Rijkens DR, van der Meer WJ, Wismeijer DW. Applicability and
[14] Unkovskiy A, Spintzyk S, Axmann D, Engel EM, Weber H, Huettig F. Additive accuracy of an intraoral scanner for scanning multiple implants in edentulous
manufacturing: a comparative analysis of dimensional accuracy and skin mandibles: a pilot study. J Prosthet Dent 2014;111:186–94.