Almukhtar, Anas (2016) Three dimensional study to quantify the
relationship between facial hard and soft tissue movement as a result
of orthognathic surgery. PhD thesis.
http://theses.gla.ac.uk/7364/
Copyright and moral rights for this thesis are retained by the author
A copy can be downloaded for personal non-commercial research or study
This thesis cannot be reproduced or quoted extensively from without first
obtaining permission in writing from the Author
The content must not be changed in any way or sold commercially in any format
or medium without the formal permission of the Author
When referring to this work, full bibliographic details including the author, title,
awarding institution and date of the thesis must be given
Glasgow Theses Service
http://theses.gla.ac.uk/
theses@gla.ac.uk
Three dimensional study to quantify the relationship
between facial hard and soft tissue movement as a
result of orthognathic surgery.
Anas Mohammed Yousif Almukhtar
B.D.S., M.Sc. Orthodontics
Submitted in fulfilment of the requirement for the
degree of Doctor of Philosophy
To
Glasgow Dental School, College of Medical, Veterinary
and Life Sciences
May 2016
I dedicate this work to the people of Iraq, my beloved home.
I know how much you are suffering
I feel for you.
Anas Almukhtar 2016
I
Summary
Introduction
Prediction of soft tissue changes following orthognathic surgery has been
frequently attempted in the past decades. It has gradually progressed from the
classic “cut and paste” of photographs to the computer assisted 2D surgical
prediction planning; and finally, comprehensive 3D surgical planning was
introduced to help surgeons and patients to decide on the magnitude and
direction of surgical movements as well as the type of surgery to be considered
for the correction of facial dysmorphology.
A wealth of experience was gained and numerous published literature is
available which has augmented the knowledge of facial soft tissue behaviour and
helped to improve the ability to closely simulate facial changes following
orthognathic surgery. This was particularly noticed following the introduction of
the three dimensional imaging into the medical research and clinical
applications.
Several approaches have been considered to mathematically predict soft tissue
changes in three dimensions, following orthognathic surgery. The most common
are the Finite element model and Mass tensor Model. These were developed into
software packages which are currently used in clinical practice. In general, these
methods produce an acceptable level of prediction accuracy of soft tissue
changes following orthognathic surgery. Studies, however, have shown a limited
prediction accuracy at specific regions of the face, in particular the areas
around the lips.
Aims
The aim of this project is to conduct a comprehensive assessment of hard and
soft tissue changes following orthognathic surgery and introduce a new method
for prediction of facial soft tissue changes.
Anas Almukhtar 2016
II
Methodology
The study was carried out on the pre- and post-operative CBCT images of 100
patients who received their orthognathic surgery treatment at Glasgow dental
hospital and school, Glasgow, UK. Three groups of patients were included in the
analysis; patients who underwent Le Fort I maxillary advancement surgery;
bilateral
sagittal
split
mandibular
advancement
surgery
or
bimaxillary
advancement surgery. A generic facial mesh was used to standardise the
information obtained from individual patient’s facial image and Principal
component analysis (PCA) was applied to interpolate the correlations between
the skeletal surgical displacement and the resultant soft tissue changes. The
identified relationship between hard tissue and soft tissue was then applied on a
new set of preoperative 3D facial images and the predicted results were
compared to the actual surgical changes measured from their post-operative 3D
facial images.
A set of validation studies was conducted. To include:
Comparison between voxel based registration and surface registration to
analyse changes following orthognathic surgery. The results showed there
was no statistically significant difference between the two methods.
Voxel based registration, however, showed more reliability as it preserved
the link between the soft tissue and skeletal structures of the face during
the image registration process. Accordingly, voxel based registration was
the method of choice for superimposition of the pre- and post-operative
images. The result of this study was published in a peer refereed
scientific journal.
Direct DICOM slice landmarking; a novel technique to quantify the
direction and magnitude of skeletal surgical movements. This method
represents a new approach to quantify maxillary and mandibular surgical
displacement in three dimensions. The technique includes measuring the
distance of corresponding landmarks digitized directly on DICOM image
slices in relation to three dimensional reference planes. The accuracy of
the measurements was assessed against a set of “gold standard”
Anas Almukhtar 2016
III
measurements extracted from simulated model surgery. The results
confirmed the accuracy of the method within 0.34mm. Therefore, the
method was applied in this study. The results of this validation were
published in a peer refereed scientific journal.
The use of a generic mesh to assess soft tissue changes using
stereophotogrammetry. The generic facial mesh played a major role in
the soft tissue dense correspondence analysis. The conformed generic
mesh represented the geometrical information of the individual’s facial
mesh on which it was conformed (elastically deformed). Therefore, the
accuracy of generic mesh conformation is essential to guarantee an
accurate replica of the individual facial characteristics. The results
showed an acceptable overall mean error of the conformation of generic
mesh 1 mm. The results of this study were accepted for publication in
peer refereed scientific journal.
Skeletal tissue analysis was performed using the validated “Direct DICOM slices
landmarking method” while soft tissue analysis was performed using Dense
correspondence analysis. The analysis of soft tissue was novel and produced a
comprehensive description of facial changes in response to orthognathic surgery.
The results were accepted for publication in a peer refereed scientific journal.
The main soft tissue changes associated with Le Fort I were advancement at the
midface region combined with widening of the paranasal, upper lip and nostrils.
Minor changes were noticed at the tip of the nose and oral commissures.
The main soft tissue changes associated with mandibular advancement surgery
were advancement and downward displacement of the chin and lower lip
regions, limited widening of the lower lip and slight reversion of the lower lip
vermilion combined with minimal backward displacement of the upper lip were
recorded. Minimal changes were observed on the oral commissures.
Anas Almukhtar 2016
IV
The main soft tissue changes associated with bimaxillary advancement surgery
were generalized advancement of the middle and lower thirds of the face
combined with widening of the paranasal, upper lip and nostrils regions.
In Le Fort I cases, the correlation between the changes of the facial soft tissue
and the skeletal surgical movements was assessed using PCA. A statistical
method known as ’Leave one out cross validation’ was applied on the 30 cases
which had Le Fort I osteotomy surgical procedure to effectively utilize the data
for the prediction algorithm. The prediction accuracy of soft tissue changes
showed a mean error ranging between (0.0006mm±0.582) at the nose region to
(-0.0316mm±2.1996) at the various facial regions.
Anas Almukhtar 2016
V
Contents
DEDICATION……………………………………………………………..……………………………………………………………….…I
SUMMARY…………………………………………………………………………………………………………………………………..II
TABLE OF CONTENTS…………………………………………………….……………………………………………………………….V
LIST OF TABLES…………………………………………………………..……………………………………………………………...VII
LIST OF FIGURES…………………………………………………………………………………………………………………………VIII
ACKNOWLEDGEMENTS……………………………………………….…………………………………………………………………..X
AUTHOR’S DECLARATIONS……………..…….……………………..……………………….……………………………………….XII
LIST OF ABBREVIATIONS…………..…………………………………………………………….…………………………………….XIII
1
REVIEW OF LITERATURE ...................................................................................................... 1
1.1 3D IMAGE CAPTURE .............................................................................................................. 2
1.1.1 ACQUISITION OF 3D SURFACE DATA: ............................................................................................. 2
1.1.2 ACQUISITION OF 3D VOLUMETRIC DATA....................................................................................... 15
1.2 3D IMAGE PROCESSING ....................................................................................................... 20
1.2.1 3D MODELLING AND 3D IMAGE VISUALISATION............................................................................. 20
1.2.2 3D IMAGE SUPERIMPOSITION: ................................................................................................... 22
1.3 3D IMAGE ANALYSIS. .......................................................................................................... 25
1.3.1 SKELETAL ANALYSIS. ................................................................................................................. 25
1.3.2 SOFT TISSUE ANALYSIS .............................................................................................................. 32
1.3.3 HARD-SOFT TISSUE CORRELATIONS AND PREDICTION OF SURGICAL RESULTS ........................................ 53
1.4 3D PREDICTION OF SOFT TISSUE CHANGES FOLLOWING ORTHOGNATHIC SURGERY. ............................. 58
1.4.1 GEOMETRICAL ANALYSIS MODELS ............................................................................................... 59
1.4.2 FINITE ELEMENT MODEL ........................................................................................................... 59
1.4.3 MASS SPRING MODEL .............................................................................................................. 60
1.4.4 MASS TENSOR MODEL ............................................................................................................. 61
1.4.5 COMPARISON OF THE DIFFERENT DEFORMATION MODELS ............................................................... 63
1.5 AIMS .............................................................................................................................. 64
1.6 SPECIFIC OBJECTIVES ........................................................................................................... 64
2
METHODOLOGY................................................................................................................ 65
INTRODUCTION .......................................................................................................................... 66
2.1 SECTION A: MAIN RESEARCH SAMPLE RECRUITMENT .................................................................. 68
2.1.1 SAMPLE ................................................................................................................................. 68
2.1.2 CBCT SCANNING PROTOCOL ...................................................................................................... 69
2.2 SECTION B: VALIDATION OF THE HARD TISSUE CHANGES AS A RESULT OF SURGERY ............................. 73
2.2.1 COMPARISON BETWEEN VOXEL BASED REGISTRATION AND SURFACE REGISTRATION TO ANALYSE CHANGES
FOLLOWING ORTHOGNATHIC SURGERY..................................................................................................... 73
2.2.2 DIRECT DICOM SLICE LANDMARKING, A NOVEL TECHNIQUE TO QUANTIFY THE DIRECTION AND MAGNITUDE
OF HARD TISSUE SURGICAL CHANGE. ........................................................................................................ 89
2.3 SECTION C: VALIDATION OF BASIC METHODS OF SOFT TISSUE ANALYSIS ......................................... 107
2.3.1 THE USE OF A GENERIC MESH TO ASSESS SOFT TISSUE CHANGES USING STEREOPHOTOGRAMMETRY. ...... 107
2.4 SECTION D: ANALYSIS OF SKELETAL AND SOFT TISSUE CHANGES FOLLOWING ORTHOGNATHIC SURGERY.. 131
2.4.1 PRE- ANALYSIS 3D IMAGE PREPARATION .................................................................................... 131
2.4.2 MEASUREMENT OF HARD TISSUE DISPLACEMENT FOLLOWING SURGERY ........................................... 134
2.4.3 ANALYSIS OF SOFT TISSUE CHANGES FOLLOWING SURGERY............................................................. 140
2.4.4 SIMULATION OF SOFT TISSUE FOLLOWING ORTHOGNATHIC SURGERY ............................................... 146
Anas Almukhtar 2016
VI
3
RESULTS ......................................................................................................................... 149
3.1 ERROR STUDY................................................................................................................. 150
3.2 ANALYSIS OF SKELETAL SURGICAL MOVEMENTS ........................................................................ 155
3.3 ANALYSIS OF SOFT TISSUE CHANGES FOLLOWING SURGERY ......................................................... 156
3.3.1 SOFT TISSUE RESPONSE TO LE FORT I MAXILLARY ADVANCEMENT. .................................................. 156
3.3.2 SOFT TISSUE RESPONSE TO BSSO MANDIBULAR ADVANCEMENT. .................................................... 163
3.3.3 SOFT TISSUE CHANGES FOLLOWING BI-MAXILLARY ADVANCEMENT. ................................................. 167
3.4 PREDICTION OF FACIAL SOFT TISSUE CHANGES FOLLOWING LE FORT I ADVANCEMENT SURGERY. .......... 172
3.4.1 UPPER LIP ............................................................................................................................ 173
3.4.2 LOWER LIP ............................................................................................................................ 176
3.4.3 CHIN ................................................................................................................................... 179
3.4.4 NOSE .................................................................................................................................. 182
3.4.5 PARANASAL LEFT ................................................................................................................... 185
3.4.6 PARANASAL RIGHT ................................................................................................................. 188
4
DISCUSSION.................................................................................................................... 192
4.1 STUDY SAMPLE................................................................................................................ 193
4.2 DICOM IMAGE PROCESSING AND ANALYSIS ........................................................................... 197
4.2.1 IMAGE SUPERIMPOSITION ....................................................................................................... 197
4.3 MEASUREMENT OF SKELETAL DISPLACEMENT ......................................................................... 198
4.4 MEASUREMENT OF SOFT TISSUE CHANGES IN RESPONSE TO ORTHOGNATHIC SURGERY. ...................... 201
4.4.1 THE GENERATION OF THE “AVERAGE FACE” ................................................................................ 203
4.4.2 THE CORRESPONDING 3D FACIAL SOFT TISSUE CHANGES IN RESPONSE TO LE FORT I MAXILLARY
ADVANCEMENT. ................................................................................................................................ 205
4.4.3 THE CORRESPONDING 3D FACIAL SOFT TISSUE CHANGES IN RESPONSE TO BSSO MANDIBULAR
ADVANCEMENT. ................................................................................................................................ 210
4.4.4 THE CORRESPONDING 3D FACIAL SOFT TISSUE CHANGES IN RESPONSE TO BIMAXILLARY ADVANCEMENT. 213
4.5 SOFT TISSUE PREDICTION ................................................................................................... 218
4.5.1 MASS SPRING MODEL (MSM) ................................................................................................. 218
4.5.2 FINITE ELEMENT MODEL ( FEM) ............................................................................................... 219
4.5.3 MASS TENSOR MODEL (MTM) ................................................................................................ 221
4.5.4 MEASUREMENT OF PREDICTION ACCURACY ................................................................................ 224
5
CONCLUSIONS & SUGGESTIONS ...................................................................................... 229
5.1
5.2
5.3
CONCLUSIONS ................................................................................................................. 230
SUGGESTIONS FOR FUTURE STUDIES ...................................................................................... 231
POSSIBLE APPLICATIONS .................................................................................................... 232
6
REFERENCES ................................................................................................................... 233
7
APPENDICES ................................................................................................................... 258
7.1 APPENDIX 1 PRESENTATIONS AND AWARDS............................................................................ 259
7.1.1 VERBAL PRESENTATIONS ......................................................................................................... 259
7.1.2 POSTER PRESENTATIONS ......................................................................................................... 260
7.1.3 AWARDS .............................................................................................................................. 261
7.2 APPENDIX 2 PUBLICATIONS ................................................................................................ 262
7.2.1 PUBLISHED JOURNAL ARTICLES ................................................................................................. 263
7.2.2 ACCEPTED FOR PUBLICATION ................................................................................................... 263
Anas Almukhtar 2016
VII
List of Tables
Table 1: Types and percentage of published methods of soft tissue analysis 2000-2015............................... 35
Table 2: Definitions and abbreviations of facial soft tissue landmarks by Farkas 1980(67) ........................... 35
Table 3: Surgical jaw correction movements. ............................................................................................... 71
Table 4: Image pairs configurations ............................................................................................................. 71
Table 5: Paired sample t-test to compare methods accuracy for each tissue type. ....................................... 83
Table 6: Paired sample t-test to compare the accuracy between different tissue types. ............................... 83
Table 7: Pearson correlation analyses showing ‘correlation coefficient’ and ‘significance’ between different
tissue types and methods............................................................................................................................. 83
Table 8: Combinations of simulated surgery movements ............................................................................. 90
Table 9: The intra and inter examiner landmarking errors (Euclidian distances) between the repeated
readings at each landmark. ....................................................................................................................... 101
Table 10: Interclass correlation and the inter- and intra-examiner errors (three dimensional distance)
between the repeated readings. ................................................................................................................ 101
Table 11: The inter and intra examiner errors in the three dimensions (one sample t-test) ........................ 102
Table 12: The differences between the two methods of measurements (Inter class correlation). ............... 102
Table 13: The differences between the two methods of measurements (one sample t-test) ....................... 103
Table 14: Results of the pilot study of a repeated measurements on clinical cases (Paired sample t-test).. 103
Table 15: Definitions of landmarks for validation of the accuracy of 3D image conformation. ................... 114
Table 16 Mean Euclidean distance and standard deviation for landmarking errors for each of the 34
landmarks.................................................................................................................................................. 123
Table 17: Mean surface distance in millimeters ......................................................................................... 125
Table 18: The mean Euclidean distances (mm) of the 19 corresponding landmarks between the conformed
and original mesh for all facial expressions. ............................................................................................... 126
Table 19: Mean Euclidean distance between the corresponding landmarks for each facial expression. ..... 127
Table 20: Landmarks definitions used for the measurements of skeletal displacement.............................. 135
Table 21: Landmarks used for generic mesh conformation. ........................................................................ 139
Table 22: Landmarking error (orthogonal distance)between the repeated digitization of landmarks used for
generic mesh conformation. ...................................................................................................................... 151
Table 23: Landmarking error (Euclidean distance) between the repeated digitisations of landmarks used for
generic mesh conformation. ...................................................................................................................... 153
Table 24: Measurements (mm) of skeletal displacement following orthognathic surgery ........................... 155
Table 25: Sample size reported in previous studies. ................................................................................... 194
Anas Almukhtar 2016
VIII
List of Figures
Figure 1: Moire's photography. ...................................................................................................................... 6
Figure 2: Diagram showing the basics of stereophotogrammetry (triangulation effect) ................................. 6
Figure 3: Anaglyph photography .................................................................................................................... 7
Figure 4: Active stereophotogrammetry....................................................................................................... 10
Figure 5: Passive stereophotogrammetry. .................................................................................................... 10
Figure 6: 3D system for active stereophotogrammetry................................................................................. 11
Figure 7: Comparison between different 3D imaging systems. .................................................................... 12
Figure 8: Marching cube algorithm. ............................................................................................................. 23
Figure 9: Mathematically generated landmarks........................................................................................... 42
Figure 10: Landmarks plotting .................................................................................................................... 42
Figure 11: Colour coded map. ...................................................................................................................... 47
Figure 12: Types of absolute distance measurements .................................................................................. 47
Figure 13: Dense correspondance surface analysis. ...................................................................................... 51
Figure 14: Basic element of the finite element model. .................................................................................. 62
Figure 15: Basic elements of the mass-spring model. ................................................................................... 62
Figure 16: CBCT image defect. A step at the face was formed due to patient movement during the CBCT. ... 72
Figure 17: CBCT image capture setup........................................................................................................... 72
Figure 18: Voxel based registration. ............................................................................................................ 77
Figure 19: The registration template. .......................................................................................................... 78
Figure 20: Standard region for analysis. ....................................................................................................... 81
Figure 21: Output colour map -registration accuracy ................................................................................... 81
Figure 22: Comparison of the accuracy of the two 3D image registration methods. ..................................... 87
Figure 23: Excluded cases. ........................................................................................................................... 87
Figure 24: 3D Surgery simulation and measurement setup .......................................................................... 93
Figure 25: Reference planes (Ondemand3D software). ................................................................................ 96
Figure 26: Landmarks digitization. The full set of landmarks ....................................................................... 98
Figure 27: Bland Altman Plot for sagittal measurements. .......................................................................... 100
Figure 28: Bland Altman vertical measurements.………………………………………………………………………………………100
Figure 29: DI3D Stereophotorgammetry system......................................................................................... 110
Figure 30: DI3D Capture software main panel ............................................................................................ 110
Figure 31: DI3D View software main panel ................................................................................................ 111
Figure 32: DI3D system calibration board .................................................................................................. 111
Figure 33: Full set of landmarks indicators placed on participant's face ..................................................... 115
Figure 34: The six Facial expressions .......................................................................................................... 116
Figure 35: The conformation software. ...................................................................................................... 118
Figure 36: Voxel based registration. ........................................................................................................... 133
Figure 37: Segmented 3D models from the DICOM image using Maxilim software .................................... 136
Figure 38: Reference planes ....................................................................................................................... 136
Figure 39: The generic mesh ...................................................................................................................... 143
Figure 40: Steps of generic mesh conformation.......................................................................................... 144
Figure 41: Soft tissue changes in the three dimensions............................................................................... 147
Figure 42: Dense anatomical correspondence (Euclidean) for Le Fort I maxillary advancement ............. 158
Figure 43: Corresponding soft tissue changes (directional) for Le Fort I maxillary advancement . ............... 160
Figure 44: Dense anatomical correspondence (Euclidean) for BSSO mandibular advancement................... 163
Figure 45: Corresponding soft tissue changes (directional) for BSSO mandibular advancement ................. 165
Figure 46: Dense anatomical correspondence (Euclidean) for bimaxillary advancement ............................ 167
Figure 47: Corresponding soft tissue changes (directional) for bimaxillary advancement ........................... 170
Figure 48: Accuracy of the prediction at the upper lip region (Box plot)...................................................... 174
Figure 49: Accuracy of the prediction at the upper lip region (directional colour map) ............................... 175
Figure 50: Accuracy of the prediction at the lower lip region (Box plot) ..................................................... 177
Figure 51: Accuracy of the prediction at the lower lip region (directional colour map) ............................... 178
Figure 52: Accuracy of the prediction at the chin region (Box plot) ............................................................. 180
Figure 53: Accuracy of the prediction at the chin region (directional colour map) ...................................... 181
Figure 54: Accuracy of the prediction at the nose region ( Box plot). ......................................................... 183
Figure 55: Accuracy of the prediction at the nose region (directional colour map)...................................... 184
Figure 56: Accuracy of the prediction at the left paranasal region (Box plot).............................................. 186
Figure 57: Accuracy of the prediction at the left paranasal region (directional colour map) ....................... 187
Figure 58: Accuracy of the prediction at the right paranasal region (Box plot) ........................................... 189
Anas Almukhtar 2016
IX
Figure 59: Accuracy of the prediction at the right paranasal region (directional colour map) ..................... 190
Figure 60: Difference between the 3D model and the radiographic shadow in representing the contour of
the greater palatine foramina.................................................................................................................... 200
Figure 61: Comparison between classical colour map and the corresponding analysis for soft tissue analysis
following Li Fort I maxillary advancement ................................................................................................. 206
Figure 62: Anatomy of mid-face muscle( quoted from Grant's atlas of anatomy) ....................................... 210
Figure 63: Comparison between classical colour map and the corresponding analysis for soft tissue analysis
foloowing mandibular advancement ......................................................................................................... 211
Figure 64: Anatomy of mandibular muscles (quoted from Grant’s atlas of anatomy) ................................. 213
Figure 65: Comparison between classical colour map and the corresponding analysis for soft tissue analysis
following bimaxillary advancement. .......................................................................................................... 214
Anas Almukhtar 2016
X
Acknowledgements
"My Lord, indeed I am, for whatever good You would send down to me,
in need" (Al_Qasas,28). Thanks to Allah, the almighty, for giving me the
willingness and strength to complete this work.
I wish to express my most sincere gratitude and appreciation to my principal
supervisor Professor A. Ayoub for his guidance, patience and encouragement
throughout the development of the project. One simply could not wish for a
better or friendlier supervisor.
My deepest gratitude also goes to my co-supervisors Professor B Khambay, not
only for his bright ideas, which influenced the project’s methodology in many
aspects, but also for his persistent demand for perfection which led to the
quality of this project; Dr. X Ju, for his remarkable software development skills
without which this project would never see the light; and last but not the least,
Professor J McDonald whose never-failing sympathy and support helped me
passing the darkest parts in my study path.
I would like to express my appreciation to the Glasgow Dental School
administration especially the Dean, Prof J Bagg, who offered his unlimited
support when I was desperately in need. A word of thanks is also spoken to our
brilliant secretary, Liz, for managing administrative issues with a high precision.
My sincere gratitude and appreciation are to my financial sponsors, the higher
committee for education and development in Iraq (HCED), for their dedication to
improve the quality of higher education for a better future.
Many thanks for all those who contributed to the success of my study; Mr
Benington, our brilliant orthodontics consultant and Carol, our lively dental
nurse, who provided all the clinical supervision and technical support.
Anas Almukhtar 2016
XI
The only way to overcome the pain and the agony of the PhD was sharing it with
sincere friends. Especially mentioning Mohammed, Ali, Jamie and Faith. The
warmth of their caring and support kept me contained and focused on my goals.
My parents watched me from a distance while I worked towards my degree,
sacrificed their comfort to let me find mine and hid this from me so I wouldn’t
feel any regret. Without their love, affection and encouragement this work
would not have been possible.
No words can express my thanks and appreciation for my wife. The success in
this project was totally attributed to her endless love and support. Thanks for
every late night I returned home and every early morning I left, thanks for
bearing my ridiculousness at every furious outburst, thanks for maintaining our
family while I was busy in my agonizing study. Your sacrifices are well known and
very well appreciated. My children Mustafa and Maryam are the joy of my life.
They only have ever known me as a student. Despite only seeing them at the
weekends and sometimes not, they have never complained. Good news for
them, “you have your father back”.
There are many more people I could thank, but time, space, and modesty
compel me to stop here.
Thank you
Anas
Anas Almukhtar 2016
XII
Author's Declaration
I declare that, except where explicit reference is made to the contribution of
others, that this thesis is the result of my own work and has not been submitted
for any other degree at the University of Glasgow or any other institution.
Signature
Printed name: Anas Almukhtar
October 2015
Anas Almukhtar 2016
XIII
List of Abbreviations
Abbreviations
Full name
3D
Three dimensions
DICOM
Digital Imaging Communications in Medicine
BSSO
Bilateral Sagittal Split Osteotomy
VBR
Voxel based registration
SBR
Surface Based Registration
SD
Standard deviation
SE
Standard Error
FEM
Finite Element Model
MSM
Mass Spring Model
MTM
Mass Tensor Model
PCA
Principal Component Analysis
ICP
Iterative Closest Point
PA
Procrustes Analysis
PPA
Partial Procrustes Analysis
IC
Interclass Correlation
Sig
Significance level
CBCT
Cone Beam Computerized Tomography
CT
Computerized Tomography
MRI
Magnetic Resonance Imaging
HU
Hounsfield Unit
IF*
Incisive Foramen
*Abbreviations for landmarks are defined in their relevant chapters.
Anas Almukhtar 2016
XIV
1
C
Review of literature
ontents
1.1
3D IMAGE CAPTURE .............................................................................................................. 2
1.1.1
ACQUISITION OF 3D SURFACE DATA: ............................................................................................. 2
1.1.2
ACQUISITION OF 3D VOLUMETRIC DATA....................................................................................... 15
1.2
3D IMAGE PROCESSING ....................................................................................................... 20
1.2.1
3D MODELLING AND 3D IMAGE VISUALISATION............................................................................. 20
1.2.2
3D IMAGE SUPERIMPOSITION: ................................................................................................... 22
1.3
3D IMAGE ANALYSIS. .......................................................................................................... 25
1.3.1
SKELETAL ANALYSIS. ................................................................................................................. 25
1.3.2
SOFT TISSUE ANALYSIS .............................................................................................................. 32
1.3.3
HARD-SOFT TISSUE CORRELATIONS AND PREDICTION OF SURGICAL RESULTS ........................................ 53
1.4
3D PREDICTION OF SOFT TISSUE CHANGES FOLLOWING ORTHOGNATHIC SURGERY. ............................. 58
1.4.1
GEOMETRICAL ANALYSIS MODELS ............................................................................................... 59
1.4.2
FINITE ELEMENT MODEL ........................................................................................................... 59
1.4.3
MASS SPRING MODEL ............................................................................................................. 60
1.4.4
MASS TENSOR MODEL ............................................................................................................. 61
1.4.5
COMPARISON OF THE DIFFERENT DEFORMATION MODELS ............................................................... 63
1.5
AIMS .............................................................................................................................. 64
1.6
SPECIFIC OBJECTIVES ........................................................................................................... 64
Anas Almukhtar 2016
1
Chapter One
Review of Literature
Introduction
The innovation in three dimensional (3D) imaging technologies allowed the
capture of a patient’s face for a comprehensive diagnosis and treatment
planning. These technologies succeeded in producing what is known today as the
“Virtual patient”, a new term introduced to the medical field which describes
the creation of a digital copy of the patient’s body on a computer screen where
various diagnostic and virtual surgeries can be performed. Orthognathic surgical
analysis and planning benefited from the availability of 3D imagining technology
which promises improvement in the diagnosis and treatment planning as well as
enriching the information obtained for facial anthropometry studies.
1.1 3D image capture
In general, the available 3D data of the face are either surface images captured
by imaging facilities that do not penetrate through the facial skin such as
Stereophotogrammetry and laser scanners, or volumetric images captured by
imaging facilities that penetrate through human tissue such as CT scans , MRI
and 3D ultrasound probing.
1.1.1
Acquisition of 3D Surface data:
Systems for surface data acquisition are capable of capturing the external
surface of the face and do not record internal or skeletal structures. These
include 3D photography and 3D laser scanners which have a variety of clinical
applications including evaluation of orthognathic surgery outcome, evaluation of
the changes following cleft lip repair, study of growth pattern, the variation of
facial features among various populations and creating a mean face for the
population.
1.1.1.1 Three dimensional (3D) photography
Four main photographic techniques were introduced to capture the 3D
morphology of human face; Three-dimensional Facial Morphometry (3DFM);
Contour photography; Moire’s topography; and Stereophotogrammetry. Of these
Anas Almukhtar 2016
2
Chapter One
Review of Literature
four methods, Stereophotogrammetry has been the most successful in producing
an accurate and reasonably practical method of facial 3D imaging.
1.1.1.2 Three-dimensional Facial Morphometry (3DFM)
Moorrees and Lebert, 1962 introduced the computerized mesh diagram
analysis(1). Their idea was to capture the face using a pair of normal lateral and
frontal photographs. A matrix of intersecting lines was then plotted on the
photograph on which a set of landmarks were digitized. Based on the
deformation of this matrix due to the shift from the frontal to the lateral
photograph, the vectors for each landmark were obtained which provided the
three dimensional orientation of each of them.
The basic purpose of this
approach was not to capture the face in 3D but only for the analysis of the three
dimensional position of each landmark (Moorrees et al., 1976(2).
Ferrario et al., 1998 (3) introduced a computerized mesh diagram analysis and
studied its application in soft tissue facial morphometry; the aim of their study
was to create norms from the vector distortion matrix and to compare the
difference in facial features between male and female.
1.1.1.3 3D Contour photography
E. J. Lovesey 1974 (4) introduced this method
for facial anthropometric
measurements. The technique was simple; a special continuous pattern of black
and white lines or squares was projected on patients face. Frontal and lateral 2D
photographs are captured and the changes of the contours of the pattern’s lines
were studied. Areas of rapid change of contour like the eyes and nose
represented a major problem for this technique. Measurement of facial features
was subjective and not accurate enough for a robust facial anthropometric
analysis.
1.1.1.4 Moire’s topography
The method is based on the projection of lines (fringes) on the face at the time
of image capture; the distortion of the line on the facial features is detected by
Anas Almukhtar 2016
3
Chapter One
Review of Literature
the software which is designed to quantify fringes distortion and to create a
three dimensional image. The technique however is sensitive to facial
positioning especially in the evaluation of the differences between pre-surgical
and post-surgical images, the change in head position would affect the pattern
of lines distortion on the face. Another limitation of this method is the potential
error in recording rough surfaces.
Motoyoshi et al., 1992 (5) introduced one of the earliest methods of capturing
the human face in 3D for objective anthropometric analysis using a cathode-ray
tube (CRT) technique. Their technique was based on Moire’s photography and
the subject was photographed while projecting about 800 points of light onto
the face using a projector and a set of three cameras (central, right and left
sides).
The three photographs obtained were then transmitted to a computer as
pictorial image data by an image scanner. The 2D image from each camera was
recognized by the computer as a network of projected points. The 3d
coordinates of each point was derived from their position into the captured set
of photographs taken from the front and from each of the cameras at 45 °
angles, (figure 1).
The precision of the method was tested by imaging a facial plaster cast. The
three-dimensional coordinates of the anatomical points were compared with
those generated from an electromagnetic digitizer. The mean of errors and the
standard deviation were less than 0.08 ± 0.23 mm in all dimensions (5).
1.1.1.5 3D Static Stereophotogrammetry
Stereophotogrammetry refers to the special photography technique where at
least two cameras are configured as a stereo-pair to recover the depth of the 3D
object by interpolating the third dimension of each point on the 3D object
surface being imaged. The degree of disparity between the two dimensional
location of the corresponding points on both images generates the third
dimension.
Stereophotogrammetry is based on the triangulation effect, (figure
Anas Almukhtar 2016
4
Chapter One
Review of Literature
2). An imaginary triangle represents the two cameras and the object to be
imaged. The distance between the two cameras is fixed while the distance of
each camera to the point on the object to be imaged varies according to their
distance from the captured point; this difference is described as the degree of
disparity. The (X1,Y1) and (X2,Y2) coordinates of the captured point within the
right and left cameras images produce the z coordinate of that point using a
mathematical formula, Equation (1).
Equation 1: Stereophotogrammetry disparity equation. Where the denominator x2 − x1 is
known as the stereo disparity, d is the distance between the two cameras and f is the focal
length of the camera lens, (6) Please refer to figure (2).
In cases where two pairs of cameras were used to capture the face, two 3D
images are produced, one for the right half of the face and the other for the left
side of the face. A software programme is used to link the two stereo-images to
produce the full 3D image of the face. This method has been clinically validated
for its level of accuracy (7). The algorithms used to match the images from the
paired sources evolved through different stages of development.
Burke and
Beard 1967 in Cambridge, UK (8) were the first to describe the concept of stereo
3D imaging of the face for medical purpose. Their method was not based on
triangulation effect but on image stereoscopy (anaglyph photography).
Two
cameras were used adjacent to each other. The patient was lying in a supine
position and the cameras hanging over the head. Alternative red and blue layers
of transparent sheets were applied, these were placed in the front of the right
and left camera respectively, (figure 3). By careful superimposition of the right
and left image the viewer detected the depth of the image. The method does
not produce 3D coordinates, so may not be ideal for facial analysis. However this
system was adopted for multimedia and gaming industry.
Anas Almukhtar 2016
5
Chapter One
Review of Literature
Figure 1: Moire's photography: Diagram showing the technique of using three
cameras at 45º angle from each other combined with the fringes projector.
Figure 2: Diagram showing the basics of stereophotogrammetry (triangulation effect).
The depth of each pixel could be calculated from the known distance between the two
cameras d; the cameras focal length f; the 2D coordinates of the pixel (x and y)
utilising the disparity between captures (from each camera) in Equation (1). Image
quoted from (6).
Anas Almukhtar 2016
6
Chapter One
Review of Literature
Figure 3: Anaglyph photography: The use of two colour filters to produce visual depth
perception. Burke and Beard 1967 (8)
In general, there are two types for stereophotogrammetry: active and passive
stereophotogrammetry, and a combination of the two known as hybrid. Active
stereophotogrammetry, (figure 4), is based on utilization of a structured light to
produce the 3D image. Specific pattern projects onto the patient’s face while
two (or more) cameras are used to capture the information provided by the
deformation of the pattern on the patients face from different viewpoints.
Monochrome cameras are used for easier detection of the deformation pattern.
The addition of a coloured camera will facilitate the capture of the facial
texture.
The 3D surface model is generated by applying the triangulation principle
extracted from the stereo images; this is accomplished by calculating the third
dimensional coordinate of each 2D point (pixel) visible in both camera images.
The purpose of the projected textured pattern on the object to be captured is to
simplify the automatic detection of corresponding points to interpolate the third
dimension.
On the other hand, passive stereophotogrammetry, (figure 5), does not depend
on pattern projection on the surface to generate the 3D image. The process of
finding correspondences between images is more complex and high-quality
cameras are necessary to capture surface details and to provide sufficient
texture information to act as natural patterns to build the 3D image. Skin pores,
Anas Almukhtar 2016
7
Chapter One
Review of Literature
freckles and scars would act as a distinctive pattern to help find the
correspondences and generate the 3rd dimension of the captured image. The
lighting conditions must be carefully controlled as strong ambient light may
result
in
glare
thus
diminishing
the
surface
details
(9). Hybrid
stereophotogrammetry combines the advantages of both active and passive to
achieve a better quality of the produced 3D imaging.
Ras et al., 1996 (10) introduced the active stereophotogrammetry concept for
facial analysis using two stereo coupled cameras at 50 cm distance from each
other and a flash spot fitted between the cameras to project a dotted pattern
on the face.
Hajeer et al., 2003 (11) used stereophotogrammetry imaging to test the
reliability of the stereophotogrammetry-based 3D imaging system (C3D) and to
determine the effect of orthognathic surgery on the 3D soft-tissue morphology.
The imaging setup consisted of one colour digital camera to capture the natural
appearance of the face; two monochrome digital cameras serving as a 'stereopair' for building the 3D model; one white light flash synchronised to operate
with the colour camera and one speckle texture projection synchronised to
operate with the monochrome cameras, (figure 6).
The use of the monochrome cameras helps to reduce the image noise and
facilitate 3D model build-up whereas the speckle texture projection pattern
provide the necessary information for the software to build 3D facial images.
System accuracy was validated and an error of less than 0.3 mm ± 0.13 mm in all
dimensions was reported.
In order to obtain accurate 3D images, calibration of the software is necessary,
ideally before each session of imaging, however one calibration at the start of
the day may be sufficient to ensure a satisfactory consistency throughout that
day. Calibration is the digital refinement for the software to detect the amount
of disparity of a set of points on a flat surface captured, at least four times,
from different angles.
Anas Almukhtar 2016
8
Chapter One
Review of Literature
Despite the use of various techniques to facilitate 3D image production, the
main challenge in stereophotogrammetry is to establish which pixels from the
left and right images correspond to the same point in 3D space. A common
approach
is
to apply
maps
of
the
stereo
disparities
based
on
the
correspondences provided by each image, and from this map, the surface of the
object is built on the basis of triangulation.
The most common stereophotogrammetry systems are the 3dMD and Di3D.
Although they are both based on stereo imaging, the algorithm and imaging
techniques
are
different
in
various
aspects.
Di3D
employ
passive
stereophotogrammetry, which generates 3D surface images solely on the basis of
natural patterns, such as skin pores, freckles, scars and so forth. Therefore, the
3D reconstruction depends on the integrity of the pixels and requires highresolution cameras.
3dMD
combines
active
and
passive
stereophotogrammetry
triangulation
strategies into its systems called ‘hybrid’ stereophotogrammetry. The cameras
are based on industrial vision standards containing sensors of higher quality and
consistency than (SLR) cameras.
Tzou et al., 2014 (9) compared the clinical applicability of five 3D imaging
systems. DI3D and 3dMD were the stereophotogrammetry imaging systems
involved in this study. The two systems were similar in many aspects. However a
few differences were seen regarding the speed of processing and type of
capturing technique, (figure 7).
Di3D was superior to 3dMD in its simplicity of the system as it uses high
resolution cameras and flash lights normally used for photography with no need
for the industrial sensors used by 3dMD.
Anas Almukhtar 2016
9
Chapter One
Review of Literature
Figure 4: Active stereophotogrammetry. A diagram showing how the projected pattern
helped in the identification of the corresponding pixels in order to build the disparity map
(9).
Figure 5: Passive stereophotogrammetry. A diagram showing how the high resolution
cameras helped in the identification of the corresponding pixels in order to build the
disparity map (9).
Anas Almukhtar 2016
10
Chapter One
Review of Literature
A
B
Figure 6: D3D system for active stereophotogrammetry. A diagram (left side) shows the
cameras panel in relation to the patient (A), an image (right side) showing the D3D system
(B). Hajeer et al., 2006 (11)
On the other hand, 3dMD provide a relatively wider capture angle and higher
processing speed. Despite these differences, the two systems were comparable
in the accuracy and quality of captured texture images. According to the author,
3dMD was the better system (9).
Khambay et al., 2008 (12) carried out an experimental study to assess the
accuracy of Di3D system image capture as well as the errors associated with 3D
image landmarking used for facial anthropometry. Operator error was measured
by
repeatedly
locating
landmarks
on
the
three-dimensional
image.
Reproducibility error of the images was calculated by capturing threedimensional images of the facial casts on two separate occasions; the Euclidean
distance between the two matched sets of coordinates was then calculated. The
Di3D system error was assessed by calculating the three-dimensional global
positions of landmarks on the three-dimensional images and comparing them
with those obtained by a coordinate measurement machine CMM (gold standard).
The results showed that the operator error in placement of landmarks on the
three-dimensional model was 0.07 mm, range 0.02–0.11 mm. The reproducibility
of the Di3D capture was 0.13 mm, range 0.11–0.14 mm. The mean distance
between the CMM and Di3D landmarks, which constitutes the Di3D system error,
was an average of 0.21 mm, range 0.14–0.32 mm.
Anas Almukhtar 2016
11
Chapter One
Review of Literature
Figure 7: Comparison between different 3D imaging systems. DI3D system (right side) and
3dMD (left side) showed the highest performance level. (9)
Anas Almukhtar 2016
12
Chapter One
Review of Literature
1.1.1.6 3D motion stereophotogrammetry (4D)
3D motion capture systems have been used for objective measurements of facial
dynamics since early in the 21st century. Various systems have been introduced
for body motion detection and analysis (13–16). The method is based on two
main principles, the direct (markers based) motion detection (13) and indirect
(marker free) motion detection systems (15,16). Marker based systems rely on
the positions of special markers placed on specific areas to track the motion.
More than one pair of cameras is required to capture the infrared signals from
active emission markers or passive reflective markers attached. Vicon TM, C3D
system
TM
, Microsoft KinectTM and Motion Analysis
TM
are common brands for
markers based motion detection systems. These systems are capable of accurate
tracking of the captured information (13). However, the placement of the
markers on patient’s faces adds to the complexity of the procedure (14).
On the other hand, indirect 3D motion detection (marker-free systems) is based
on a non-contact motion analysis using video recordings. An infrared projector is
used to project an infrared dots pattern on the subject’s face, while he/she
animates (15,16).
Currently, the two commercially available marker-free 3D motion analysis
systems are the 4D capture system (Di4D
and 3dMDface
TM
TM
Dimensional imaging, Glasgow, UK)
Dynamic system (3dMD, Atlanta, GA).
Di4D capture system is based on passive stereophotogrammetry which depends
on surface texture to build the 3D facial model. An indirect motion detection
system utilized a pair of mono chrome video recording cameras and a single
coloured video recording camera. The mono chrome cameras were used to
enhance the accuracy of the capture and to avoid noise from the coloured
texture information while a normal white light was used for the video capture
where each of the cameras simultaneously captures 60 frames per second.
Landmarks are digitized on the first frame of the captured image and the
software automatically tracks individually placed landmarks throughout the
sequence of the captured frame during performing facial expressions.
Anas Almukhtar 2016
The
13
Chapter One
Review of Literature
accuracy of the landmarks tracking has been validated by AL-Anezi et al., 2013
(17) and clinically applied in various orofacial dynamics studies (18,19).
3dMD system, on the other hand, is based on active stereophotogrammetry using
random infrared speckle projection. The latest version uses two pods with three
cameras each (two grey scale and one coloured cameras) with infrared speckle
projector in addition to normal white light. The system is capable of capturing
still images and a video at up to 60 frames per second. The infrared projection
helps increase the accuracy of triangulation and the output image is a
continuous point cloud that theoretically eliminates the error associated with
stereo image stitching (20). However, this may impact on the accuracy of the
automatic tracking of facial landmarks throughout the captured frames of facial
expressions.
1.1.1.7 3D laser facial scan
The development of laser scanning techniques provides a sophisticated method
for capturing the orofacial region in three dimensions. It has been used for
evaluation of surgical outcomes and stability of the results (21,22). High
resolution facial imaging could be achieved using laser scanning; up to 20,000
coordinates could be generated on the facial surface. 3D laser scanning has been
validated for its accuracy and reproducibility (23,24). The results showed an
error of less than 0.56 mm in all scans.
Traditional laser scanning has the shortcomings of slow capture of the face, it
takes 8- 10 seconds to scan the face, so any change in the patient's head position
or facial expression during scanning will distort the scanned image; the patient's
eyes should be closed during scanning for protection; soft-tissue surface texture
is not captured, which results in difficulties in identification of some landmarks
which are dependent on surface colour (11). Recent advancements such as using
white-light laser approaches (surface texture colour), using eye-safe laser
scanning which allows a safer scan and reduction of scanning time have
enhanced the popularity of this approach. Laser scanning of the face has been
considered in numerous studies including the quantification of facial symmetry
Anas Almukhtar 2016
14
Chapter One
Review of Literature
(25,26), evaluation of gender related facial dimorphism (27), study of facial
morphology among different populations (28), evaluation of growth changes in
facial morphology (29) and assessment of surgical outcome (30,31).
1.1.1.8 Electromagnetic 3D digitizer
This is not an imaging tool but a method to produce a set of 3D coordinates of
selected landmarks on a 3D object which is accomplished by using a flexible arm
to carefully digitize the positions of landmarks directly on patients’ faces
(32,33). The advantages of this method are that it is non-invasive and landmark
identification is usually assisted by palpation of the underlying bony structures,
such as gonial angle, for better accuracy (32,34). The outcome is a set of 3D
landmark coordinates that can be used for further analysis.
1.1.2 Acquisition of 3D volumetric data
These are the imaging tools that can penetrate through the surface to capture
internal structures in addition to the surface image. Traditionally facial
cephalometric radiography was the main tool to evaluate facial skeletal
structures. At the late quadrant of the past century, the introduction of 3D
radiography
such
as
computerized
tomography
(CT)
and
Cone
Beam
Computerized Tomography scan (CBCT) led the revolutionary change of the
assessment of facial morphology, and later on the development of Magnetic
Resonance Imaging (MRI) and 3D Ultrasonography took place to augment the
amount of information provided for the whole body, and in particular, for the
craniofacial region.
1.1.2.1 Computed tomography (CT)
Computed tomography (CT) is a highly specialized method of tomography and
provides anatomical image slices through the body. This was developed in 1972
by Sir Godfrey Hounsfield and Allan McLeod Cormack and was manufactured by
EMI (35). CT utilizes the variation of the attenuation of the x-Ray beam
traversing various body tissues to map an image of grey scale which represents
the shape and position of internal body structures. CT depends on multiple
Anas Almukhtar 2016
15
Chapter One
Review of Literature
sensors which are placed opposite to the x-ray beam source on a rotating gantry
where both rotate around the body of the patient at the area to be scanned.
Data from the sensors are transmitted to the computer to be processed into
multiple axial 2D slices building the three dimensional image structures for the
scanned volume (35).
In order to interpret the output data from CT an image matrix should be
created. This matrix is composed of a sequence of numbers arranged in columns
and rows, each square of the matrix is referred to as a picture element, and
when arranged in a matrix these elements are called pixels. The pixels are
arranged into a slice composing a single two dimensional slice image, and when
combining multiple slices, the thickness of the slice creates an additional
volume element or voxel, and each voxel has a CT number or Hounsfield Unit
(HU). A HU is the number that is allocated to each individual voxel within the
image and is displayed on the monitor as a level of brightness. HU value is
dependent on the relative comparison of an x-ray attenuation coefficient of the
tissue that is present within the voxel compared to an equal volume of water.
Water is used as the reference material as it has a uniform density and is
abundant within the human body; It is assigned a HU value of zero. Molecular
structures that have more density than water are assigned with a positive HU
value and structures that have less density are assigned with a negative HU
value. The range of values varies from -1000 for air to +4000 for metals, (35).
Computed tomography has been widely used in the field of dentistry providing
cross-sectional implant imaging (36), for evaluating a pathology in the
maxillofacial region (37), in planning for craniofacial reconstruction (38), and for
predictive surgical planning (39,40). CT allows the user to access the internal
structures of bone and teeth in a virtual environment. When displayed in a 3D
format it can provide valuable information as the images provide clear
information about the recorded anatomical structure in a variety of directions as
well as cross-sections (41–43).
Anas Almukhtar 2016
16
Chapter One
Review of Literature
1.1.2.2 Cone beam computed tomography (CBCT)
Cone beam computed tomography (CBCT) is relatively recent introduction and
has similar features to that of CT. CBCT was specifically designed for use in the
maxillofacial region for the visualisation of hard tissue and has gained broad
acceptance in dentistry in the last 5 years (37). CBCT differs from conventional
CT in many aspects:
1. While the x-ray source in CT is generated from a high output rotating
anode, CBCT can use a relatively low energy fixed anode tube comparable
to dental panoramic x-ray machines.
2. CT scanners image patients in a series of axial plane slices that can be
either stacked or form a continuous spiral motion, while CBCT captures
the image in one 360° rotation. The maxillofacial region can be captured
as a single image or the field to be imaged can be specified to capture a
limited regional area of interest (38).
The CBCT scanner includes a rotating gantry to which an x-ray source or
projector and detector are attached opposite to each other around patient’s
head. The x-ray beam projects as a cone-shaped beam through the centre of the
area of interest and the image is recorded on a flat panel detector on the
opposing side. The gantry rotates once around the region of interest projecting
and capturing 150 to 600 sequential planer projection images of the field of
view. The projection images pass through a complex algorithm where the x-ray
beam attenuation is calculated for each volume unit (voxel). This procedure is
normally completed within 10-30 seconds (44). The developed image files are in
a DICOM format (Digital Imaging and Communications in Medicine); this is the
universal format for 3D images in the medical field.
The dose of radiation is significantly reduced compared to the conventional CT
and it is approximately equal to a full mouth periapical series (45). This
depends on the setting and the model of CBCT device being used which ranges
from 29 Sv to 477 Sv compared with the conventional CT output of 2000 Sv (44).
Anas Almukhtar 2016
17
Chapter One
Review of Literature
Some CBCT scanners enable facial capture while the patient is seated in an
upright position. This is particularly beneficial when the image is used for
orthognathic surgery planning.
CBCT has some other minor advantages compared to normal CT, these include a
significant reduction in artefacts created by patient movement as the images are
captured in one single rotation of the device (44); CBCT devices are significantly
smaller and less expensive than conventional CT (46); and the isotropic
resolution is relatively high ranging from 0.076 mm – 0.4 mm which makes the
CBCT accurate enough to be used in craniofacial imaging. On the other hand the
contrast resolution which can display soft tissue information is clearer in
conventional CT scans compared to that of CBCT (37).
When imaging patients using CBCT, any intra-oral metallic objects (e.g.
restorations, jewellery, implants and orthodontic appliances) create streak
artefacts (43,47). This is to some extent a common problem shared with the CT
and CBCT scanners alike. A few published research papers have proposed
solutions for this problem (42,43). However, none of them were satisfactory
enough to be considered clinically.
1.1.2.3 Magnetic resonance imaging (MRI)
Magnetic resonance imaging (MRI) was first introduced in July 1977 (48). The
principles of MRI differ greatly from CT, CBCT and conventional radiography as
no ionizing radiation (x-rays) is required to generate the MRI image (49,50). MRI
is generated from electromagnetic energy produced by a powerful magnet. By
generation of a powerful magnetic field the hydrogen atoms within the body are
aligned. Radio waves are then transmitted to alter this alignment resulting in
the hydrogen emitting a weak radio signal which is amplified by the scanner
(51). MRI is currently regarded as the gold standard for the imaging of soft
tissue, but it is of limited value for the imaging of hard tissues (48). MRI has
equal resolution but much greater soft tissue contrast than CT scanning; this
allowes a better visualisation of the soft tissues (50).
Anas Almukhtar 2016
18
Chapter One
Review of Literature
A few drawbacks with MRI scanners have been reported: it is relatively time
consuming; equipment is expensive; it is noisy and some patients fee
claustrophobic while in the gantry tube during image capture (50); MRI is
contraindicated in patients who have a considerable amount of implanted
metallic devices (48), metallic dental restorations, and orthodontic brackets as
these materials can produce image distortion (52,53). In addition to that, MRI
does not provide the natural photographic appearance of the texture of the
facial surface(52).
1.1.2.4 Ultrasonography
Ultrasonography is an imaging technique that captures the reflection of
transmitted pulses of ultra sound. These pulses are emitted from a probe in a
direct connection to a patient’s skin, and a special conductive gel is needed to
improve the contact between the ultrasound probe and the skin. When the ultra
sound wave crosses the junction between two tissues with different densities
part of the wave is reflected back and detected as an echo by the probe (54).
The depth of tissue is calculated from the duration of time for the echo to be
detected. Three dimensional images are created by acquiring multiple cross
sectional 2D images.
3D ultrasonography is a relatively new 3D imaging technique and has many
advantages such as the relatively low cost of the equipment and the absence of
ionizing radiation exposure. Modern development of high resolution ultrasonic
imaging saw the introduction of the Colour Doppler Ultrasonography (CDUS) (55)
and Power Doppler Ultrasonography (56) with their promising ability of
differential diagnoses of malignant tumours.
Ultrasonography
in
maxillofacial
surgery
is
limited
for
diagnoses
of
temporomandibular disorders (57,58). The use of ultrasonography as an imaging
technique for maxillofacial surgical planning is still at an experimental stage,
and there are major problems associated with data acquisition, reduction and
storage (59).
Anas Almukhtar 2016
19
Chapter One
Review of Literature
The procedure is time consuming and requires a compliant patient as well as a
highly skilled operator (60). Ultrasonography is not able to visualise bone
abnormalities (49). Due to the need for direct skin contact, probe touching and
depression of the patients skin may cause distortions of their spatial positions
(60).
1.2 3D image processing
1.2.1 3D modelling and 3D image visualisation
3D modelling in the medical field has an increasing importance in diagnosis,
simulation, and treatment planning. Most of the 3D models used in various
medical applications are rendered from 3D images acquired using one of the
multi-range scanners, most commonly CT and MRI. CT scanning detects the x-ray
beam attenuation coefficient and creates a multi-slice volume image which
contains the information about the internal body structure within the scanned
volume in the form of grey scale gradient. Internal structures of close intensities
such as muscles, nerves and blood vessels are not well differentiated using this
method.
However, skeletal tissue surface and soft tissue surface are easily
identified and segmented from the image data. MRI detects the Hydrogen nuclei
resonance in the magnetic field and internal structures such as nerves and blood
vessels are clearly identifiable. However, skeletal structures are not well
represented due to low water content (48).
The 3D image volume is divided into isotropic cubes which are trapped between
image slices, and each of them is called a “voxel”. The size of the voxel is
governed by image resolution which is represented as the number of the slices
generated from the scanned volume. Each voxel has 8 vertices at the corners
(four on each successive slice). Voxels are provided with a grey scale intensity
level which represents the anatomical structure captured in that volume. Three
dimensional (3D) models segmented from these types of images are usually
rendered in 3D polygonal mesh form. Therefore, it has the capability of
representing details of biological objects’ surface topography in addition to the
selective mesh density option which allows rendering high density polygonal
meshes at areas with detailed anatomical surface features, such as areas around
Anas Almukhtar 2016
20
Chapter One
Review of Literature
the eyes and nose in human face while choosing lower density at regions where
less detail is needed, such as the cheeks areas. This will lower data size and
improve processing time while high density is preserved where needed (61).
Several algorithms have been proposed for building-up a polygonal surface mesh
from a scanned image. Commonly used 3D model construction algorithms are
slice based surface contour detection (62) ray casting (63) and the marching
cube algorithm, which is the most commonly used algorithm in the medical field
and was first introduced by Lorensen and Cline in 1987 (61).
1.2.1.1 Marching cube algorithm
In 1987, Lorensen and Cline (61) proposed an innovative method of building a 3D
model from volumetric images such as CT and MRI. The process involved
detecting the anatomical structure surface from the grey scale intensity of the
voxels in a volumetric image and rendering it as a 3D polygonal surface mesh
through the following two steps; First, locating the surface corresponding to a
user-specified HU value and creating triangles representing the intersection of
the surface boundary with the voxel sides. Then, to ensure a quality image of
the surface, the algorithm calculates the normals to the surface at each vertex
of all the triangles.
Marching cubes uses a divide-and-conquer approach to locate the surface in a
logical cube (voxel) created from eight pixels, (figure 8); four from two adjacent
slices joined to the corresponding four from adjacent slice. The algorithm
determines how the surface intersects this cube; creates a triangle then moves
(or marches) to the next cube. The algorithm tests each voxels against the
assigned HU value, if the data value at that voxels exceeds (or equals) the value
of the surface we are constructing then these voxels are inside (or on) the
surface. Voxels with values below the selected HU value are outside the surface
(61).
Anas Almukhtar 2016
21
Chapter One
1.2.2 3D image superimposition:
Review of Literature
In orthognathic surgery, image superimposition is mainly aimed at comparing the
change of jaw bones for surgical outcome assessment or growth monitoring.
Therefore, superimposition has to be on a stable area that hasn’t been
significantly changed by the variables being assessed. Classic 2D analysis relies
on the anterior cranial base and forehead and so does the 3D superimposition.
These areas are at a significant distance from the area of surgical change and
reasonably stable during growth (64). In General, three main types of 3D image
superimposition were commonly applied by most of the reviewed literature.
These were landmarks based registration; surface based registration and volume
(voxel) based registration.
1.2.2.1 Landmark based registration
General Procrustes Analysis (GPA) involves translation, rotation and scaling to
match two 3D objects by superimposition. In orthognathic surgery assessment,
scaling is avoided to preserve the geometry of the structures being studied,
therefore, Partial Procrustes Analysis (PPA) is preferred which includes object
translation and rotation only. The procedure involve identifying corresponding
landmarks on two related images followed by translation and rotation of one of
the images bringing the corresponding landmarks into the best fit relying on
their centroid match (65). The centroid is the geometric centre of the shape to
be analysed. This method is also referred to as manual registration since the
landmarks are manually identified. This method is easy to apply but the
potential error is associated with the manual landmarking.
Anas Almukhtar 2016
22
Chapter One
Review of Literature
Figure 8: Marching cube algorithm. Diagram presented by Lorensen W, Cline H, 1987
showing the basics of marching cube algorithm. For each voxel, a cube is formed by
locating 8 vertices, four on each adjacent slice.
Anas Almukhtar 2016
23
Chapter One
Review of Literature
1.2.2.2 Surface based Registration
From its name, this method relies on surface topography matching between the
two images. The procedure is entirely automated but often preceded by a free
hand registration or landmark based registration to provide a better starting
point for the automated process and thus shorten the time required and increase
its accuracy. The method’s main objective is to bring the corresponding vertices
of the two superimposed surfaces as close as possible to each other in the region
of interest. The algorithm utilised for this is called Iterative closest point (ICP)
which involves an iteration of moving (translation and rotation) followed by
calculation of the mean square root distance between the two meshes measured
from each vertex to the closest correspondence on the opposite mesh (66). The
iteration continues until this measurement reaches its minimum value to
maximise the superimposition of the two surfaces.
1.2.2.3 Voxel based registration
The method has been widely used for various medical applications and research
purposes including diagnoses, treatment planning and assessment of a variety of
cases utilizing CT, CBCT, MRI, and 3-D power Doppler ultrasound.
Voxel based registration algorithm utilizes the grey scale intensity difference of
the superimposed DICOM images voxels to approximate the overlapping DICOM
image to the best fit maximizing mutual information based on the grey scale
density calculated between the two images voxel by voxel (67).
This method utilizes relatively complex algorithms for voxel based registration
compared to those used for surface based registration and thus, it requires extra
time to execute and more powerful computer processors are required to handle
the image registration process. In addition, the higher cost of the software
packages capable of voxel based registration might reach up to 10 times the cost
of software packages required for surface based registration.
Anas Almukhtar 2016
24
Chapter One
Review of Literature
Researchers who used voxel based registration have claimed more accuracy in
registration (68,69). However, a recently published article of our work,
Almukhtar et al., 2014 (70), showed that there was no statistically significant
difference between the two methods.
1.2.2.4
Vestibular registration
Vestibular orientation (VO) is a technique developed initially to serve the needs
of comparative anatomy and ontogeny. The lateral semi-circular canal (LSCC) is
used to define the horizontal plane. Following the orientation of the skull to the
horizontal plane, vertical and median planes were derived to serve as a
reference planes for superimposition of sequential images to observe the effect
of growth or surgery on various parts of the skull. A limited number of published
articles have been found for this approach (71).
1.3 3D image analysis
1.3.1 Skeletal analysis
Radiographic imaging including lateral and frontal cephalometric radiography
provides the traditional images for the analysis of the facial skeleton. Their
marked short comings have been pinpointed in numerous scientific publications
(72–74). The method is associated with errors
in the projection, errors in
landmark location and mechanical errors in tracing.
Even though 3D
cephalometry was not practically applicable until recently, Baumrind and Frantz
1971 stated “The conclusion appears inescapable that the total amount of
information contained in head films is not sufficiently large to make clinically
meaningful predictions even if we were able to assess all the contained
parameters perfectly by the use of stereo-head films or by the integration of
information from lateral and frontal films (as in Broadbent’s method)” (73).
Adams et al., (2004) (75) compared the lateral cephalometric measurements
against the direct physical measurements which were considered the gold
standard. The measurements obtained from a single cephalometric radiograph
were
significantly different (75). Van Vlijmen et al., 2010 (76) showed that
Anas Almukhtar 2016
25
Chapter One
Review of Literature
measurements obtained from 2D cephalograms were significantly different from
those obtained from a CBCT image.
The Golden ratios of ‘Leonardo da Vinci’ in the fifteenth century and the Books
of Proportions by Dürer in the beginning of the seventeenth century are the best
example of the historical description of the standard measurements of the
human face.
One of the first cephalometric analyses was introduced by Downs 1956 to
quantify facial morphology. Downs polygon was an effective method of
cephalometric analysis which included three measurements; the facial angle (NPog) to Frankfort plane; angle AB to the facial plane and the angle of facial
convexity (N-A to A-Pog). The
analysis focused on
the skeletal and dental
components, facial soft tissues were not considered (77,78).
Steiner introduced his analysis of the skeletal, dental and soft tissues separately
(79). He was the first to consider the anterior cranial base (Sella to Nasion) as
the line of reference to evaluate the position of the jaw bones (79) instead of
the Frankfort Horizontal line suggested by Downs (77). According to this analysis,
the lips, in well-balanced faces, should touch a line extending from the soft
tissue contour of the chin to the middle of an “S” formed by the lower border of
the nose. This line is referred to as the S-line (80). Steiner's S-line is still used in
orthodontics and orthognathic research in addition to Steiner's skeletal and
dental measurements (81).
Burstone was the first to define a set of landmarks on the soft-tissue profile
(82). The nasolabial angle and facial contour angle are two of many soft-tissue
measurements that he proposed in this analysis.
Vertical relationship was first emphasized by Sassouni (83). He introduced the
terms “skeletal open bite” and “skeletal deep bite” depending upon the
Anas Almukhtar 2016
26
Chapter One
Review of Literature
divergence or convergence of the four horizontal anatomic planes of his analysis.
However, the analysis is not routinely applied for the dentofacial diagnosis (84).
Evaluation of the facial width, facial height and facial contour were dependent
on the facial angle, the XY axis and the facial plane according to Ricketts in
1961(85). Facial contour was measured as the angle between the facial plane (NPog) and A-Pog. The aesthetic line (E-line) is one of the common measurements
to assess the balance of the facial profile (86,87). The E-line extends from softtissue pogonion to pronasale (tip of the nose).
Other analyses including Burstone's method (82), Di Paolo's quadrilateral analysis
(88), Bergman's analysis, and Butow analysis (89) were introduced for
cephalometric analysis of the craniofacial morphology.
Many measurements have been applied clinically to assess soft-tissue facial
changes. The most common ones are the vertical facial proportions, facial
asymmetry measurements, upper anterior teeth exposure at rest, dental
exposure on smiling, middle to lower facial third ratio, upper lip to lower lip
height ratio, nose width and length, nasolabial angle, upper lip prominence,
lower lip prominence, inter-labial gap, labiomental fold, zero-meridian, chin
prominence, chin-neck angle, soft-tissue angle of facial convexity, E-line of
Ricketts, S-line of Steiner, Z angle of Merrifield, and Holdaway's soft-tissue
measurements (22,24,25,26).
Various reference planes recorded the position of anatomical structures
including the Sella-Nasion plane (SN), the Frankfort horizontal (FH), the
mandibular plane and the maxillary plane.
Despite the short comings associated with 2D cephalometry, these analyses are
still routinely applied for diagnoses and treatment planning in orthodontics and
orthognathic surgery.
Anas Almukhtar 2016
27
Chapter One
Review of Literature
1.3.1.1 3D Cephalometric analysis
The development of CT and the dramatic decrease in radiation dose with the
CBCT brought three-dimensional analysis of the head and face to the scene and
made it possible to expand the knowledge of the craniofacial structures into the
third dimension. Errors associated with conventional 2D cephalometry could be
eliminated with the advent of 3D cephalometry (90).
The First attempt to apply cephalometric analysis on 3D models of the skull was
established by Robert W. Fuhrmann 2002 (91). His aim was to copy a traditional
2D cephalometric analysis onto a 3D surface model of the skull segmented from
a CT scan image. His approach was not successful due to the lack of clear 3D
definition of anatomical landmarks. However it opened the doors for progressive
research in that field. In 2006 a simultaneous publication by Gwen R. Swennen
2006 (90,92) and Raphael Olszewski 2006 (93) introduced two new 3D
cephalometric
analyses. The
main
difference
from
the
traditional
2D
Cephalometry was that the anatomical landmarks were digitized on the 3D
surface model of the skull and soft tissue. This subsequently led to the need for
paired digitization of the bilateral landmarks including Orbitale, condylon,
gonion, and porion. 3D Planes were introduced including the Frankfort Plane
which connects two orbitale and a mid-point between right and left porion, the
mandibular plane which connects menton and two gonion landmarks, the
maxillary plane which connects the anterior and posterior nasal spines, Midsagittal plane which connects Nasion and sella points, and is oriented
perpendicular to the horizontal plane and the vertical (coronal) plane which is
established perpendicular to both planes at the position of the sella.
Transforming a line into a plane facilitated the assessment of anatomical
structures in the third dimension.
In his article, Gwen R. Swennen validated the accuracy and reproducibility of his
analysis by measuring inter and intra examiner errors associated with the
identification
of
the
landmarks
(36
Landmarks),
linear
distances
(78
measurements) and angular measurements (164 angles) on 26 CT scan images.
Anas Almukhtar 2016
28
Chapter One
Review of Literature
The results showed low inter and intra examiner errors with high correlation
coefficient (90).
3D-ARCO analysis on the other hand was introduced by Olszewski in the same
year (93). In this the 2D Delaire’s cephalometric analysis was transformed into
3D to evaluate craniofacial deformities. The philosophy of this analysis is that a
reference plane must not rely on landmarks located on the analysed anatomical
structure. In asymmetry cases, for example, the sagittal plane that depends on
external facial landmarks is not acceptable. The author proposed a sagittal
reference plane (orbito-maxillary-sagittal plane) based on landmarks found on
the foramina associated with trigeminal and optic cranial nerves. In this article,
13 CT scan images were utilised for validation of the analysis and a low inter and
intra examiner error were observed.
The published data show that the error associated with 3D-ARCO analysis was
significantly lower than that with Swennen analysis, especially when linear
measurements were compared (94).
Recently, Lee et al., 2014 (95) readopted Delaire’s measurements for 3D
analysis. Their concept was similar to the 3D-ARCO analysis except that the
landmarks were placed on the slices of the DICOM image instead of the 3D
surface model which was claimed in previous studies to be more accurate and
reproducible (96).
Several studies have reported the accuracy of 3D cephalometric analysis (97–
100); Others have compared between 2D and 3D analysis (75,76,94,101); further
studies have looked at the improvement of the accuracy in landmarking the
slices of the DICOM images (95,100,102,103). Two systematic reviews have been
published recently (104,105), they concluded that there is still limited research
to support the validity of the 3D cephalometric analysis. The large variation of
approaches and measurements made it difficult to extract an objective
comparison, and future approaches should provide a more standardised method
of conduction.
Anas Almukhtar 2016
29
Chapter One
Review of Literature
The change in the Cephalometric analysis from 2D to 3D and the variation in the
concept of the measurements between the two approaches highlighted the need
to establish new norms among populations of different ethnicity. Few attempts
have been made in this field. Most recently the norms in a South Korean
population based on 3D architectural and structural cephalometric analysis have
been published (95). Gender difference was the most prominent characteristic
among the population where adult males were found to have higher ramus
length, width, and total facial length.
1.3.1.2 3D surface analyses using colour coded distance map
The 3D colour coded distance map has been applied to analyse the 3D changes in
the facial skeleton after orthognathic surgery (69,106–108). This tool provides a
measurement of the relative distance between two 3D surface meshes by
measuring the distance of each vertex from one mesh to the closest point on the
compared mesh. The measurements are then verified by changing the vertices’
colour in a scale ranging from blue which is the maximum –ve measurement to
red which is the maximum +ve measurement. It is also possible to consider
numerical values by calculating generalized mean and standard deviation,
maximum and minimum measurements. This method is more commonly
associated with soft tissue analysis. Thus, a thorough description has been
provided in the following sections related to soft tissue analysis.
Carvalho et al., 2010 applied a colour coded distance map to each anatomical
component of the mandible and the maximum value represented the main
displacement of that anatomical part (108).
Their method gives the opportunity to use the total surface area to analyse the
skeletal movement and provide a single measurement for comparison. This point
may or may not be at the site of anatomical landmarks.
Anas Almukhtar 2016
30
Chapter One
Review of Literature
1.3.1.3 Changes in centroid position
Cevidanes et al., 2005(109) evaluated the 3D displacement of atomically defined
areas of the mandible affected by orthognathic surgery by measuring the
absolute displacement between its centroids (the centroid is the geometric
centre of a 3D object).
This method simplified the analyses of the 3D movement of a structure by
measuring the displacement of a single point (centroid) since its displacement
depends on the movement of all the vertices of the 3D object.
The short coming of this method lies in that it measures the absolute translation
of the 3D object, the rotation movements will not be recorded if it did not alter
the position of the centroid. The position of the centroid is highly sensitive to
the object’s geometry so that any alteration of the morphology of the
anatomical structure as a result of surgery (removal of the anterior nasal spine
or positioning of the fixation plate) may change the position of the centroid, this
will lead to inaccuracy in measuring surgical movements.
1.3.1.4
Volumetric measurements
The 3D modelling technology offers the ability to virtually measure the volume
of the 3D object. Maal et al., 2012 (110) analysed the displacement of jaw bones
after
orthognathic
surgery
by
superimposing
the
preoperative
and
postoperative to measure the differences in volume. Although this method gives
us an idea about the increase in volume of the jaws after surgery, it does not
identify the new position of the anatomical structure after the surgery.
1.3.1.5 3D Vector analyses
Recently, Park et al., 2013 (111) introduced a new method for assessment of the
asymmetry of mandibular anatomy by analysing the vectors of mandibular
functional units. The magnitude and directional difference of these vectors were
measured to identify the contribution of each part of the mandible to the total
Anas Almukhtar 2016
31
Chapter One
Review of Literature
mandibular asymmetry. This method is highly informative and provides valuable
information about the origin of mandibular asymmetry. However it is specific for
asymmetry cases in addition to the associated landmarking error.
1.3.2 Soft tissue analysis
Direct facial anthropometric measurement is theoretically ideal for the analysis
of facial morphology. However, this requires special skills for the examiners and
has a considerable range of errors compared to 2D or 3D indirect measurements
(33). Among indirect measurement methods, photogrammetry seems to be more
frequently used. However, to be of an acceptable accuracy, it requires a
standardized clinical photographic technique (112–115). Common extra-oral
photographs used for facial assessment are full-face with relaxed lips, full-face
with smile, 45-degree oblique and profile images (116).
During the late 20th century, standardized photographic techniques have been
developed and were adopted by several investigators for facial anthropometry
(112,113,117,118). In comparison with direct anthropometry, standardized
indirect anthropometry allowed the measurements to be carried out on a still
photo rather than on a relatively moving surface of the face, and the negative
impact of the compressibility of the skin is eliminated (119–121). The
measurements can be repeated without recalling the patient and the data can
be permanently stored for future comparative examinations. However, the basic
concept of photogrammetric analysis is to measure the shortest distance
between two landmarks projected on a 2D image (flat plane), this may be
different from the actual geodesic distance between the two points. Two
dimensional photographs may also obscure some anatomical structures which
limits the comprehensiveness of measuring facial characteristics. A compromised
accuracy of land marking was also reported (117).
The introduction of the 3D (x, y and z) Cartesian coordinate system by René
Descartes in the 15th century allowed the spatial orientation of physical objects
to be measured in a digital framework. Various 3D capturing techniques were
Anas Almukhtar 2016
32
Chapter One
Review of Literature
introduced (Section 1.1.1). These are capable of providing three dimensional
information (point cloud data) that allow volumetric surface and curve analyses
to be carried out in addition to the classical linear and angular measurements.
In orthognathic surgery, various methods were applied to quantify facial
dysmorphology and measure the changes following surgical correction. The
methods
ranged
in
their
complexity
from
simple
linear
and
angular
measurements up to the more complex statistical models including curve
analysis and PCA. More than 200 published scientific articles were reviewed in
this chapter, in about 50% of these studies analysis of the facial soft tissue was
the main objective. Table 1 shows the methods applied in the published data to
study the impact of orthognathic surgery on the covering soft tissue.
1.3.2.1 Landmarks based analysis
Landmarks based analysis is the most common method for the evaluation of soft
tissue morphology. However, the method is limited to the analysis of the
changes at a set of particular points which does not describe the complex nature
of the facial morphology. Measuring the shape changes at the surface between
the selected points is not considered in this method. Unlike the hard tissue, this
argument seems to be more appealing here since the soft tissue surface is not a
solid structure and the response of a surface area to the changes in the
underlying skeletal changes is too complex to analyse at a certain set of
anatomical landmarks. The limited number of reproducible points of the facial
soft tissue restricts the value of the landmarks based analysis.
Farkas et al., 1980 (112) suggested 16 facial landmarks that are still used these
days as the gold standard for facial soft tissue analysis (table 2).
Hwang et al., 2012 (122) studied facial asymmetry on 60 stereophotogrammetry
images using Farkas’ landmarks. The orthogonal distance of each landmark to a
common three dimensional reference planes were measured. The results showed
that facial asymmetry was more obvious in the lower third of the face, ch, Ala
Anas Almukhtar 2016
33
Chapter One
Review of Literature
and Me landmarks showed the highest asymmetry scores. Similar studies
considered more facial landmarks (123–127). G Sforza et al., 2007 (128) digitized
50 landmarks to evaluate the changes following the surgical correction of
skeletal class III jaw deformity in 7 cases. The surgical results were compared
with the facial appearance of 87 healthy females as a control group. The results
showed that facial deformities were significantly improved after orthognathic
surgery.
In addition to the standard set of anatomical landmarks mathematically
generated ones were also utilized for facial analysis. These were identified by
the intersection of two or more lines passing between anatomical features or at
a specific distance and direction from an adjacent landmark.
These extra landmarks allow a more comprehensive analysis of the facial
morphology. Terajima et al., 2008 (129), Kim et al., 2011 (130) and Park et al.,
2013 (131) applied a grid of intersecting lines comprised of 10 horizontal equally
distributed lines at 4.5 mm distance and 27 vertical lines equally distributed at
5mm distance from each other. The grid was orientated vertically and the top
first line passes through the right and left “orbitale” landmarks. The
preoperative and post-operative CBCT images were superimposed, and 270
points of intersections between the vertical and horizontal lines were dropped
on the soft and hard tissue for analysis. Hoefert et al., 2010 (65) used a set of
mathematically generated landmarks for image superimposition and analysis
such as the mid-point from (go) to (me) and mid-point from (ex) to (ch) (8,9)
(figure 9).
Anas Almukhtar 2016
34
Chapter One
Review of Literature
Table 1: Types and percentage of published methods of soft tissue analysis 2000-2015
Rank
Type of analysis
Percentage
1
Landmarks based Linear and angular analysis
50%
2
Colour coded distance map
28.7%
3
Absolute surface distance
11.95%
4
Vector analysis
8.75%
5
Volumetric measurements
5%
6
Principal component analysis
3.75%
7
Dense correspondence analysis
2.5%
8
Curve analysis
1.25%
9
Centroid and GPA
1.25%
10
Statistical landmarks plotting
1.25%
Table 2: Definitions and abbreviations of facial soft tissue landmarks by Farkas 1980(67)
Landmarks
Abbr. Definition
Glabella
Gla
Most prominent midline point between eyebrows
Nasion
Na
Deepest point of nasal bridge
Pronasale
Prn
Most protruded point of the apex nasi
Subnasale
Sn
Midpoint of angle at columella base
Labial superius
Ls
Midpoint of the upper vermilion line
Stomion
Sto
Midpoint of the mouth orifice
Labial inferius
Li
Midpoint of the lower vermilion line
Menton
Me
Most inferior point on chin
Exocanthion*
Exc
Outer commissure of the eye fissure
Endocanthion*
End
Inner commissure of the eye fissure
Alar curvature*
Ala
Most lateral point on alar contour
Cheilion*
Ch
Point located at lateral labial commissure
* Paired (right and left) landmarks.
Anas Almukhtar 2016
35
Chapter One
Review of Literature
The spatial information of soft tissue landmarks was either digitized directly on
patients face using an electromagnetic digitizer in some studies (34,128,132,133)
or virtually by digitizing landmarks on a 3D image viewed on a computer
screen(129,134–137). Infra-red (IR) sensors were also utilized to detect the 3D
position of IR emitting landmarks fixed on the patient’s face (138,139).
The accuracy of landmarking is variable according to the technique used for
imaging as well as the method of digitization. Fourie et al., 2011 (140) evaluated
the accuracy of landmarking of three different imaging systems (laser scanner,
CBCT and stereophotogrammetry). Twenty-one linear measurements were
developed based on 15 landmarks placed on the three images, and these were
compared to those obtained from direct anthropometric measurements. The
difference in the mean absolute distances between the three measurements was
less than 0.89mm. However the study did not compare the absolute digitization
error at individual landmarks.
Plooij et al., 2009 (141) evaluated the
reproducibility and reliability of 3D landmarking of stereophotogrammetry
images. 49 soft tissue landmarks were digitised twice by 2 observers. The images
were loaded into specialized software (Maxilim) where the orthogonal distance
of each landmark to 3D reference planes was recorded. Paired sample t-test and
correlation coefficient evaluated the reproducibility and the reliability of the
digitisation. The results showed no significant differences between the interand intra- observer measurements with a high correlation coefficient.
Othman et al., 2013 (142) evaluated the reproducibility of facial soft tissue
landmarks on a stereophotogrammetry image which showed
no statistically
significant difference between the repeated readings.
Gwilliam et al., 2006 (143) digitised 24 soft tissue landmarks on six 3D facial
scans. Thirty orthodontists of varying experience were asked to landmark the
facial scans to establish inter-operator reproducibility. The results showed that
only 50% of the landmarks were associated with less than 1 mm error when redigitised by the same observer. Whereas only 8% of the landmarks showed less
than 1mm of error in reproducibility when digitised by different observers.
Anas Almukhtar 2016
36
Chapter One
Review of Literature
Similar results were obtained by Toma et al., 2009 (144). 21 landmarks were
placed on 30 laser scanned facial images by two observers. This was repeated
after two weeks. Bland-Altman plot was used to assess the reliability of each
landmark and only 50% of the landmarks showed less than 1mm error and less
than 0.5 mm error was observed in 35% only.
The results of the validation studies led to the conclusion that despite the high
level of precision in digitising landmarks the accuracy level of 0.5 mm was only
achievable in few of the landmarks.
Lower landmarking error was observed by Khambay et al., 2008 (12). A
reproducibility study was carried out to assess operator errors of landmark
digitization on high resolution stereophotogrammetry images captured by the
DI3D system. Operator error was measured by repeatedly locating landmarks on
the three-dimensional image. The results showed that the operator error in
placement of landmarks on the three-dimensional model was 0.07 mm, range
0.02 mm – 0.11 mm.
1.3.2.1.1 Linear and angular measurements
Around 50% of the reviewed published papers used this approach. Changes in
facial soft tissue after surgery have been assessed using two main methods:
Measuring the linear and angular distance between landmarks; and their 3D
orthogonal distances against common 3D reference planes (134,136,143–147).
Some of these were part of a 3D soft tissue cephalometric analysis. The
measurements were made on the pre- and post-operative images separately and
the results were compared to find the differences. In a minor fraction of these
studies the assessments were made by measuring the net distances between
corresponding landmarks on both images, and in this case an additional step of
image superimposition was needed. Facial heights, A-P position of soft tissue A,
B Prn, Sn, and soft tissue Pog were the common linear and angular
measurements in addition to facial asymmetry measurements.
Anas Almukhtar 2016
37
Chapter One
Review of Literature
Chen et al., 2012 (148) utilized linear measurements between corresponding
landmarks to investigate profile changes after surgical treatment of mandibular
prognathism. Thirty patients (20 females and 10 males) underwent vertical
ramus osteotomy. Preoperative and postoperative cephalograms were analysed;
landmarks were identified and compared. The mean setback of the pogonion
(Pog) was 11.7 mm. The setback ratios of labrale inferius (Li) / incision inferius
(Ii), labiomental sulcus (Si) / point B, and soft tissue pogonion (Pog) / pogonion
(Pog) were 0.98, 0.99, and 0.95, respectively. There were no sex related
changes in soft tissue.
Ubaya et al., 2012 (147) carried out a study involving linear and angular
measurements on 3D stereophotogrammetry images of 40 patients following
orthognathic surgery and compared them to a control group of 112. The study
aim was to evaluate the 3D naso-maxillary complex soft tissue morphology
following Le Fort I maxillary advancement. The results showed that facial
morphology post-surgery was similar to the reference group, except the nasal
base width which was wider by 2.3 mm in males and 2.6 mm in females. In the
orthognathic group, the females had a smaller nasolabial angle by 9.78° than the
reference group.
1.3.2.1.2 Vector analysis
This approach was considered by about 8.75% of the reviewed articles (149–151).
Vectors analysis considers the magnitude and direction of the linear
displacement of corresponding landmarks. This is particularly beneficial to study
the mechanism of oro-facial changes in 3D or 4D. This method can produce
valuable information when combined with a more comprehensive type of
analysis including dense correspondence analysis which measures the changes of
every vertex on the mesh of the surface of the face (152).
1.3.2.1.3 Thin plate spline (TPS)
This approach was considered in 1.25% of the reviewed literature (153). TPS was
first introduced by F L Bookstein (154). It measures the bending deformation of
Anas Almukhtar 2016
38
Chapter One
Review of Literature
the geometry of a particular shape guided by a group of corresponding
landmarks. A mathematic algorithm then smooths these deformations by
minimizing the localized bending energy of the shape produced by the original
change.
Bugaighis et al., 2010 (153) examined the facial soft tissue of 4 categories of
cleft lip and /palate children (40 with a unilateral cleft lip and palate (UCLP), 23
with a unilateral cleft lip and alveolus (UCLA), 19 with a bilateral cleft lip and
palate (BCLP), and 21 with an isolated cleft palate (ICP)). Patients’ faces were
captured using stereophotogrammetry and compared to 80 age and gender
matching control group. After applying thin plate spline algorithm to measure
the differences of the facial morphology among the study groups, the
transformation matrix quantified the change in the mean distance of the facial
landmarks from their centroids. The principal component analysis was applied to
identify the morphological difference among the cleft groups.
1.3.2.1.4 Principal component analysis (PCA)
PCA is a statistical multivariate correlation analysis aimed at simplifying large
data analysis by limiting the number of variations of shape differences to those
with a significant contribution to the total variability. This approach does not
only simplify the analysis but also emphasizes the level of importance of each
associated aspect of variations by analysing the distribution of its effect around
the mean. The Eigenvalues associated with variation vector for each variable are
measured. The “eigenvalue” and “eigenvector” are the determinants of the
transformation matrix of that shape deformation. In other words, shape
deformation has both magnitude and direction which could be calculated by
analysing its transformation matrix. When PCA is applied to analyse variation
among populations or to quantify facial shape deformation as a result of
orthognathic surgery, the input from each case may show an individual variation
in its deviation from the mean shape deformation. Face length, width, alar base
width may act as contributing factors to the total variability. The percentage of
the variability associated with each factor can be calculated, their direction can
be illustrated in the 3D dot plot. The factor that contributes to the highest
Anas Almukhtar 2016
39
Chapter One
Review of Literature
percentage of variability is considered the first principal component (PC1) and
has its unique direction and magnitude. The next contributing factor with the
next higher percentage is (PC2) and its direction of variation is orthogonal to the
first component. The calculation of the subsequent smaller Principal components
continues until a total percentage of variability <80% is reached which is
considered satisfactory to measure the level of variability between shapes. In
most cases 5-8 principal components describe about 80% of the level of
variability of shapes.
Difference in facial forms as a result of growth, gender, and ethnic variations
has been assessed using PCA (155–157). The effect of orthognathic surgery was
analysed by assessing in the same manner.
Toma et al., 2012 (157) applied principal component analysis to analyse facial
variation among 4747 British school children. Their aim was to identify key
components of variation in a sample of 15.5 years old children. Laser scanned
images were obtained and 21 reproducible facial landmarks were identified and
their coordinates were recorded. The images were superimposed using
Procrustes analysis (PA) and principal component analysis was applied to each
landmark cluster which consisted of 21 landmarks for the whole sample. 14
principal components were identified which in total constituted for 82% of the
variations. The first, second and third principal component represented 46% of
the variations which were the facial width, facial length and nose projection
respectively. The impact of gender differences in facial shapes was minimal.
Principal component analysis is a useful tool in describing the pattern of
variation in large study populations where other analyses are short of description
or labour intensive.
However, the method provides descriptive rather than
quantitative analysis.
1.3.2.1.5 Statistical landmarks plotting
Sforza et al., 2010 (133) applied this approach as an adjunctive method which
helped in describing the effect of associated asymmetry in faces on the level of
attractiveness from childhood to early adulthood. The coordinates of 50 facial
Anas Almukhtar 2016
40
Chapter One
Review of Literature
landmarks were digitised for 148 male and 232 females (and in 669 controls
consisting of 397 male and 272 female) using an electromagnetic digitizer. The
landmarks were plotted on a Radar chart (web chart) where each of the mean
landmarks were digitized on a radial line originated from the centre (zero). The
mean of asymmetry measured at each landmark determined the position of this
landmark on its radial line, (figure 10). A continuous line is then connected
between the landmarks of each category in a circular fashion around the centre.
Superimposing different categories on the same plot described clearly the
variation among different categories. The results showed that asymmetry in
attractive children was less than that in normal control group. This was not the
case with adults where a considerable amount of asymmetry was found in
attractive subjects especially I the male sample.
1.3.2.2 Surface based (landmarks free) analysis
The inclusion of the entire facial surface mesh in the analysis overcomes the
major limitations of the landmarks based analysis. The analysis of surface mesh
includes colour coded map, volumetric analysis, absolute surface distance,
dense correspondence analysis and curves analysis. Despite the comprehensive
nature of these analyses most of them are more descriptive in nature unless they
are combined with refinements by landmark selection. Colour coded map ,as an
example, will give an excellent visual description of the facial surface change
after orthognathic surgery. However, it will only be objective enough when
combined with landmarking either to measure the surface distance at a specific
anatomical landmark area or to separate a desired surface patch (158,159).
Anas Almukhtar 2016
41
Chapter One
Review of Literature
Figure 9: Mathematically generated landmarks applied by Hoefert et al., 2010 (65)
Figure 10: Landmark plotting applied to evaluate facial asymmetry by Sforza et al.,
2010 (133)
Anas Almukhtar 2016
42
Chapter One
Review of Literature
1.3.2.2.1 Colour coded map
This approach was the second in terms of the percentage of published articles
reviewed in this section (28,160–166). Around 29% of the reviewed literature
applied the colour coded distance map (163).
The entire facial surface is involved in the analysis; it produces visually
displayed, easily interpreted findings for clinicians and researchers, the method
provides both generalized surface measurements as well as localized patches on
the area of interest.
To apply this method, the two facial image meshes to be compared should be
superimposed to allow direct measurements between them. Its mathematical
algorithm considers every vertex from one of the meshes (Mesh A) as a landmark
and seeks for the closest point on (Mesh B) to be considered as a correspondent
point for distance measurements. The results are comprehensive information
which can be utilized in describing general and local facial changes after
surgery. Three main outcomes are expected when adopting this method;
Generalised colour coded distance map which can be displayed in a range of
colours describing the distance between the two tested surface meshes, (figure
11). The shell displacement expressed as a spectral colour range in which,
specific colour is allocated to each vertex according to its distance from the
corresponding point (closest point) on the adjacent mesh surface. Considering
the geometric centre of the two superimposed meshes as a directional reference
point, the vertices of the outer mesh are highlighted in red while the vertices of
the inner mesh are highlighted in blue. The higher the distance between the
meshes in the direction away from the geometric centre (the outer mesh) the
more the shift of the colour towards the red side of the spectrum and it has the
positive sign. On the other hand the higher the distance between the two
meshes in the direction towards the geometric centre of the object (the inner
mesh) the more the shift of the colour toward the blue side of the spectrum and
it has the negative sign. Consequently, the midpoint of the spectrum which is
the green colour is allocated to the vertices with zero distance on both meshes.
Variable regions on the face will take variable colour according to the magnitude
Anas Almukhtar 2016
43
Chapter One
Review of Literature
and direction of their geometrical differences which results in a descriptive
colour map. To explain the concept in more details, an example of orthognathic
surgery results of a patient with Class II malocclusion is shown in, (figure 11).
The combined (superimposed) images (A), the Post-operative (B) and Preoperative (C) images are shown. The combined images view shows only the
positive part of the spectrum (green –red) when viewed from the outside of the
3D facial image where it will show the negative part of the spectrum (greenblue) when viewed from inside of the 3D facial image. The Post-operative view
showed generally green areas at the areas of forehead and part of the midrace
where no effect of surgery were evident while the lower face showed a marked
shift towards the red colour which indicates a positive (forward) displacement of
the soft tissue as a result of mandibular advancement as part of the treatment.
The maximum positive and maximum negative values can be manually adjusted
to emphasize minor changes.
Kau et al., 2006 (163) introduced a slight modification to this method. Instead of
using the spectral colour scale ranging in the positive and negative direction
around the zero point, threshold points were determined which represented the
level of clinical significance. This was named “tolerance value”. Vertices with
lower distance measurements were displayed in black colour whereas those with
higher distance measurements were displayed in white colour. This allowed the
colour coded map to be interpreted in a more objective way. The method was
applied to evaluate surgical results, gender and age differences as well as
variation among populations (29,160,167).
The colour coded map gives a visual description of the extent and location of the
surface change. However its accuracy can be criticised by the fact that it relies
on the geometric proximity of adjacent surface shells to create the
correspondence rather than the real anatomical one. This limitation is clearly
evident when asymmetric faces are evaluated.
Another modification is the patch displacement analysis (158,168), the mean
and standard deviation of the distances between a group of interconnected
Anas Almukhtar 2016
44
Chapter One
Review of Literature
vertices (surface patch) in Mesh (A) and their corresponding nearest points on
Mesh (B). It is possible to provide the mean and standard deviation for the whole
face mesh. However the localization of the measurements to the region of
interest provides a localized and more precise estimation of the resultant
deviations. The precision in selecting patch boundaries which is based on
identifying anatomical landmarking is the preferred approach in this analysis.
The mean and standard deviation of the distance between the two meshes in
addition to the maximum and minimum measured distances are the numerical
values which were considered in most of the studies (158,169). These
measurements described the differences between the meshes especially when
localized surface patches are being examined. However the lack of anatomical
correspondence and the need for standardized accurate landmarking are among
the limitations of the method.
1.3.2.2.2 Absolute surface distance
This type of analysis has been considered in about 11% of the reviewed articles
(29,66,160,170–173). It has also been used along with other types of analysis
mainly with colour coded map (158,169). The mean distance between the two
adjacent facial meshes is calculated by measuring the distance from each vertex
on one mesh to the nearest point on the other. The method is dependent on the
accuracy of the superimposition of the images which should precede the
distance analysis.
In general, analyses carried out using this method fall into one of the following
four categories with marginal difference in calculated mean distance, (figure
12). The four types as stated by Miller et al., 2007 (174) are:
Normals: In this type the distance from each vertex on mesh (A) is
measured to the nearest point on Mesh (B) which lie on a line
perpendicular to that vertex regardless of the presence of closer points.
Anas Almukhtar 2016
45
Chapter One
Review of Literature
Radial: The distance is measured between the two meshes along radial
lines originated from the geometric centre (centroid) of the mesh (A)
regardless of the presence of closer points.
Closest point: This type measures the net distance from every vertex on
mesh (A) to the closest point on mesh (B).
Correspondence with sensitivity to movement (CSM): This type follows the
geometrical similarity in identifying the correspondences between the
two surfaces.
Despite the type of the analysis, this method calculates the absolute distance
between the two meshes and produces the mean, standard deviation, minimum
and maximum values. For a more specific analysis, this method can be applied
on a local patch on the surface (66,70). In that case accurate landmarking is
essential to standardise the dimensions of the selected patch.
1.3.2.2.3 Dense correspondence analysis
This method was applied for the analysis of shape change in about 2.5 % of the
published articles on the topic (152,175,176). Although it is considered as a
landmark free analysis, it could also be considered as a comprehensive landmark
based analysis since it treats every vertex on the 3D surface as an individual
corresponding landmark. It was introduced to overcome the lack of anatomical
correspondence associated with the previous methods. This was achieved by
providing a reproducible index of all vertices on the assessed images. This index
provides the guide to create an actual anatomical correspondence between the
compared
images.
The
method
combines
the
advantages
of
the
comprehensiveness of the landmark free surface analysis and the objective
precision of landmarks based analysis.
Anas Almukhtar 2016
46
Chapter One
Review of Literature
Figure 11: Colour coded map. Pre-operative (a), post-operative (b), and superimposed (c),
and colour distance map (d). Kau et al., 2006 (263)
Figure 12: Types of absolute distance measurements: Normal surface distance A, radial
surface distance B, Closest point surface distance C and corresponding points surface
distance D. Ryckman et al., 2010 (134)
Anas Almukhtar 2016
47
Chapter One
Review of Literature
The analysis is rather complicated and involves multiple stages:
a) Construction of Generic mesh.
The facial soft tissue surface is captured by a three dimensional acquisition
devise e.g. stereophotogrammetry or laser scanner. Depending on the resolution
of the scan the mesh may comprise thousands of points. Conventional landmark
analysis based on anatomical structures involves a small number of these points
which reduces
the utilisation of the captured data. Ideally each point on the
surface mesh should be a landmark; however, the manual identification of
thousands of points (landmarks) would be impossible. The application of generic
surface overcomes this problem and allows a more comprehensive assessment of
the captured data.
A generic facial model is a simplified polygonal mesh representing the
morphometric information of an average face with known 3D co-ordinates of
each point of the mesh in addition to a reproducible polygonal index. The mesh
is generic so it can be adapted universally to any face for the purpose of
morphometric analyses. Generic meshes are a key tool in studies that deal with
graphical representation of surface morphology to analyse the areas of
difference and similarity.
The generic mesh model can be generated using virtual 3D modelling software. It
can also be an average facial model based on the facial topography of the study
population (167).
The topography of the generic mesh should include the common features of the
study population and should not have extreme or unique characteristics out of
the normal range, for that reason, an average face could be always chosen from
the study group to guarantee accurate future analyses.
Anas Almukhtar 2016
48
Chapter One
Review of Literature
According to Kau and Richmond 2010 (177), four methods can be used to extract
the average face. These include; straight forward point-wise averaging in the zdimension(A-P); averaging in the radial direction of the average facial cylinder;
averaging in the radial direction of the average face sphere; averaging in the
locally normal direction to a template shell.
The first three methods, extract the average face using statistical algorithms
that measure the relative distance to a plane, a vertical central line and a facial
central point. The fourth method depends on the extraction of the average face
by superimposing the entire study sample to a template shell produced by one of
the previous averaging methods using the generalised procrustes algorithm
followed by ICP.
To create the generic mesh model, most of the biological morphometric studies
advocate the average polygonal 3D mesh over other modelling meshes. The
choice is justified as the polygonal mesh has the capability to accurately
represent a biological shape especially in detailed features areas, folds and
edges combined with the possibility to produce uniform polygons of relatively
equal sizes throughout the mesh. These characteristics facilitate creating a
polygonal index necessary for a uniform assessment of different anatomical
areas on the facial mesh (178).
b) Conformation
Conformation is the adaptation of the generic mesh to the facial surface mesh in
a process known as “elastic deformation”. The generic mesh is warped onto the
underlying facial morphology. This conformation process can be achieved
through a two-step algorithm: In the first step (GPA) generalized Procrustes
algorithm (179) is applied followed by thin-plate spline (TPS) warping (154).
This brings the landmarks of both the facial surface and the generic mesh into
exact alignment and produces a smooth transformation for the other parts of the
mesh by minimizing the bending energy for the close adaptation of the surfaces.
Anas Almukhtar 2016
49
Chapter One
Review of Literature
The second step accomplishes fine local adaptation of the generic mesh to facial
morphology by deforming each vertex of the generic mesh into the location of
the closest vertex on the target mesh surface guided by surface topography.
This algorithm results in cloning of the target face mesh and creating a
simplified, uniform and indexed polygonal mesh which carries the same facial
features of the original mesh.
c) Superimposition
In orthognathic surgery, dense correspondence analysis was used to analyse
localized change on the facial soft tissue as a result of orthognathic surgery. In
order to emphasize these changes, the analysed meshes (pre surgical and postsurgical) should be superimposed on stable areas unaffected by surgery. The
most common method of superimposition is the surface based registration,
(Section 1.2.2).
d) Dense correspondence analysis
The generic mesh is an indexed mesh; each vertex of the mesh has its analogues
vertex on any other mesh with the same index. This characteristic provides a
real dense anatomical correspondence for a comprehensive analysis of the whole
facial surface with a high level of precision.
Dense correspondence analysis was applied to a variety of facial anthropometric
analysis. In their study, Claes et al., 2001 (175) applied this method to
investigate growth pattern of normal and syndromic patients. In this case, the
scale of the colour index is different. Although it involves the full colour
spectrum (blue to red), the blue colour refers to (zero) whereas the red colour
refers to the maximum value of differences between the corresponding vertices
of the dense surface mesh. A similar study was carried out by Al-Hiyali et al.,
2015 (180). Their analysis involved motion 3D imaging (4D) to investigate the
impact of orthognathic surgery for correction of asymmetry on facial animation,
(figure 13).
Anas Almukhtar 2016
50
Chapter One
Review of Literature
Principal component analysis can also be applied for the comprehensive
evaluation of the surface change. Curve analysis as described by Higgins 2014
(181) can be applied on the uniform structure of the indexed generic mesh to
optimize curve correspondence. The method depends on multiple steps involving
various procedures for image superimposition and conformation. Each of these
steps has its associated margin of errors.
Figure 13: Dense correspondence surface analysis. Note the minimum value on the colour
scale is ZERO. Al Hiyale et al., 2015 (180).
1.3.2.2.4 Volumetric analysis
The method was applied in 5% of the reviewed articles on the soft tissue analysis
following
orthognathic
surgery
(155,159,182,183).
The
analysis
provides
information about the change in the volume of the whole face or at specific
regions (33,184). The method is considered an adjunctive to other analyses.
Chau et al., 2008 (182) applied angular and linear measurements in addition to
the colour coded map to detect changes in volume around the nose region.
Anas Almukhtar 2016
51
Chapter One
Review of Literature
1.3.2.2.5 Curve analysis
This approach has been recently developed and applied in 2.5 % of the reviewed
literature (123,181,185).
Similar to dense correspondence, 3D curve analysis is a comprehensive
landmarks based analysis to describe surface morphology. Human facial features
including cheeks, lips, nose and chin are represented by smooth recognizable
curves that describe the morphological characteristics. These curves are
detectable, reproducible and comparable (181). Curve analysis is a promising
research tool to describe and quantify the change in facial form after jaw
corrective surgery.
Higgins 2009 (181) demonstrates the application of curve analysis to evaluate
the effect of surgical correction of cleft lip and palate and Principal Component
Analysis was applied to describe the characteristics of facial morphology. Two
methods of curve detection were applied. The first was named “planner based
curve identification”. Curves were identified by the creation of a plane that
passes through the face, the points along the surface – plane interface
represents the curve to be analysed. Some geometrically complex areas around
the nostrils and lips regions may require a combination of more than one plane
to accurately identify the desired curve. This method was criticized for its
failure to produce satisfactory curves in all areas of interest in addition to the
complexity of the computational process for curves identification (181).
The second curve identification method was “surface curvature based curve
identification”. This type utilised surface topography for curve detection
minimizing the need for landmarking to only two at the ends of the curve. It is
well known that facial soft tissue surface composed of a variety of surface
topography features as peak, pit, saddle, ridge and valley. It is therefore logical
to use surface characteristics for curve detection. The “ridge” was the ideal
geometric feature to detect facial curvatures. 3D curve analysis is a promising
tool, however, the accuracy of the method has not been fully evaluated.
Anas Almukhtar 2016
52
Chapter One
Review of Literature
1.3.3 Hard-Soft tissue correlations and prediction of surgical
results
A review of the scientific literature from the mid-20th century until present days
showed the growing interest in understanding the behaviour of the facial soft
tissue following surgical repositioning of the underlying jaw bones. A common
aim among researchers has been to find an approach to accurately predict the
soft tissue outcome following orthognathic surgery.
Until recently, studies have been limited to the use of two-dimensional (2D)
cephalometric radiography to analyse the relationship between soft and hard
tissue. The main approach was dependent on linear or angular measurements to
track the post-operative positional changes of certain anatomical hard and soft
tissue landmarks and to extract a mathematical ratio between their movements
(186–188).
The introduction of digital cephalometry resulted in an improvement in the
accuracy of landmark location and measurements (189). However, the inherent
inaccuracies of imaging and analysing a three-dimensional structure in 2D still
remained (190). It is worth noting that the measured change of landmark
position reflects the behaviour of the soft tissue at this particular point and do
not indicate changes in the surrounding region. This is critical due to the multifactorial flexible behaviour of the soft tissue. Adjacent points to the digitised
landmark on the face may react entirely differently to the same underlying hard
tissue surgical movement.
In the beginning of the 21st century, positive steps were taken following the
introduction of 3D modelling into the medical diagnostic field in attempt to
understand the correlation between the skeletal and soft tissues. The
advancement of CT Scanning and MRI technology provides the accurate recording
of 3D topography of the facial hard and soft tissues and allows a comprehensive
analysis of the recorded morphology.
Anas Almukhtar 2016
53
Chapter One
Review of Literature
The first part of this section aimed at reviewing previous studies on the
correlations of soft and hard tissue movement following orthognathic surgery
using both 2D and 3D imaging modalities.
1.3.3.1 Two Dimensional correlations and ratios
1.3.3.1.1 Maxillary osteotomies
The effects of Le Fort I advancement on the naso-labial soft tissue complex have
been thoroughly investigated (130,187,191–208).
In general, four points on the soft tissue profile were the basis for the linear and
angular measurements to quantify the relationship between soft and hard tissue
movements following maxillary osteotomy. These points were the nasal tip
(Prn), the deepest point at the naso-labial angle (Sn), the vermilion border of
the upper lip (Ls) and the vermilion border of the lower lip (Li). These were
correlated to the skeletal movements of the following points; the anterior nasal
spine (ANS), the anterior maxilla (A point), the crest of the upper alveolar bone
(Par) and upper incisor edge (UI) respectively (187,191).
In response to maxillary advancement, several studies reported that the upper
lip at the vermilion border moved anteriorly by a mean ratio of 0.7:1 (209)
(192). However, larger ratios approaching 0.9:1 have been found (201,210). On
the other hand, smaller ratios have been also reported
(211). A possible
explanation for this variation in the soft tissue response to maxillary surgery was
caused by the associated soft tissue alteration including the V-Y closure of the
mucosal surface of the upper lip and cinching of the alar base which is routinely
carried out to avoid flaring. These may result in an increased thickness of the
soft tissue at the naso-labial region (201). In other studies maxillary
advancement was combined with a variable amount of surgical impaction which
will alter soft tissue response to surgery (210), while the small test sample (8
patents including 2 cleft lip cases) may be the cause for the smaller ratios (209).
Anas Almukhtar 2016
54
Chapter One
Review of Literature
Based on the previous literature findings, the movement of the upper lip is
highly variable and dependant on several factors including soft tissue thickness
(196,210), the direction and amount of tooth movement and the initial position
of the upper lip (212,213), Patents gender (214) and ethnic group (215) .
The thicker the lips the lower the soft to hard tissue movement ratio as the soft
tissue bulk tends to absorb the skeletal movement and reduces the detected
changes (196). Another parameter which may influence soft tissue response to
Le Fort I maxillary advancement is the presence or absence of a gap between
anterior maxilla and the upper lip. In most cases of maxillary deficiency this
space exists which may reduce the ratio of soft tissue movement following
maxillary advancement since the initial movement of the hard tissue will be
within the existing space before it contacts the mucosal surface of the upper lip.
The effect of skeletal maxillary advancement may extend to other soft tissue
structures including the nose and lower lip (192,209). However, these effects are
variable and studies have reported the difficulty in analysing the correlation
between the hard and soft tissues at these regions (210).
In general, maxillary advancement surgery moved the tip of the nose and
subnasale anteriorly with a ratio ranging from 0.2:1 to 0.7:1 (216). Few studies
have shown the effect of maxillary advancement on the position of the lower lip,
the ratio was found to be ranging between 0.2:1 to 0.57:1 (192,209).
The effect of maxillary advancement is not limited to the forward movement of
the overlying soft tissues. Several studies detected an upward vertical shift of
the naso-maxillary complex and shortening of the upper lip. However the results
were variable with a large range of the standard deviation (210). On the other
hand, Carlotti et al.,1986 (201) showed that there is no change in the upper lip
length. A possible explanation may be the surgical procedure for this group of
patients included a V-Y closure which was used to maintain lip length.
Anas Almukhtar 2016
55
Chapter One
Review of Literature
The effect of skeletal maxillary advancement on the upper lip of cleft lip
patients has been investigated (217).
Soft tissue responses to maxillary
advancement in patients with cleft palate were compared to non-cleft patients
with deficient maxilla. The sample of 50 patients was divided equally into two
groups; cleft lip and non-cleft with deficient maxilla. Both groups were treated
with a Le Fort I maxillary advancement with no vertical change. The results
showed that the cleft lip group expressed as higher ratios of soft to hard tissue
movements compared to non-cleft patients in both the horizontal plane 0.66:1,
0.54:1 and in vertical plane 0.53:1, 0.23:1 respectively. This may have been due
to the scar tissue formation in the upper lip following primary lip repair and the
difference in the magnitude of maxillary advancement which was not
standardized in this study.
In a more recent study Louis et al., 2001(187) investigated the effects of
maxillary advancement for treatment of class III patients with sleep apnoea.
The vermilion border of the upper lip was advanced 0.8 and moved superiorly
0.16, subnasale point at the naso-labial angle was advanced 0.39 and moved
superiorly 0.16 and the nasal tip was advanced 0.16 and moved superiorly 0.16
per unit of maxillary advancement measured at the upper incisal edge. The
results of the study were different from those reported in previous findings. The
upper incisal edge was the only hard tissue reference point to represent the
complex skeletal movements which was a major limitation of the study.
1.3.3.1.2 Mandibular Osteotomy
There are numerous studies investigating the response of the soft tissue to
mandibular osteotomies since the mid-20th century (218,219). In the early
articles, changes in the lip and chin soft tissue, following skeletal correction of
mandibular position were based on profile cephalometric measurements. Similar
findings were reported, for each 1mm posterior movement of the bony chin, the
soft tissue chin moved 1 mm in the same direction while the lower lip moved 0.6
mm. In other words, the ratio of soft to hard tissue movement after mandibular
setback and advancement surgery was estimated to be 1:1 at the chin region
whereas the ratio was 1:2 for mandibular advancement, 1:1 for mandibular
Anas Almukhtar 2016
56
Chapter One
Review of Literature
setback at the lower lip region. More recent studies have also supported the
same ratios (148,186,220,221).
Several factors influenced the soft tissue behaviour in response to mandibular
movements, these include, the thicker the lips the lower the soft to hard tissue
movement ratio as the soft tissue bulk tends to absorb the skeletal movement
and reduces the ratio; females in general tend to show higher ratios than males;
a nonlinear correlation has been found between the amount of skeletal
movement and the extracted ratios as higher ratios were observed with the
increased magnitude of skeletal movements (214,220,221).
The effects of mandibular setback on the soft tissue profile as a whole rather
than just the lower lip and chin has been reported (222). The study found that
not only changes in lower lip and chin region occurred as a result of mandibular
surgery but all the facial profile is affected including the upper lip and nasolabial region. The main effect of mandibular setback surgery on the soft tissue
profile is an increase in facial convexity, deepening of the mentolabial fold,
lengthening of the upper lip and an increase in naso-labial angle. It was also
noted that changes in the soft tissue profile following small mandibular setbacks
were less predictable compared to large setbacks and females demonstrated
greater soft tissue movement in response to skeletal repositioning compared to
males, this was statistically significant for the upper lip and chin (p<0.05). Lastly
changes in facial aesthetics following orthognathic surgery were highly
dependent on skeletal stability of the surgical procedure (222).
Interestingly two recent systematic reviews have been published by Joss et al.,
2010 on the effects of soft tissue profile following mandibular surgery for
mandibular setbacks (186) and mandibular advancement (223). The review
concluded that the current evidence-based analysis of soft tissue changes
following orthognathic surgery is poor. This is mostly due to inherent problems
of the retrospective nature of the published data, poor study designs and lack of
standardized outcome measurements. However the review did report the ratio
of soft to hard tissue movement was around 1:1 in chin area especially for the
Anas Almukhtar 2016
57
Chapter One
Review of Literature
mentolabial fold, whilst there was greater variability with the lower lip which
ranged from 1:1 for mandibular setbacks and 1:2 in mandibular advancements
and to a less extent in the upper lip region.
No published studies were found stating A-P, lateral or vertical ratios in 3D. This
was probably as a result of the better understanding of the 3D nature of the
facial changes following orthognathic surgery which made it completely
irrelevant to describe surface change with a single numerical ratio.
1.4 Prediction of soft tissue changes following
orthognathic surgery.
At the start, computer aided surgery with the ability to predict soft tissue
changes were confined to profile view only. Common software packages were
CASSOS and Dolphin. These two software packages were based on interpolating
the published 2D ratios to create continuous profile contour changes. Jones et
al., 2007 (224) validated the accuracy of the profile prediction using CASSOS
software. Their findings stated that the prediction was within acceptable
accuracy with a higher range of error around the lips regions especially in the
vertical dimension.
In order to achieve a successful three dimensional prediction of the facial soft
tissue changes following orthognathic surgery a well-developed mathematical
model is required that can mimic the actual deformation behaviour of the facial
soft tissues as a result of skeletal displacement. These are known as
approximation models (225).
Various models have been proposed for this function (225–227). These include:
Geometrical analysis models; Finite Element Model; Mass Spring Model and Mass
Tensor Model.
Anas Almukhtar 2016
58
Chapter One
1.4.1 Geometrical analysis models
Review of Literature
In these models the displacements of soft tissue vertices were estimated in
relation to the movements of its corresponding hard tissue vertices (228,229).
Xia et al., 2000 (229) introduced a soft tissue simulation model built on a purely
geometrical analysis. The introduced algorithm was based on transferring the
skeletal movement to the soft tissue. The procedure utilized the available 2D
hard-soft tissue ratios and applied them to different regions on the face.
According to their method, two algorithms were adopted: Surface Normal-based
Model Deformation Algorithm and Ray Projection-based Model Deformation
Algorithm. The first was based on the directions of the normal (perpendicular)
at each vertex to allocate hard-soft tissue correspondence. This was applied to
the chin area (from labiodental fold and down). A variation from this algorithm
was based on an imaginary ray passing through the hard tissue to the soft tissue
in a radial fashion. This was applied to the upper two thirds of the face
(labiomental fold and up). This type of simulation provides a low computation
time combined with a photorealistic representation of facial tissue changes.
Recent well known software packages such as Dolphin 3D uses an algorithm not
so far from this type for prediction planning. However not enough publications
validating the accuracy of the prediction planning were found.
1.4.2 Finite Element Model
The finite element model (FEM) was first introduced to help engineers to
understand the reaction of different materials to various external forces such as
the reaction of a bridge to applied loads from crossing vehicles. The principal
idea was to divide a continuum into smaller elements known as finite elements
to simplify their analysis such as converting a circle into a hexagon (6 elements)
or a pentagon (8 elements) or more elements until a satisfactory level of
simplification was reached in balance with preservation of the geometrical or
physical properties of the tested material. In 3D surgical simulation, the reaction
of the face as a continuum was simplified by the polygonal mesh construction
where each vertex represent an element across which force action and reactions
Anas Almukhtar 2016
59
Chapter One
Review of Literature
were individually analysed based on a fixed coefficient related to the type of
the tissues involved (230). FEM was one of the first deformation models that has
been extensively applied to orthognathic surgery prediction planning (228,231).
Keeve et al., 1998 (230) introduced and explained his idea of using FEM in 3D
planning. In their approach they used a grid of six node prisms. The prism
elements are defined by their corner nodes, as shown in figure (14). In their
displacement-based finite-element model, given displacements are specified for
certain nodes on the bone surface. These displacements cause stress which in
turn creates internal strain forces. To bring these forces into equilibrium, a
system of differential equations was developed delivering the displacements of
the unconstrained nodes of the soft tissue surface (230).
FEM has been shown to give a reasonably accurate simulation of tissues affected
by maxillofacial surgery (230). However, The high computational power and
memory consumption were the main short coming (227,232).
1.4.3 Mass Spring Model (MSM)
An alternative deformation model termed Mass Spring Model (MSM) has been
developed to reduce the processing labour (233,234). The basic idea is to
connect each element (usually one triangle) on the skeletal surface with a
corresponding element on the soft tissue surface using either ray projection or
closest point algorithm. This connection is represented by a spring at equilibrium
state in the pre-operative model. Moving the skeletal tissue will compress or
decompress the spring producing stress that will be transferred to the soft tissue
surface as the spring trying to rebound to equilibrium state. The amount of soft
tissue displacement depends on the stiffness of the spring and its relation to
adjacent elements. As the facial soft tissue is composed of layers of different
anatomical structures including muscles, facia, and fat pad, mass-spring
algorithm has the ability to incorporate more than one spring with variable
stiffness for each tissue type which increases its accuracy in soft tissue
simulation, (figure 15). This approach has a considerable advantage over the
Anas Almukhtar 2016
60
Chapter One
Review of Literature
finite element analysis. Most importantly is its ability to combine a large number
of ‘elements’, which results in a better modelling accuracy in addition to its
faster simulation time. However,
MSM
lacks the
biomechanical foundation
(227). An example of surgery planning software that is based on this algorithm is
“3dMD vultus”.
1.4.4 Mass Tensor Model
MTM defined by Cevidanes et al., 2011 (225) as “a mixture of the easy
architecture of the MSM and the biomechanical relevance of FEM”.
Cotin et al., (1999) (235) explored the possibility of a hybrid model that
combined both the FEM and MSM in an attempt to overcome the disadvantage of
the mass spring model lacking the biological foundation, which was later termed
the Mass Tensor Model (235). This deformation model provided the architectural
simplicity of the mass spring model augmented with the biomechanical
relevance of the finite element model. The processing time was greatly reduced
(227) which allowed a real time simulation using a standard PC for routine
clinical practice. An example of surgery planning software that is based on this
algorithm is “Maxilim” (236).
Schendel et al., 2013 reported a high accuracy of 3dMD Vultus prediction of the
post-operative soft tissue changes. An average difference of 0.27 mm between
the simulated and the actual soft tissue meshes was reported, the highest
difference was 0.6mm at the mental fold (237). However the methodology of
this study is questionable, the analysis included areas of the face that are not
affected by surgery, and these were considered in the calculations of the mean
changes of soft tissue following surgery. Anatomical regions were assessed by
measuring
distances
at
corresponding
landmarks
which
limits
the
comprehensiveness of the analysis.
Anas Almukhtar 2016
61
Chapter One
Review of Literature
Figure 14: Basic element of the finite element model. Note that the prism could not be
divided into layers. Keeve et al.,1998 (230)
Figure 15: Basic elements of the mass-spring model. Note that the prism could be divided
into layers. Keeve et al.,1998 (230)
Anas Almukhtar 2016
62
Chapter One
Review of Literature
1.4.5 Comparison of the different deformation models
Molleman et al., 2007 (227) compared the finite element model, Mass spring
model, and mass tensor model for accuracy and clinical versatility. The results
showed a superior outcome achieved with the mass tensor model in orthognathic
planning prediction accuracy. In general, mass tensor model and finite element
model predictions demonstrated the highest accuracy, but the mass tensor
model achieved the shortest processing time (227).
Anas Almukhtar 2016
63
Chapter One
1.5 Aims
Review of Literature
To introduce a new method for prediction of soft tissue changes following
orthognathic surgery.
1.6 Specific objectives
The following are the aims objectives and the hypotheses to be tested.
1. To identify the most reliable approach of 3D image superimposition with
the highest level of accuracy and to develop an informative method for
the 3D measurement of surgical skeletal movement following orthognathic
surgery.
2. To validate the accuracy of the conformation of the generic facial mesh
and to apply the “anatomical dense correspondence analysis” for facial
anthropometry.
3. To provide a statistical algorithm for prediction of soft tissue changes in
response to orthognathic surgery.
Hypotheses to be tested:
1. The conformation of the generic mesh followed by the application of the
dense correspondence analysis is valid for the assessment of facial
morphology.
2. The prediction algorithm is valid with a satisfactory level of accuracy.
3. The newly developed measurement for surgical movement is accurate and
reproducible.
Anas Almukhtar 2016
64
2
C
Methodology
ontents
INTRODUCTION .......................................................................................................................... 66
2.1
SECTION A: MAIN RESEARCH SAMPLE RECRUITMENT .................................................................. 68
2.1.1
SAMPLE ................................................................................................................................. 68
2.1.2
CBCT SCANNING PROTOCOL ...................................................................................................... 69
2.2
2.2.1
SECTION B: VALIDATION OF THE HARD TISSUE CHANGES AS A RESULT OF SURGERY ............................. 73
COMPARISON BETWEEN VOXEL BASED REGISTRATION AND SURFACE REGISTRATION TO ANALYSE CHANGES
FOLLOWING ORTHOGNATHIC SURGERY..................................................................................................... 73
2.2.2
DIRECT DICOM SLICE LANDMARKING, A NOVEL TECHNIQUE TO QUANTIFY THE DIRECTION AND MAGNITUDE
OF HARD TISSUE SURGICAL CHANGE. ........................................................................................................ 89
2.3
2.3.1
2.4
SECTION C: VALIDATION OF BASIC METHODS OF SOFT TISSUE ANALYSIS ......................................... 107
THE USE OF A GENERIC MESH TO ASSESS SOFT TISSUE CHANGES USING STEREOPHOTOGRAMMETRY. ...... 107
SECTION D: ANALYSIS OF SKELETAL AND SOFT TISSUE CHANGES FOLLOWING ORTHOGNATHIC SURGERY.. 131
2.4.1
PRE- ANALYSIS 3D IMAGE PREPARATION .................................................................................... 131
2.4.2
MEASUREMENT OF HARD TISSUE DISPLACEMENT FOLLOWING SURGERY ........................................... 134
2.4.3
ANALYSIS OF SOFT TISSUE CHANGES FOLLOWING SURGERY............................................................. 140
2.4.4
SIMULATION OF SOFT TISSUE FOLLOWING ORTHOGNATHIC SURGERY ............................................... 146
Anas Almukhtar 2016
65
Chapter Two
Methodology
Introduction
This study relies on the use of two different imaging modalities. The first is
volumetric conebeam CT imaging technology that simultaneously captures both
the skeletal and soft tissue data of an individual.
The second is
stereophotogrammetry which captures only surface data.
Using appropriate software it is possible to separate or segment the skeletal and
soft tissues captured within the 3D volume of the CBCT scan whilst still
maintaining their relative positions to one another in 3D space.
It is also
possible to convert the volumetric data into surface data, thus producing a
common file format for CBCT and stereophotogrammetry.
To avoid repetition the materials and methods section has been subdivided into
the following sections.
Section A
Describe the recruitment of the main research sample.
Section B
Describe two experiments to validate the measurement of hard tissue changes as
a result of surgery.
Comparison between voxel based registration and surface registration to
analyse changes following orthognathic surgery.
Direct DICOM slice landmarking, a novel technique to quantify the
direction and magnitude of hard tissue surgical change.
Anas Almukhtar 2016
66
Chapter Two
Methodology
Section C
Describe a series of experiments to assess soft tissue changes. These include:
1. The use of a generic mesh to assess soft tissue change using
stereophotogrammetry.
2. The application of dense correspondence technique on the soft tissue
surface data of the CBCT scan to achieve the following:
a) Construct an average face.
b) To quantify the direction and magnitude of soft tissue change following
surgery.
Section D
Describe skeletal and soft tissue measurements and quantification of the
relationship between the soft tissue changes and the hard tissue changes as a
result of surgery; in effect combining the above sections.
Anas Almukhtar 2016
67
Chapter Two
Methodology
2.1 Section A: Main research sample recruitment
In this section, details concerning the recruitment of the study sample used in
the main project will be described.
2.1.1 Sample
This is a retrospective study based on the conebeam CT DICOM images of
orthognathic surgery patients who had been treated at the University of Glasgow
Dental Hospital and School and the Southern General Hospital. Ethical approval
to access and use the data was obtained from the West of Scotland Research
Ethics Service (Reference 12/WS/0133).
Patients were selected from the
database at the Dental School and cross-referenced with the Southern General
Hospital theatre list from 2008-2014. All patients were of Caucasian decent and
had been diagnosed with a dentofacial deformity that required correction by
orthognathic surgery. All the patients were treated by the same surgeon.
Preoperative and postoperative CBCT scans were taken for routine pre-surgical
workup prior to orthognathic surgery.
The preoperative CBCT scans were
acquired within one month of surgery and the postoperative scans were obtained
at a minimum 6 months after surgery using the same CBCT machine (i-CAT
Classic, Imaging Sciences, Hatfield, UK). The protocol for CBCT image capture is
discussed in details elsewhere (Section 2.1.2).
2.1.1.1 Inclusion criteria
Patient related
Patients were Caucasian ethnic background.
Patients 17 years of age or above with no further anticipated growth.
Patients underwent orthognathic surgery treatment, Table (3).
No history of previous operations involving facial soft tissue.
No history of previously treated dentofacial deformity including cleft lip
and palate.
Anas Almukhtar 2016
68
Chapter Two
Methodology
Conebeam CT related
All patients had pre and post-operative CBCT scans. The pre-operative
image must have been taken no more than one month pre-operatively and
the post-operative images within 6 to 12 months after surgery.
All conebeam CT scans were extended field of view (EFOV) (22 cm) scans.
Surgery related
Only patients who were treated with single piece Le Fort I and BSSO
osteotomies ± genioplasty were included.
2.1.1.2 Exclusion criteria
Defective images caused for various reasons. More than 20% of the original
collected sample was excluded due to this reason, (figure 16).
Failed to meet the specific inclusion criteria detailed above.
Patients treated with segmental osteotomies.
A total of 137 pre- and post-operative images were located. Only 100 images
were included according to the research sample inclusion criteria. Image defects
were the main reason for exclusion. Table (3) details the classification of
malocclusion and the surgical procedures undertaken.
2.1.2 CBCT scanning protocol
Standardised pre-operative and post-operative conebeam CT (CBCT) images
were taken for all patients using a Classic iCAT (i-CAT Classic, Imaging Sciences,
Hatfield, UK) 0.4mm voxel, 22cm Extended Field of View (EFOV) option. An
experienced radiographer in the Radiology Department at the Glasgow Dental
Hospital and School under took all the scans.
Prior to CBCT scanning the chin support on the iCAT scanner was removed and
replaced with a forehead band to avoid any chin soft tissue distortion. The
patients were positioned sitting upright with their back supported by the built-in
Anas Almukhtar 2016
69
Chapter Two
chair.
Methodology
Using the laser patient orientation system incorporated in the iCAT
scanner the patients were positioned with their Frankfort plane and interpupillary parallel to the horizontal line together with the vertical line in the
mid-sagittal facial line. The patient’s head was then secured to the headrest
with a securing band placed as high as possible on the forehead, (figure 17).
Patients were instructed to relax with lips in repose by saying “Mississippi”,
licking their lips and saying “N” and gently putting their teeth together
(Zacharrsion, 1998).
Patients who were overclosing, mainly due to vertical maxillary deficiency, had
had an intra-occlusal wax wafer constructed by the Clinician responsible for
their care prior to the scan; this was used during the scan and subsequent
planning. The patients were asked to remain still during the CBCT scan. The
images were saved in DICOM format and exported for later use.
2.1.2.1 DICOM Image anonymization
DICOM images incorporated patient indefinable information. This information
was not necessary for the current research and the data was anonymized. Two
online free software packages were used for this purpose “VTK DICOM
anonymiser” and “DICOM files renamer”. These two software packages were
capable of anonymising the images by removing any tagged information in
addition to renaming each slice with the case CHI number of the patient and a
unique patient code. An encrypted EXCEL worksheet (Microsoft®, Redmond, CA)
was created containing patients names and their corresponding CHI numbers and
codes. This was kept for future need of retrieving information regarding
individual patients.
Anas Almukhtar 2016
70
Chapter Two
Methodology
Table 3: Surgical jaw correction movements.
Surgical procedure
A-P corrections combinations
Number of cases
Le Fort I
Maxillary advancement
33
Le Fort I + Genioplasty
Maxillary advancement
19
Mandibular advancement
12
Mandibular setback
2
Mandibular advancement
5
Mandibular setback
1
BSSO
BSSO + Genioplasty
Maxillary and Mandibular
advancement
Combined (Bimaxillary)
Maxillary advancement and
Mandibular setback
Maxillary and Mandibular
advancement
Combined +
Genioplasty
Maxillary advancement and
Mandibular setback
Total
11
7
7
3
100
Table 4: Image pairs configurations
Patient
Registration
method
tissue specific image pair
Soft tissue (pre- and registered post-operative)
VBR
Patient X
SBR
Hard tissue (pre- and registered post-operative)
Soft tissue (pre- and registered post-operative)
Hard tissue (pre- and registered post-operative)
Total
Anas Almukhtar 2016
8 models
71
Chapter Two
Methodology
Figure 16: CBCT image defect. A step at the face was formed due to patient movement
during the CBCT.
Figure 17: CBCT image capture showing the patient sitting upright. The chin rest was
removed and the head strap was used instead.
Anas Almukhtar 2016
72
Chapter Two
Methodology
2.2 Section B: Validation of the hard tissue changes as a
result of surgery
In order to measure the changes as a result of orthognathic surgery the pre- and
post-operatives images need to be firstly superimposed on structures which have
remained stable and not changed as a result of surgery. It is common practice
to use the anterior cranial base in two-dimensional cephalometry (209,222) and
in 3D imaging (168,238). Other structures such as the zygomatic regions have
also been proposed (69) as an alternative.
Following superimposition the
changes in landmarks, representative of the structure being assessed, are
calculated and used as a measurement of the outcome for skeletal movement.
In this section, two experiments will be described; the first to objectively assess
two 3D image registration methods. The second to validate a novel technique
that measures the three dimensional skeletal displacement of the maxilla and
mandible.
2.2.1 Comparison between voxel based registration and surface
registration to analyse changes following orthognathic
surgery
2.2.1.1 Introduction
The two types of image registration are rigid registration methods and were
discussed earlier (Section 1.2.2). In summary, surface based registration (SBR)
was the initial method described for 3D image superimposition (10,11). The
principle involves approximating two surfaces using either Partial Procrustes
Analysis PPA or iterative closest point (ICP) or both of them.
Voxel based
registration (VBR) on the other hand utilizes the grey scale difference of the
voxels to align the two DICOM images to the best superimposition achieving the
least total grey scale density difference between the two images. Voxel-based
registration is useful were it is difficult to detect distinct surface topography
features or when the whole body of the registered structure needs to
superimpose regardless of the type of tissues involved.
Anas Almukhtar 2016
73
Chapter Two
Methodology
Studies reporting the use of voxel based registration have claimed high accuracy
in registration (16-18). However, to date, no research has been published
comparing the accuracy of voxel based and surface based registration methods.
The choice between voxel based registration and surface based registration was
vital for this PhD project. The availability of the image resources for both types
of registration provided a unique opportunity to experiment each of them and to
choose the type of image registration suitable for the project.
2.2.1.2 Aim
The objective of this study was to determine if there was a statistically
significant difference in the accuracy of image superimposition between two
registration methods i.e. surface based and voxel based.
2.2.1.3 Methods
2.2.1.3.1 Sample
The study sample composed of the pre- and post-operative CBCT images of 31 of
orthognathic surgery patients. These images were randomly selected from the
total 100 previously described (Section 2.1.1).
All the patients had had
orthognathic treatment to correct their facial deformity. The preoperative CBCT
scans were acquired within one month of surgery and the postoperative scans
were obtained at a minimum 6 months after surgery.
2.2.1.3.2 Voxel based registration
The principals of voxel based registration were explored in details (Section
1.2.2.3). In summary, voxel based registration (VBR) is a specialised method of
DICOM images superimposition. The associated algorithm utilises the grey scale
intensity of the voxels composing the DICOM image to produce the best match
between two overlapping images through an automated iterative move and
measure sequence. The number of iterations, the percentage of voxels involved
and the region of interest were manually selected, (figure 18).
Anas Almukhtar 2016
74
Chapter Two
Methodology
The pre- and post-treatment DICOM images for each patient were imported into
Maxilim software (Medicim-Medical Image Computing, Belgium). Using the voxel
based registration add-on module provided by the manufacture a region of
interest including the anterior cranial base and part of the forehead, was
selected on the pre-treatment and post-treatment images. This region would be
used for image registration; the number of iterations was set to 30 to ensure
maximum alignment. The registration process took approximately 20 minutes to
complete and the registered DICOM images were saved in their new positions for
further analysis.
Following registration of the pre- and post-operative volumetric DICOM images
for each patient, the soft and hard tissue 3D surface models were segmented
using Maxilim software. Details of the segmentation process are previously
provided (Section 1.2.1). The segmented hard and soft tissue models were
exported as (.stl) files and saved for the next step of the analysis.
2.2.1.3.3 Surface based registration
The principals of surface based registration were discussed in details in Section
(1.2.2.2). In summary, the method is a generalized 3D image registration
technique used for superimposition of two 3D surface mesh models. The method
has been extensively used for engineering, medical, security and military
applications, (239,240). In this study, the automated Iterative Closest Point (ICP)
algorithm was used in addition to a preliminary manual landmark based
registration (PPA) as a preparatory step. The (ICP) algorithm searches’ for a
match of the topographic features of the overlapping surface meshes where the
square root distance of the involved vertices with the adjacent mesh are
minimised. This is achieved in an iterative move and search fashion.
The soft and hard tissue models of the unregistered pre- and post-operative
DICOM images were segmented using Maxilim software and were exported as STL
files. These models were then loaded into VRMesh software (Virtual Grid, Seattle
City, U.S.A) and arranged in two groups: pre-operative group (soft and hard
tissue models) and post-operative group (soft and hard tissue modes). An
Anas Almukhtar 2016
75
Chapter Two
Methodology
additional group was created and named template which was an exact copy of
the pre-operative group. Soft and hard tissue models in the template group were
cropped to the region of interest only (stable region unaffected by the surgery)
to act as the bases for the surface registration procedure. The cranial base
extended to involve the frontal bone, and the forehead extended to involve the
eyes were the regions of interest for the hard and soft tissue models
respectively, (figure 19). Since the templates were created from a copy of the
pre-operative models, they were in the same 3D position as the full preoperative model. Following this, manual registration (PPA) was carried out as a
preliminary step to approximate the models. All the registration steps involved
rigid registration with no scaling. For manual registration three landmarks were
on the hard tissue template; left and right zygomatico-frontal sutures and
nasion. The corresponding landmarks were placed on the post-operative model.
The same procedure was performed for the soft tissue template and the postoperative image using left and right exocanthian and nasion. This procedure
approximated the two images and provided a closer starting position for the next
automated step. The post-operative models (source) were always superimposed
by rotation and translation on to the template (target).
To further improve the alignment of the post-operative image to the template
the (ICP) function within VRMesh was used. The iterations were set at 500 and
50% of the vertices in the region were involved in the ICP registration process.
The full procedure took an average 17 minutes for each case including template
construction. The template was then deleted and the original pre-operative
image re-imported. As the template was based on the pre-operative image and it
did not move during alignment, as it was the target image, the post-operative
images were aligned on target regions of the pre-operative image. Both the
registered pre- and post-operative (soft and hard tissue) models were exported
and saved for further analysis.
Anas Almukhtar 2016
76
Chapter Two
Methodology
A
B
Figure 18: Voxel based registration. Selecting the volume of interest (A); the registered
images (B), soft (left) and hard (right).
Anas Almukhtar 2016
77
Chapter Two
Methodology
A
B
Figure 19: The registration template. Soft tissue (A) and Hard tissue (B) .
Anas Almukhtar 2016
78
Chapter Two
Methodology
2.2.1.3.4 Analysis of registration accuracy
A standardised region for analyses was chosen for all the image pairs on the
forehead and anterior cranial base using VRMesh software, (figure 20). The
inferior boundary of the region was determined by a horizontal plane passing
through right and left outer canthi; the superior boundary was denoted by a
horizontal plane parallel to the inferior limit and located 20 mm above glabella;
the posterior boundaries were limited by a coronal plane passing through sella.
To insure standardisation of the region, the eight models for each patient were
loaded in the same group prior to boundary selection and the same set of planes
were used to croup all the models in the group.
These 8 models include two pairs for each registration method: VBR pre- and
registered post-operative (soft and hard tissue models) and SBR pre- and
registered post-operative (soft and hard tissue models). Isolation of the region
was performed for all the loaded images simultaneously using the same cutter
planes and as stated earlier above. At the end of the process, the eight models
were exported and saved as VRML files for further analysis.
To perform the analysis, the images were loaded into in-house developed
software. The software measured the nearest distance between two points on
two adjacent meshes and produced the mean, standard deviation, maximum and
minimum distances between them. The images were loaded pair wise (pre- and
post-operative) as shown in table (4).
The results (mean, standard deviation, maximum and minimum distance) were
exported as an EXCEL file (Microsoft®, Redmond, CA). Only 90% of the vertices
were included in the analysis. This insures exclusion of the outliers due to
erroneous data, which may have been attached to the isolated regions.
In addition to these measurements, the software provided a visual output of the
distance measurements using a colour coded map and the distribution of the
Anas Almukhtar 2016
79
Chapter Two
Methodology
measurements around the mean. These were saved as jpeg image files for visual
observation only, (figure 21).
2.2.1.3.5 Statistical analysis
A paired Student t-test was used to determine if there was any statistical
difference between SBR and VBR for pre- and post-operative images for both soft
and hard tissues selected regions (p<0.05). An ANOVA and post-hoc Duncan test
was used to detect any significant difference between each method of
superimposition and tissue type i.e. hard or soft tissue. A Pearson correlation
coefficient was used to test the correlation between superimpositions for the
four groups (SBR hard, SBR soft, VBR hard and VBR soft). A one sample t-test was
used to test if the absolute mean difference between the post-operative soft
tissue 3D models aligned by each registration method was greater than 0.5 mm.
2.2.1.4 Results
Figure 22 shows the descriptive analysis of the four superimpositions groups.
The four superimpositions were ranked from the lowest to the highest absolute
mean distances between corresponding 3D meshes. Voxel based registration and
surface based registration of the hard tissues showed the same values in the
absolute mean distances between the models, 0.05 mm ± 0.21mm and 0.05 mm
± 0.26mm respectively.
For soft tissue superimposition the absolute mean
distances between the meshes were larger on the voxel based registration than
that on surface based registration, 0.29 mm ± 0.33mm and 0.23 mm ± 0.56mm
respectively.
For both hard and soft tissue the paired Students t-test showed no statistically
significant difference between the two superimposition methods, (table 5).
Anas Almukhtar 2016
80
Chapter Two
Methodology
Figure 20: Standard region for analysis. Note that the soft and hard tissue of the two images
were loaded and cropped at the same time.
Figure 21: output colour map -registration accuracy
Anas Almukhtar 2016
81
Chapter Two
Methodology
A one way ANOVA and post hoc Duncan test were used to investigate the
statistical significance of the differences between any pair of the four groups
(SBR hard, SBR soft, VBR hard and VBR soft). The result of ANOVA test showed a
statistically significant difference between the four groups. The post hoc Duncan
test showed that the type of tissue i.e. hard or soft tissue influenced the
accuracy of superimposition using either surface based or voxel based
registration methods. A statistically significant difference was found between
superimposition of the soft and hard tissue models within the same method. The
difference between the VBR hard and VBR soft superimpositions was statistically
significant (p<0.001); the absolute mean difference was 0.23 mm, (table 6).
However, the difference between SBR hard and SBR soft was not statistically
significant (p=0.712).
Statistical correlation between different groups was analysed using a Pearson
correlation test, (table 7). VBR hard and SBR hard superimpositions showed a
strong positive correlation (r = 0.886). VBR soft and SBR soft showed a weak
positive correlation (r = 0.126). This implies that the superimposition of the hard
tissue did not show variability between the two methods whereas the soft tissue
superimposition showed high variability.
Anas Almukhtar 2016
82
Chapter Two
Methodology
Table 5: Paired sample t-test to compare methods accuracy for each tissue type.
Pair 1
SBRhard - VBRhard
-.01
-.06
-.01
.02
.01
Sig.
(2tailed)
.39
Pair 2
SBRsoft - VBRsoft
-.18
.05
.28
.05
0.54
.24
Lower
Uppe
r
Mean
SD
SE
Table 6: Paired sample t-test to compare the accuracy between different tissue types.
Pair
1
Pair
2
Lower
Upper
Mean
SD
SE
Sig.
(2-tailed)
sbrHard - sbrSoft
-0.42
0.29
-0.06
0.96
0.17
0.712
vbrHard - vbrSoft
-0.31
-0.16
-0.23
0.21
0.04
0.000
Table 7: Pearson correlation analyses showing ‘correlation coefficient’ and ‘significance’
between different tissue types and methods.
SBRhard
VBRhard
SBRhard
VBRhard
SBRsoft
VBRsoft
Pearson Correlation
1
0.886**
0.190
0.102
Sig. (2-tailed)
X
0.000
0.343
0.613
N
27
27
27
27
Pearson Correlation
0.886**
1
0.126
0.182
Sig. (2-tailed)
0.000
X
0.532
0.363
27
27
27
27
Pearson Correlation
0.190
0.126
1
0.126
Sig. (2-tailed)
0.343
0.532
X
0.532
27
27
27
27
Pearson Correlation
0.102
0.182
0.126
1
Sig. (2-tailed)
0.613
0.363
0.532
X
27
27
27
27
N
SBRsoft
N
VBRsoft
N
**. Correlation is significant at the 0.01 level (2-tailed).
Anas Almukhtar 2016
83
Chapter Two
Methodology
The one sample t-test showed the absolute mean difference between the preand post-operative soft tissue position when VBR was used to align soft tissue
images or SBR was used was not statistical greater than 0.5mm (p = 0.73). The
clinical significance was determined to be 0.5mm from a previous study (20).
2.2.1.5 Discussions
This study aimed to evaluate the accuracy of voxel based registration compared
to surface based registration method and to determine if the difference between
them is statistically significant. Accordingly, the research method was based on
31 pairs of preoperative and postoperative CBCT scans of patients treated by
orthognathic surgery. The study investigated the accuracy of both methods in
registering the postoperative image to the corresponding preoperative images.
Despite the fact that both methods of registration use the information provided
by a CBCT generated DICOM image, voxel based registration deals with the raw
information of the DICOM image by comparing the grey scale intensity of the
voxels composing the corresponding DICOM images; on the other hand, surface
based registration requires an extra step involving 3D model rendering to
generate a three dimensional surface mesh model, on which the surface based
registration is performed. This additional step may introduce a possible source of
error since the algorithm used for segmenting the 3D model depends on
Hounsfield value (HU value) of DICOM images of the CBCT. The form and
dimension of the 3D surface model is dependent on the HU value (19) which in
turn may be affected by image quality and tissue density. In addition, this extra
step increases processing time and implies the need for multiple software
packages for 3D model rendering which is unnecessary in the case of voxel based
registration.
Another parameter worth considering when comparing the two methods is the
amount of information utilised for the registration purpose. Surface based
registration uses the 3D information provided by surface mesh topography of the
3D model. Whereas voxel based registration uses the grey scale values of all the
Anas Almukhtar 2016
84
Chapter Two
Methodology
voxels embedded in and around the anatomical structure and is not dependent
upon surface features. In other words, surface based registration deals with the
“shell” covering the 3D structure while the voxel based registration deals with
all the contents of the volume selected which may theoretically increase the
accuracy of the method. However the use of such information implies the need
for a more efficient computers and a longer processing time (16).
Despite the fact that both methods use the ICP algorithm for superimposition,
which involves repetitive translation-rotation movement and measurements
between the two 3D objects to reach the best matching superposition, the two
approaches are considerably different. Surface based registration applies an
estimation of the optimal translation and rotation between the three
dimensional shapes by minimizing the mean square distance between the
surfaces. The distance is measured between a specified percentage of the points
randomly selected on one 3D mesh and the corresponding 3D surface mesh.
Unlike voxel based registration in which the estimation of the optimal
translation and rotation between the 3D volumes is determined by the mean
square difference in the grey scale intensity between a specified percentage of
voxels randomly selected on one image volume and the overlapped voxels in the
corresponding one.
Loss of the sharpness of a 3D image during capture may be a source of error due
to confusion in the estimation of the grey scale level of the voxels and therefore
registration. However, the degree of DICOM image sharpness has a similar effect
on the surface based registration but indirectly and may not be detected due to
the automatic surface smoothing of the image. The accuracy of 3D model
segmentation from DICOM image is affected by the quality of the DICOM image.
In other words, the algorithm will have to decide where to place the boundaries
of the hard tissue when building a skull model from a DICOM image with loss of
sharpness and the resultant 3D model will represent the estimated dimensions
rather than the original.
Anas Almukhtar 2016
85
Chapter Two
Methodology
Four of the samples used in this study were considered as outliers with values
reaching up to 6 times the general attitude of the sample and introducing errors
by significantly changing the mean values of all of the superimposition groups.
They were excluded from the study sample for this reason, (figure 23).
In all cases, surface based registration demonstrated a higher variability in
superimposition as indicated by the larger standard deviation, (figure 22). This
may be due to the SBR algorithm relying on well-defined surface features for
registration which are present on the hard tissue but are not a prominent
feature of the relatively homogenous surface of the soft tissue forehead.
With
respect to VBR registration, the distribution of the voxel’s grey scale intensity
was thought to be the reason for a lower variation in the superimposition
process, which was reflected as a lower standard deviation.
Further investigation using Pearson correlation coefficient test was carried out
to observe the correlation between different registration methods within each
pre- and post-operative data set. A strong positive correlation (r = 0.886) was
found between the hard VBR and SBR of the hard tissue models There were
weak positive correlations among all other groups of the study. This result
highlighted two important observations; firstly, surface based registration for
hard tissue was as accurate and consistent as the voxel based registration. A
possible explanation may be the high level of feature specific information
available on the hard tissue surface which improves the performance of surface
based registration.
The relatively smooth surface of the soft tissue model
reduces the accuracy of the registration and increases the variability of the
results. On the other hand, voxel based registration relies on the grey scale
intensity of the DICOM image voxels rather than the soft and hard tissue model
surface features topography which makes it more consistent in both regions.
Anas Almukhtar 2016
86
Chapter Two
Methodology
0.561
0.6
0.5
0.4
0.334
0.294
0.259
0.3
0.206
0.230
0.2
0.1
0.050
0.047
0
mean
SD
sbr hard
mean
SD
vbr hard
mean
sbr soft
SD
mean
SD
vbr soft
Figure 22: Comparison of the accuracy of the two 3D image registration methods. Note that
the standard deviation is higher at surface based registration of the soft tissue models.
7
6
5
SBR-soft
Series1
4
VBR-soft
Series2
3
SBR-hard
Series3
2
VBR-hard
Series4
1
0
1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031
Figure 23: Excluded cases. Graphic representation of the error values showing the reason
to exclude some of the study sample.
Anas Almukhtar 2016
87
Chapter Two
Methodology
The other finding was the weak positive correlation between the soft and hard
tissue models registration using voxel based registration (r = 0.126). Unlike
surface based registration, the voxel based registration algorithm translates and
rotates all the tissues captured in the DICOM image simultaneously. Hence, a
strong correlation would be expected between the soft and hard tissue models
alignment measurements.
This result may be explained by the effect of
variation of facial expression during the preoperative and postoperative image
capture and the possibility of soft tissue thickness change as a result of weight
changes in the time interval between the two scans. The fact that the voxel
based registration algorithm relies on the grey scale intensity of the entire
image may result in excluding these small differences in soft tissue contour as
outliers during the registration process. This finding suggests that voxel based
registration produces a more accurate representation of soft tissue changes as a
result of surgery over surface based registration. Surface based registration of
the soft tissue aligns the pre and post-operative images irrespective of the
underlying hard tissue and therefore will “force” the two surfaces as close as
possible; whilst VBR may be restrained by the underlying hard tissue since it is
involved in the registration process. The differences between the two methods
of registration are unlikely to have any clinical significance.(20)
2.2.1.6 Conclusions
No statistically significant differences were detected between the voxel based
and surface based registration methods. However, voxel based registration
showed more consistency in representation of the actual soft and hard tissue
positions as indicated by lower mean standard deviation. Soft tissue surface
based registration does not take into account changes in tissue thickness.
Anas Almukhtar 2016
88
Chapter Two
Methodology
2.2.2 Direct DICOM slice landmarking, a novel technique to
quantify the direction and magnitude of hard tissue surgical
change.
2.2.2.1 Introduction
Numerous studies have reported on the three-dimensional changes of the
skeletal hard tissue following orthognathic surgery, (Section 1.3.1).
Previous
methods used include 3D surface landmarking (137), colour coded distance maps
(241) and volumetric changes (159). Each of these methods has deficiencies that
limit their clinical application.
For instance, changes in landmark position
indicate the change of one point rather than a complete 3 dimensional
structure, which can move with six degrees of freedom and volume changes, are
not indicative of positional changes.
To address these shortcomings this study was designed to assess the accuracy
and reproducibility of a novel method, based on landmarking DICOM image
slices, to quantify the three-dimensional positional change of the maxilla and
the mandible following orthognathic surgery.
To assess the accuracy of landmarking, true physical measurements were
established by simulating 16 orthognathic surgery osteotomies on a plastic skull.
This part of the project was carried out in collaboration with the University of
Hong Kong. Measurements were directly recorded from a plastic skull for each
simulated jaw movement and were considered the “gold standard” for
comparison with the proposed internal 3D landmarking method.
2.2.2.2 Methods
2.2.2.2.1 Sample
Simulated surgical procedures involving Le Fort I osteotomies and bilateral
sagittal split osteotomies were carried out on a plastic skull. Sixteen different
combinations of vertical and antero-posterior (A-P) movements were performed,
(table 8). The spatial coordinates of three landmarks on the maxilla and five
landmarks on the mandible were recorded and the A-P and vertical
Anas Almukhtar 2016
89
Chapter Two
Methodology
displacements at each of these landmarks were calculated. In total 108 (68 for
maxilla and 40 for mandible) measurements were produced.
This was a collaborative study established with the University of Hong Kong,
China. Thus, the method was carried out in two main parts: orthognathic surgery
simulation and physical measurements (The University of Hong Kong) and digital
image measurements (Glasgow University).
Table 8: Combinations of simulated surgery movements
Anterior-Posterior (AP) (mm)
Vertical downgraft (mm)
Mandible
2, 4, 8 and 10
0
Maxilla
0, 3, 6 and 9
0, 3, 6 and 9
0, 3, 6 and 9
2
4
8
2.2.2.2.2 The physical measurements (Gold standard measurements)
2.2.2.2.2.1 3D surgical simulation setup
A specific device was constructed for the purpose of simulation of the surgical
osteotomies and displacements. The device composed of a plastic skull which
was secured to a universally adjustable camera mount fixed to a 20mm thick
acrylic base. The level of the acrylic base could be adjusted to make sure it was
horizontal using a spirit level. This arrangement allowed the skull to be rotated
left and right, tilted up and down and tipped side to side, (figure 24).
An adjustable stage was secured to the acrylic block immediately below the
occlusal plane of the maxilla. A height adjustable platform was fabricated
relying on spacers of known height to adjust the vertical position of the maxilla.
The adjustable stage constrained movement of the maxilla in the sagittal
direction only, whilst the adjustable platform controlled vertical movement only
by removing or adding “spacers” of known thickness. This ensured the maxilla
could only be moved in two directions with no or minimal rotation.
Anas Almukhtar 2016
90
Chapter Two
Methodology
2.2.2.2.2.2 Physical measurements
Using the 3D surgical simulation setup described above, a “locating mask” was
produce that allowed re-location of the maxilla back to the skull following
separation. A line was drawn representing the future Le Fort I osteotomy cut.
Two holes were drilled on the right and left sides above and below the
osteotomy cut.
Four 3 mm stainless steel screws were used to secure the
locating mask to the skull base. The locating mask was removed and the maxilla
was detached from the skull by carrying out a Le Fort I osteotomy. Three 5 mm
diameter spherical plaster markers were secured using sticky wax to the maxilla
at the right greater palatine (GPR), left greater palatine (GPL) and the incisive
foramina (IF). Using the locating mask the maxilla was secured to the skull using
sticky wax; after which the locating mask was removed.
To assess mandibular changes, a bilateral sagittal split osteotomy (BSSO) was
performed on the plastic mandible and reassembled using two 3mm stainless
steel screws per side. As in the maxilla, 5 plaster sphere markers were placed
on the mandible using sticky wax; left and right lingual foramen (LL and RL), left
and right mental foramen (LM and RM) and genial tubercle (Ling).
A standard Dentatus face bow (Dentatus International AB, Sweden) was fitted to
the skull; securing the acoustic meatus with laboratory putty and resting the
orbital pointer in the right orbit. A circular spirit level was placed on the
anterior region of the face bow which allowed orientation of the skull into a true
horizontal position based on the Frankfort plane and parallel to the acrylic base.
The skull was secured in this position.
The 3D surgical simulation setup with face bow in place was positioned in the
cone beam CT (CBCT) scanner (iCAT, Imaging Science, Hatfield) so it was
horizontal based on the spirit levels. The face bow was removed immediately
prior to the 22 cm Extended Field Of View (EFOV) scan at 0.4mm voxel
resolution being performed. This baseline scan horizontally orientated the skull
within the 3D scan volume and was saved as a DICOM file.
Anas Almukhtar 2016
91
Chapter Two
Methodology
2.2.2.2.2.3 Simulated maxillary and mandibular movements
As the skull needed to be removed from the acrylic base to carry out the
simulated surgical movements six pre-marked reference points were used to
make sure it was correctly re-positioned on the acrylic base. The six pre-marked
points were nasion, landmarks on the right and left zygomatic buttresses,
landmarks on the maxillary left and right molar region and an additional one
between the central incisors. The landmarks on the molars and in between the
incisors were also used to measure the simulated maxillary movements. Sagittal
measurements were taken using a Vernier caliper mounted perpendicular to the
acrylic base. Vertical measurements were recorded using a vertical height
caliper (Chesterman, Sheffield, UK).
Following removal of the mandible, re-attachment, correct re-alignment of the
skull and base line measurements, the maxilla was secured to the adjustable
platform using sticky wax; and released from the base of the skull. Using the
adjustable stage and platform the maxilla was moved to the desired position, resecured to the base of the skull with sticky wax and then released from the
adjustable platform. Duplicate measurements of the positions of the reference
points before and after simulated maxillary movement were taken five times by
the same operator. Cone beam CT scans using the EFOV option and 0.4mm voxel
resolution were taken at 0, 3, 6 and 9 mm maxillary advancement; together with
a 2, 4 and 8 mm “downgraft” for each 3mm increment of maxillary
advancement. In total, 12 different positions of the maxilla were recorded and
each scan saved as a DICOM file.
For simulated mandibular movements the mandible was re-attached to the skull
and the lower border of the mandible secured to the adjustable platform.
Following removal of the screws the anterior segment of the mandible was
Anas Almukhtar 2016
92
Chapter Two
Methodology
A
B
Figure 24: 3D Surgery simulation and measurement setup: Lateral view (A), frontal view (B)
Anas Almukhtar 2016
93
Chapter Two
Methodology
translated forward and re-secured with sticky wax to the maxilla. Cone beam
CT scans following the previous protocol were taken at 4, 6, 8 and 10mm
mandibular advancements.
To accurately determine maxillary and mandibular movement all DICOM image
was converted to a surface mesh using MeVisLab (MeVis Medical Solutions Ltd.,
Germany) with surface value of 600 and resolution 1 and saved in STL format.
The original baseline skull STL file was loaded into VRMesh and all the remaining
STL files were aligned to this baseline skull using the anterior cranial base as a
common area of superimposition. Each image was re-saved in its new aligned 3D
position. Using Minimagics (Materialise, Belgium) it was possible to import any
two registered STL files, create a profile of both images and then measure the
sagittal and vertical distances between the two profiles at any point. This was
performed for all maxillary and mandibular movements.
2.2.2.2.3 Digital measurements
Measurements were made using OnDemand3D software (Cybermed, Seoul, South
Korea). The landmarking procedure involved three steps: pre- and postoperative DICOM image superimposition; 3D image orientation and creation of
reference planes (x, y and z planes) and lastly a modified 3D cephalometric
analysis of the orthogonal measurements of 8 landmarks placed on the DICOM
image slices. These steps were followed by calculation of the actual threedimensional movement of each landmark due to changes in jaw position.
2.2.2.2.3.1 DICOM image superimposition
The pre- and post-operative CBCT scanned DICOM images were imported into
OnDemand3D software. Superimposition of the two images was accomplished
using voxel-based registration. This involved two steps: manual alignment of the
two images followed by automatic registration. The registration process was a
series of iterative movements aimed at achieving the “best fit” based on the
grey scale intensity between the two overlapping images, voxel by voxel(69).
The region of interest for superimposition was the anterior cranial base as this
was a stable and unaffected by surgery (242).
Anas Almukhtar 2016
94
Chapter Two
Methodology
2.2.2.2.3.2 Creation of reference planes
Following superimposition of the pre-and post-operative DICOM images a
common reference plane could be constructed. Three reference planes were
created, an axial plane based on left and right porion and right orbitale), a
sagittal plane based on nasion and sella points and oriented perpendicular to the
horizontal plane, and lastly, a coronal plane was created perpendicular to both
the previous planes at the sella point. The x, y and z co-ordinates of any
landmark placed on any of the DICOM slices could then be extracted
orthogonally with reference to these three planes, (figure 25).
2.2.2.2.3.3 Orthogonal measurements
In total 8 landmarks, 3 in the maxilla and 5 in the mandible were placed. The
positions of these landmarks were designed to assess the three dimensional
orientation and movements of maxilla and mandible. Each landmark was placed
at the centre of the spherical plaster ball placed on the skull prior to scanning as
described above. To facilitate onscreen landmark placement, the centre of each
sphere was identified by simultaneously viewing the sphere on the DICOM slices
in the three dimensions (sagittal, axial and coronal), (figure 26).
Orthogonal measurements of each landmark on the pre-operative slice images to
the common reference planes were recorded and the x, y and z coordinates of
the 8 landmarks were exported to Microsoft EXCEL (Microsoft®, Redmond, CA) for
further analyses. The same procedure was repeated for the post-operative slice
images of each of the simulated movement.
Anas Almukhtar 2016
95
Chapter Two
Methodology
Figure 25: Reference planes (Ondemand3D software). Fine tuning of the reference planes is
possible through changing the three dimensional position using the panel on the right side.
Anas Almukhtar 2016
96
Chapter Two
Methodology
2.2.2.2.4 Error study
The inter and intra-operator reliability of landmark digitisation was assessed by
the same operator re-digitising the same points two weeks apart and a second
independent operator digitising the landmarks. Inter- and intra-examiner
landmarking errors were evaluated by analysing the differences between the
repeated readings using a one sample Students t-test and Interclass Correlation
Coefficient (ICC) (SPSS Version 22, IBM). Inter- and intra-examiner Euclidean
landmarking distances errors were calculated using the formula below, equation
(2).
Equation 2: 3D Pythagoras equation where D is the Euclidean distance and x, y and z are
the linear measurements in the three respective dimensions
2.2.2.2.5 Clinical application test
The landmarking method was carried out on 5 randomly selected cases.
Landmarks were successfully placed on the DICOM slices at the same anatomical
locations used in this validation project and displacement measurements were
calculated. The same procedure repeated two weeks later and measurements
were compared using paired Students sample t-test.
2.2.2.2.6 Statistical analysis
The absolute distance between the pre- and post-operative positions of each
landmark in x, y and z dimensions from the DICOM data were compared to the
measurements obtained from the simulated orthognathic surgeries using a one
sample t-test, Interclass Correlation Coefficient (ICC) test and Bland-Altman
plot.
Anas Almukhtar 2016
97
Chapter Two
Anas Almukhtar 2016
Figure 26: Landmarks digitization. The full set of landmarks was individually digitized at the centre of the indicator. The
presence of the 3D viewer on the right hand side enhanced the accuracy of landmarking. Starting from the top left Incisive
foramen (IF), Grater palatine Right (GPR) Greater palatine (Left), Lingual foramen Right (LR), Lingual foramen Left (LL), Mental
foramen Right (RM), Mental foramen Left (LM),and Genial tubercle (Ling).
Methodology
98
Chapter Two
Methodology
2.2.2.3 Results
2.2.2.3.1 Inter- and intra-examiner’s landmarking error
The magnitude of the intra-examiner and inter-examiner landmarking errors are
shown in table (9). The mean landmarking distance errors were 0.35 ± 0.17 mm
and 0.30 ± 0.15 mm for inter- and intra-examiner tests respectively. There was a
highly significant correlation between the repeated readings for intra- and interexaminer error test in all three dimensions, (table 10). A one sample t-test for
repeated readings showed no significant difference (p>0.05) in all dimensions (x,
y and z), (table 11).
2.2.2.3.2 Accuracy of DICOM slice landmarking compared to the gold standard
measurements
There was a significant correlation between the two measurements in both the y
and z dimensions (r = 0.999, p = 0.0001), (r = 0,998, p = 0.0001) respectively,
whereas in the x dimension there was no significant correlation (r = 0.000,
p=0.500), (table 12).
A one sample t-test showed that there was no statistically significant difference
for the y and z dimensions, (table 13). The x dimension, however, showed a
significant difference.
The mean difference between the two absolute measurements was 0.34 ± 0.20
mm, 0.22 ± 0.16 mm, 0.18 ± 0.13 mm in the y, z and x dimensions respectively.
Figures (27 and 28) show Bland-Altman plots for the sagittal and vertical data.
The results in table (14) show a high reproducibility of the measurements on
clinical cases. The same landmarks were digitised on the superimposed pre- and
postoperative DICOM images. Paired sample t-test showed a low significance
values (p = 0.3) in x, y and z dimensions.
Anas Almukhtar 2016
99
Chapter Two
Methodology
Figure 27: Bland Altman Plot for sagittal measurements. Note the uniform distribution
around the mean.
Figure 28: Bland Altman vertical measurements. Note the uniform distribution around the
mean.
Anas Almukhtar 2016
100
Chapter Two
Methodology
Table 9: The intra and inter examiner landmarking errors (Euclidian distances) between the
repeated readings at each landmark.
Inter-examiner
Mean
Landmark distance
(mm)
IF
0.22
Intra-examiner
SD
Landmark
Mean distance
(mm)
SD
0.15
IF
0.21
0.07
GPR
0.29
0.15
GPR
0.13
0.07
GPL
0.38
0.12
GPL
0.25
0.10
LL
0.36
0.19
LL
0.60
0.49
RL
0.39
0.22
RL
0.28
0.08
RM
0.37
0.13
RM
0.25
0.15
LM
0.38
0.30
LM
0.40
0.17
Ling
0.42
0.17
Ling
0.25
0.07
Table 10: Interclass correlation and the inter- and intra-examiner errors (three dimensional
distance) between the repeated readings.
Coordinates
Interexaminer
Intraexaminer
Mean
(mm)
SD
SE
Interclass
correlation
r-value
p-value
x
0.02
0.2
0.03
0.638
0.003
y
0.01
0.17
0.03
0.907
0.001
z
-0.09
0.7
0.12
0.991
0.001
x
-0.04
0.76
0.13
0.999
0.001
y
0.14
0.32
0.06
0.999
0.001
z
0.14
0.32
0.06
1.000
0.001
Anas Almukhtar 2016
101
Chapter Two
Methodology
Table 11: The inter and intra examiner errors in the three dimensions (one sample t-test)
Intra examiner error
Landmarks
Inter examiner error
Coordinate
Mean
SD
p-value
Mean
SD
p-value
x
y
z
x
y
z
x
y
z
x
y
z
x
y
z
x
y
z
x
y
z
x
y
z
0.08
0.53
-0.13
-0.08
0.62
-0.15
-0.15
0.44
-0.23
0.26
0.34
-0.21
0.10
-0.03
-0.11
0.00
-0.08
-0.05
0.72
-0.09
-0.15
-0.25
0.06
-0.14
1.02
0.99
0.18
1.06
1.05
0.55
1.17
1.01
0.4
0.44
0.87
0.25
0.48
0.59
0.17
0.33
0.73
0.42
0.86
0.48
0.46
0.55
0.92
0.31
0.88
0.37
0.23
0.89
0.32
0.61
0.82
0.45
0.33
0.32
0.49
0.19
0.70
0.91
0.28
1.00
0.85
0.82
0.19
0.72
0.55
0.43
0.91
0.42
0.12
0.13
0.16
0.01
-0.19
0.23
0.10
0.02
0.16
0.08
-0.11
0.24
0.06
-0.04
0.11
0.01
-0.08
-0.01
0.16
-0.11
0.17
-0.04
0.07
-0.09
0.22
0.17
0.22
0.36
0.22
0.16
0.31
0.23
0.25
0.19
0.17
0.32
0.28
0.3
0.18
0.17
0.34
0.38
0.32
0.28
0.28
0.32
0.39
0.21
0.34
0.23
0.24
0.95
0.18
0.06
0.57
0.86
0.28
0.48
0.27
0.23
0.68
0.79
0.32
0.91
0.66
0.96
0.39
0.48
0.31
0.83
0.76
0.44
IF
GPR
GPL
RL
LL
RM
LM
LING
Table 12: The differences between the two methods of measurements (Inter class
correlation).
Coordinate
Mean
(mm)
SD
SE
X
Y
Z
0.34
0.08
0.02
0.21
0.27
0.22
0.03
0.03
0.03
Anas Almukhtar 2016
Interclass correlation
r-value
p-value
0.000
0.998
0.999
0.500
0.001
0.000
102
Chapter Two
Methodology
Table 13: The differences between the two methods of measurements (one sample t-test)
Landmark
Coordinate
Mean (mm)
SD
p-value
x
y
z
x
y
z
x
y
z
x
y
z
x
y
z
x
y
z
x
y
z
x
y
z
0.35
-0.01
0.06
0.33
-0.15
-0.07
0.32
-0.15
0.05
0.43
-0.25
-0.09
0.29
-0.18
-0.18
0.32
0.04
-0.08
0.35
0.05
-0.07
0.36
0.11
-0.11
0.21
0.26
0.27
0.20
0.32
0.24
0.25
0.27
0.19
0.20
0.29
0.15
0.20
0.12
0.11
0.17
0.12
0.13
0.16
0.21
0.25
0.19
0.13
0.22
0.02
0.99
0.44
0.01
0.09
0.27
0.02
0.05
0.28
0.54
0.19
0.34
0.13
0.06
0.05
0.12
0.59
0.37
0.16
0.70
0.64
0.22
0.20
0.40
IF
GPR
GPL
RL
LL
RM
LM
LING
Table 14: Results of the pilot study of a repeated measurements on clinical cases (Paired
sample t-test).
Coordinate
Mean
(mm)
SD
SE
x
y
-0.27
-0.18
0.77
0.62
0.25
0.20
95% Confidence
Interval
Lower
Upper
(mm)
(mm)
-0.87
0.32
-0.66
0.29
z
0.31
0.91
0.30
-0.38
Anas Almukhtar 2016
1.01
p-value
0.31
0.39
0.33
103
Chapter Two
Methodology
2.2.2.4 Discussion
The primary objective of this project was to introduce and validate a new
method of radiographic measurements to assess the maxillo-mandibular changes
following orthognathic surgery. Each of the currently available methods has its
own deficiency which impacts negatively on the validity of the measurements.
The proposed method attempts to overcome the problems associated with the
current approaches of 3D surface model segmentation using DICOM image slices.
It promises a reliable and a reproducible 3D landmarking to measure surgical
changes.
Lack of anatomical correspondence is a major problem associated with the
colour coded error map method that is frequently used for radiographic
superimposition (243). However, the method is commonly used to evaluate facial
soft tissue changes following surgery. It overcomes the difficulty of the limited
number of landmarks available for soft tissue analysis. The anisotropic
deformation of the soft tissues of the face in response to the surgical movement
of the underlying jaw bones is difficult to measure. This is not the case when
measuring bony movements of the facial skeleton. As a result of surgery, the
maxilla moves as a unified unit, rigid body transformation. The addition of
rotational movements such as differential impaction or central midline can
produce variation in the amount of linear translation of various regions within
the structure. However, this affects the whole skeletal structure as a unit since
the geometric integrity is preserved.
Placement of landmarks on a CBCT scan slice is an advanced technique that can
provide accurate and detailed information about the internal skeletal structures.
The reliability and reproducibility of “slice” landmarking was recently validated
(96). However, the validity of this method to assess skeletal changes following
orthognathic surgery has not been tested yet.
Three points in 3D space are enough to create and orient a plane. The
coordinates of these three points will be changed as a result of the translation of
the plane in 3D space. Rotation of the plane around axes which passes through
Anas Almukhtar 2016
104
Chapter Two
Methodology
two of the three points would change the coordinates of the third point,
whereas the rotation around any point other than these three points would
change the coordinates of the three points. This explains how three points could
be used to monitor the position of its related anatomical structure in 3D space.
If this plane forms part of a larger solid 3D object then these three points could
be used to assess the exact translation and rotation movements of that solid
object in 3D space.
This concept was adopted in the proposed approach to assess the skeletal
displacement of the maxilla and the mandible as a result of orthognathic surgery
for correction of dento-facial deformities. Three anatomical landmarks are
therefore required to be marked on each hard tissue structure to accurately
measure its displacement in 3 planes of space. To overcome the problems
associated with surface remodelling, three landmarks within the maxilla and five
landmarks within the mandible were identified directly on the DICOM image
slices. Landmark positions on the maxilla and mandible were selected for their
favourable
geometric
position,
clear
anatomical
definitions
and
high
reproducibility.
Simulating the surgical movement on a plastic skull allowed the maxilla and
mandible to be separated and translated into different positions anteriorly and
vertically with minimal lateral or rotational movements. It was not possible to
measure rotational and lateral movements directly on the plastic skull using the
current experimental setup; therefore the movement in the x dimension was
considered to be close to zero throughout the calculations. However the small
inadvertent lateral or rotational changes in the x dimension led to the
expression of a significant p-value and low correlation when compared to the
physical movement in the x dimension. Differences in measurements in the y and
z dimensions were not statistically significant with a mean difference of 0.22 ±
0.16 mm and 0.18 ± 0.13 mm respectively. A one sample t-test was used to test
for difference from the reference measurement set, the 95% confidence interval
was narrow ranging from -0.03 to 0.15 mm for the y and z dimensions, with an
upper limit of 0.15 mm confirming the clinical insignificance. The high
Anas Almukhtar 2016
105
Chapter Two
Methodology
reproducibility of the method on clinical cases validates its applicability in
clinical research environment.
The proposed method proved reliable in measuring surgical changes of the jaw
bones which was demonstrated by the high correlation coefficient between the
physical and digital measurements. The method lends itself to uncomplicated
landmarking, the inter-examiner and intra-examiner variability were nonsignificant.
2.2.2.5 Conclusions
Internal landmarking of DICOM image slices is a reliable, reproducible and
informative method for assessment of the 3D skeletal changes following
orthognathic surgery.
Anas Almukhtar 2016
106
Chapter Two
Methodology
2.3 Section C: Validation of basic methods of soft tissue
Analysis
In this part, one experiment was carried out to validate the accuracy of Generic
mesh conformation which constitutes the bases of soft tissue analysis method.
Details about stereophotogrammetry system used and a brief details 3D model
conformation ( elastic deformation ) were also explained.
2.3.1 The use of a generic mesh to assess soft tissue changes
using stereophotogrammetry.
2.3.1.1 Introduction
The use of generic meshes for analysing biological geometry has previously been
reported (152,175).
The use of “correspondence analysis”, based on generic
meshes, has been suggested as a solution for the lack of accurate anatomical
correspondence between the pre- and post-operative images associated with the
current surface analysis methods (243).
The advantage of using a generic mesh is that there are a known number of
vertices each with known co-ordinates and the triangles formed by these
vertices are indexed or ordered. Most importantly following conformation this
index is maintained and preserved. Conformation (elastic deformation) is a
process to elastically deform the generic mesh to the pre- and post-operative
patient’s images. This process will then produce two meshes with the same
number of vertices and triangles in the same order in the file structure, where
each vertex represents a corresponding point on both pre- and post-operative
conformed meshes.
The accuracy of the conformation will determine the
accuracy of the final correspondences. Even though the conformation process is
semi-automated, it relies on an initial manual landmarking process used to
match certain corresponding anatomical features and constitutes the bases for
the automated process. In addition different conformation algorithms built into
various software packages may also affect the final correspondences.
Anas Almukhtar 2016
107
Chapter Two
Methodology
The in-house conformation software developed by the research group at the
University of Glasgow provides a wide range of 3D imaging tools including: 3D
image conformation (elastic deformation), inter-surface distance measurements,
3D landmarking and colour coded distance map generation.
For
each
subject
six
stereophotogrammetry.
facial
expressions
were
captured
using
3D
One facial expression image, rest position, was chosen
and used as the “generic mesh”. The rest position generic mesh was then
conformed to the remaining five images and this was repeated for each subject.
Following conformation of the rest position “generic mesh”, to each of the facial
expressions, the 19 landmarks pre-located on the generic mesh which were not
used during conformation should “slide” into their respective positions to match
each of the corresponding landmarks on each of the five facial expressions if the
algorithm was functioning correctly. The distance between the actual landmarks
on the non-conformed expression mesh and the landmarks on the conformed
generic mesh, for the same facial expression, gives an accuracy of the
conformation process. The closer to zero, the more valid and accurate is the
process of conformation.
2.3.1.2 Aim
The aim of this section is to determine the accuracy of the conformation process
based on 34 pre-marked facial landmarks, 15 of which are used during the
conformation process and the remainder as measures of conformation accuracy.
2.3.1.3 Materials and methods
2.3.1.3.1 Stereophotogrammetry capture protocol
All three-dimensional facial images were captured using the Di3D imaging system
(Dimensional Imaging, Hillington, Glasgow). The system was based on passive
stereophotogrammetry and produced fully textured 3D images from ear to ear.
The imaging system used two pairs of high-resolution digital cameras (Canon EOS
1000D-EOS Digital SLR, Canon, Japan) together with external flashes (Esprit
Anas Almukhtar 2016
108
Chapter Two
Methodology
digital 1000DX, Bowens, England, UK) to create a stereo 3D image based on the
principle of triangulation. Please refer to Section (1.1.1.5), (figure 29).
The Di3D system had separate image capturing software (Di3DCapture) and postprocessing viewing, manipulation and analysis software (Di3DView), both running
on a high performance PC (Dell OptiPlex 960) and Windows 7 (Microsoft). Di3D
Capture is the user interface control panel of the Di3D system which enabled the
user to capture, build and save the stereophotogrammetry images; in addition to
system calibration. The “live preview” function allowed the correct positioning
of the subject prior to capture, (figure 30).
Di3DView software allowed image manipulation including translation, rotation
and magnification. A variety of tools were built into the software, some of which
included
image
landmarking
and
3D
co-ordinate
extraction,
image
superimposition, Euclidian surface distance measurements and asymmetry score
calculations. In this study, the software was mainly used for image viewing,
manipulation and assessing the quality of each image prior to saving, (figure 31).
2.3.1.3.2 System Calibration
The Di3D system was calibrated prior to the image capture sessions using the
manufactures instructions.
This process was semi-automatic and involved
capturing several images of the calibration target in different orientations and
using the “calibration” function within Di3D Capture software to complete the
process, (figure 32).
This process provided the intrinsic parameters for the
camera configuration relying on the known spacing between the centres of the
circles on the calibration target. The calibration software extracted the coordinates of the circles on the image and from this information the software
could determine the relative positions of all four cameras without any further
operator intervention.
Anas Almukhtar 2016
109
Chapter Two
Methodology
Figure 29: Di3D Stereophotogrammetry system
Figure 30: Di3D Capture software main panel
Anas Almukhtar 2016
110
Chapter Two
Methodology
Figure 31: Di3D View software main panel
A
B
Figure 32: Di3D system calibration board: calibration board (A); Calibration board in
position during the calibration procedure
Anas Almukhtar 2016
111
Chapter Two
Methodology
2.3.1.3.3 Image capture protocol
For all captures, patients were seated on a chair directly in front of the Di3D
imaging system.
Using the “live preview” screen provided by Di3D-Capture
software, the patient was positioned correctly relative to all four cameras. To
standardise the images each subject was asked to:
Remove spectacles and jewellery,
Keep all hair completely off the face and neck using a head cap, (Nurses
cap Barrie, MOLNLYCKE, UK.)
Remain still during image capture,
Say “Mississippi”, then told to swallow once and say “N” (guidelines to
obtaining rest position natural facial expression as proposed by
Zachrisson, 1998).
Following capture each image was automatically built into a 3D model.
The images were then exported in Wavefront (.obj) format for future
analysis.
2.3.1.3.4 Sample
Ten individuals (6 male and 4 female) were chosen at random and consented to
take part in the study. Volunteers were healthy adults with no history of facial
deformity or previous surgery in the facial region. Males were clean-shaven to
avoid image distortion.
2.3.1.3.5 Subject imaging
Prior to 3D facial image capture thirty-four 2mm diameter self-adhesive black
non-reflective markers (Diamante, Apparel accessories Ltd, Guangdong, China)
were placed on each subjects face using an application tool (Pick-it-up vacuum
tool, Bead smith, China). The position of the markers were selected around the
eyes, nose, mouth and cheeks in addition to the peripheries of the face i.e. the
tragus, gonial angle and chin areas, (Figure 33) and (table 15).
Anas Almukhtar 2016
112
Chapter Two
Methodology
a) The participants were then asked to rehearse six facial expressions to
ensure that the markers were secure. The six expressions are shown in
figure (34). There were:
b) maximum smile,
c) cheek puff,
d) lip purse,
e) mandibular displacement to the right (simulating facial asymmetry),
f) forward mandibular displacement and
g) Repose or rest position.
The images were captured using the Di3D stereophotogrammetry system and
protocol previously described (Section 2.3.1.3.3).
In addition subjects were
asked to keep their lips closed through all captures. This would prevent
distortion of the image by exposure of the teeth. Two images were captured for
each facial expression and the best image was chosen. Images were individually
built and viewed using Di3D view software and saved as Wavefront (.obj) file for
further analysis.
Anas Almukhtar 2016
113
Chapter Two
Methodology
Table 15: Definitions of landmarks for validation of the accuracy of 3D image conformation.
Abbr.
Landmarks
1
EBR
Eyebrows-R
2
Gla
Glabella
3
EBL
Eyebrows-L
4
5
6
7
8
Exc-R
End-R
Na
Exc-L
End-L
Exocanthion-R
Endocanthion-R
Nasion
Exocanthion-L
Endocanthion-L
9
Sbtr-R
Subtragion-R
SbtrR1/3*
SbtrR2/3*
Ala-R
Ab-R
Prn
Ab-L
Ala-L
SbtrL1/3*
SbtrL2/3*
Sutragion-R
(1/3)
Subtragion-R
(2/3)
Alar curvature-R
Alar base-R
Pronasale
Alar base-L
Alar curvature-L
Subtragion-L
(1/3)
Subtragion-L
(2/3)
10
11
12
13
14
15
16
17
18
19 Sbtr-L
Subtragion-L
20 Go-R
Gonion-R
GoR1/3*
GoR22
2/3*
23 Ch-R
21
Definition
The point just above the eyebrows at a vertical line from
the pupil.
Most prominent midline point between eyebrows
The point just above the eyebrows at a vertical line from
the pupil.
Outer commissure of the eye fissure
Inner commissure of the eye fissure
Mid-point on the nasal bridge.
Outer commissure of the eye fissure
Inner commissure of the eye fissure
The most anterior inferior point of the anterior inferior
attachment of the ear helix, just above the ear lob
One third the distance from Sbtr-R to Ala-R
Two third the distance from Sbtr-R to Ala-R
Most lateral point on alar contour
the junction between the right nostril and upper lip
Most protruded point of the apex nasi (tip of the nose)
the junction between the right nostril and upper lip
Most lateral point on alar contour
One third the distance from Sbtr-L to Ala-L
One third the distance from Sbtr-L to Ala-L
The most anterior inferior point of the anterior inferior
attachment of the ear helix, just above the ear lob
The most lateral point of the cheeks close to mandibular
angle.
Gonion-R 1/3
One third the distance from Go-R to Ch-R
Gonion-R 2/3
The third the distance from Go-R to Ch-R
Cheilion-L
26 FL-L
Philtrum crest-L
27 Ch-L
Go28
L2/3*
Go29
L1/3*
Cheilion-L
Point located at lateral labial commissure
The tip of the right philtral ridge at the upper lip vermilion
border
Midpoint of the upper vermilion line
The tip of the right philtral ridge at the upper lip vermilion
border
Point located at lateral labial commissure
Gonion-L 1/3
One third the distance from Go-L to Ch-L
Gonion-L 2/3
The third the distance from Go-L to Ch-L
30 Go-L
Gonion-L
31 Li+3*
32 Li
33 Pog+3*
34 Pog
Labiali inferius
Labiali inferius
Pogonion+3
Pogonion
24 FL-R
Philtrum crest-R
25
Labiali superius
Ls
Anas Almukhtar 2016
The most lateral point of the cheeks close to mandibular
angle.
Mid-point on the lower vermilion line 3mm higher than Li
Mid-point of the lower vermilion line
Midline point 3mm higher than pogonion
Most prominent midline point of the chin
114
Chapter Two
Methodology
Figure 33: Full set of landmarks indicators placed on participant's face
Anas Almukhtar 2016
115
Chapter Two
Methodology
Lip purse
Figure 34: The six facial expressions. The relax expression (top left) was conformed, as a
generic mesh, to the other five facial expressions. The positional change of the markers on
the (relax expression) facial image following conformation were tracked and compared with
the positions of those on the target images.
Anas Almukhtar 2016
116
Chapter Two
Methodology
2.3.1.3.6 Pre-analysis image processing
Two steps were necessary to pre-process each image before it could be analysed
following capture, these were: image conversion and conformation. The first
step was accomplished using 3DSMAX software. The second step and final
analysis were carried out using the in-house developed conformation software.
2.3.1.3.6.1 Image Conversion
The conformation software required all the captured images to be converted
from Wavefront (.obj) to VRML (.wrl) files.
The texture information,
dimensional units and the orientation of the image were maintained during the
conversion process using 3DSMax software.
2.3.1.3.6.2 Image Conformation
The
procedure
has
been
previously
described
(Section
1.3.2.2.3.b).
In summary, the conformation software provides a dual display panel, the
generic mesh image (repose facial image) was imported and shown in one panel
whilst the captured face (in this case the smile facial expression) was imported
and shown in the second viewing panel, (figure 25).
The conformation process involved two steps that were executed in the
following sequence; initial conformation (semi-manual) and final conformation
(fully automated).
2.3.1.3.6.2.1 Initial conformation
The images were magnified to allow landmarks to be placed at the centres of
the markers. Fifteen landmarks were digitised on the generic image (landmark 4
to 9, 13, 14, 15, 19, 23, 25, 27, 31 and 33), please refer to table (15), and the
same corresponding landmarks were placed on the captured facial expression
image (smile). The sequence of landmark placement in both images was
identical for the process of conformation.
Anas Almukhtar 2016
117
Chapter Two
Methodology
Figure 35: The conformation software showing the main panel with two 3D images loaded;
the target image (left side) and the generic mesh (right side).
The initial conformation elastically deformed the 15 landmarks on the generic
mesh (relax image) to align on the corresponding 15 landmarks on the smile
expression image. The remainder of the repose mesh was automatically partially
deformed through minimizing the “bending energy” to produce an approximate
fit of the two mesh surfaces.
2.3.1.3.6.2.2 Final conformation
This process fully deformed the generic mesh not only in shape but also in
position. At this point the generic mesh resembled the facial expression mesh in
both shape and position in 3D space. The conformed image was exported as a
VRML (.wrl) file and saved for further analyses.
This process was used to
conform the (relax) facial mesh to each of the remaining 5 facial expressions
Anas Almukhtar 2016
118
Chapter Two
Methodology
meshes i.e. cheek puff, lip purse, smile, asymmetric mandibular movement and
mandibular protrusion. For each subject, the outcomes were five conformed
meshes in addition to the five corresponding original facial images; in total 50
conformations were produced.
2.3.1.3.7 Accuracy measurements
For each subject the conformed mesh was compared to original mesh for each
facial expression i.e. conformed maximum smile to original maximum smile and
so forth, using two methods of analyses.
Ideally when the two meshes are
loaded in viewing software both the meshes should be identical in shape and
position.
The first analysis method was based on the absolute Euclidian
distances between the points of the two meshes as well as visual distance colour
maps.
However these measurements rely on the distances between closest-
points rather than corresponding points. To overcome this, a second method of
analysis, based on corresponding landmarks was undertaken i.e. the adhesive
markers.
2.3.1.3.7.1 Surface Euclidian distances
For each subject the conformed mesh and the original mesh for each facial
expression in turn were imported into the conformation software. The mean
Euclidian distance between the two meshes was measured according to the
normal surface distance, (Section 1.3.2.2.3). The mean, standard deviation,
minimum, maximum and the percentage of the surface points and their
distribution around the mean distance were calculated. In addition, a distance
colour map was generated for visual illustration. The data from each
measurement was saved in an EXCEL file for further analysis.
To determine the accuracy of the conformation, the Euclidian distances between
all the points on the conformed mesh and the original facial expression mesh for
all facial expressions were calculated. The mean Euclidian distances was
calculated then grouped together with all conformations measurements to
calculate the overall mean Euclidian distances between the two meshes. This
was accomplished using Microsoft EXCEL.
Anas Almukhtar 2016
119
Chapter Two
Methodology
2.3.1.3.7.2 Corresponding landmarks
The second method of analysis was based on corresponding landmarks. In total
34 markers were placed on the face prior to image capture, 15 were used during
the conformation process whilst the remaining 19 were used for the analysis.
For each subject the conformed mesh and the original mesh for each facial
expression in turn were imported into the conformation software and
landmarked. The x, y and z coordinates of each of the 19 landmarks were
generated and exported as a plain text (.txt) file. For each expression, two text
files were created, one from the conformed mesh and the other from the
original mesh. The sequence of landmarking was the same for both, the original
and the conformed images, this allowed easy identification of landmarks and
their correspondence for analysis. The files were saved for future analysis.
The Euclidean distance between each corresponding landmark was calculated
using the 3D Pythagoras equation (2). Descriptive statistical analyses and paired
sample t-test were applied to analyse the data.
2.3.1.3.8 Landmarking error
Ten randomly selected images, one from each case, were landmarked twice with
two weeks interval by the same operator (AAM). The Euclidean distances
between the repeated digitisation of the same landmark were calculated.
2.3.1.4 Results
2.3.1.4.1 Landmarking error study
The mean Euclidean distance and standard deviation for landmarking errors for
each of the 34 landmarks is shown in table (16). The overall mean error for all
the landmarks was 0.25 mm ± 0.10 mm. The overall mean error for all the
landmarks was 0.23 mm ± 0.11 mm. Landmarks 6 and 8 had the lowest error
0.11 mm ± 0.06 mm and 0.11 mm ± 0.10 mm respectively whilst landmark 30 had
the largest error 0.57 mm ± 0.64 mm.
Anas Almukhtar 2016
120
Chapter Two
Methodology
2.3.1.4.2 Conformation accuracy
2.3.1.4.2.1 Mean absolute distance between meshes
Table (17) shows the absolute mean distances between the conformed mesh and
the original “generic” mesh. The largest distance was 0.06 mm which was
observed in subject 3 across all facial expressions.
2.3.1.4.2.2 Corresponding landmarks distances
The 15 landmarks which were used for the initial stage of image conformation
were excluded. The data from the remaining 19 landmarks were used to assess
the accuracy of the conformation process.
Table (18) shows the Euclidean distances between the 19 corresponding
landmarks on the conformed mesh and original mesh for all facial expressions.
The absolute mean distances between the corresponding landmarks ranged from
0.81mm to a maximum error of the conformation process of 1.85 mm for
landmarks 26 and 20 respectively. The overall mean Euclidean distance error of
the conformation process was 1.21 mm ± 0.28 mm.
The effect of each facial expression on the accuracy of the conformation was
also investigated. Table (19) shows the mean Euclidean distance between the
corresponding landmarks on the generic and conformed facial meshes. The five
facial expressions were ranked in an ascending order according to the accuracy
of the conformation process starting by the lateral mandible shift, lip purse,
forward mandible shift, cheek puff and smile.
The lowest mean Euclidean
distance between corresponding landmarks due to the potential errors of the
conformation process of the facial mesh was associated with the lateral
mandible shift expression which was 1.06 mm ± 0.33 mm. The largest mean
Euclidean distance between corresponding landmarks due to inaccuracy of the
conformation process of facial surface meshes was 1.46 mm ± 0.51 mm which
was associated with maximum smile.
Anas Almukhtar 2016
121
Chapter Two
Methodology
2.3.1.5 Discussion
Establishing an accurate conformation process (elastic deformation) of the
generic facial mesh to resemble the detailed anatomy of the face is essential for
a valid application of dense correspondence analysis to evaluate morphological
changes.
Previous studies on image conformation and elastic deformation have evaluated
the accuracy of the process using the mean surface distance measurement
(231,244). The deficiency in this approach lies in the fact that the distances
were between the closest points of the two surface meshes, generic mesh and
conformed mesh, not the actual correspondences. It is quite possible for the
meshes to slide over each other during the conformation process, measuring the
closest distance between two meshes would not detect this source of errors.
Assessment of the accuracy of the conformation process based on specific
landmarks also carries the risk of over estimating the accuracy of the
conformation process since the rest of the mesh vertices are not considered in
this type of analysis. Therefore, the measurements of the 90th percentile of the
vertices of the two meshes were considered in this study. This is essential for
the application of dense surface correspondence analysis to evaluate facial
changes due to surgery, pathology or growth.
Anas Almukhtar 2016
122
Chapter Two
Methodology
Table 16: Mean Euclidean distance and standard deviation for landmarking errors for each of the 34 landmarks.
Landmark
number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Case number
1
2
3
4
5
6
7
8
9
10
0.1
0.28
0.18
0.03
0.13
0.09
0.12
0.05
0.08
0.13
0.16
0.52
0.08
0.23
0.23
0.15
0.04
0.53
0.21
0.19
0.36
0.10
0.08
0.12
0.34
0.12
0.10
0.33
0.34
0.14
0.29
0.37
0.48
0.22
0.09
0.03
0.09
0.07
0.12
0.10
0.27
0.19
0.08
0.09
0.14
0.03
0.20
0.10
0.34
0.09
0.09
0.08
0.19
0.16
0.12
0.10
0.21
0.14
0.04
0.19
0.15
0.49
0.12
0.18
0.13
0.16
0.10
0.34
0.40
0.15
0.21
0.20
0.09
0.04
0.40
0.08
0.13
0.09
0.17
0.16
0.20
0.08
0.12
0.18
0.17
0.21
0.21
0.06
0.23
0.07
0.10
0.15
0.14
0.16
0.41
0.19
0.44
0.33
0.33
0.16
0.09
0.36
0.17
0.09
0.19
0.16
0.13
0.07
0.29
0.03
0.12
0.21
0.11
0.06
0.16
0.11
0.08
0.29
0.04
0.15
0.13
0.10
0.06
0.08
0.20
0.02
0.12
0.14
0.12
0.11
0.13
0.15
0.12
0.23
0.18
0.12
0.13
0.33
0.2
0.12
0.39
0.06
0.24
0.03
0.48
0.17
0.23
0.10
0.15
0.16
0.20
0.14
0.10
0.09
0.06
0.16
0.05
0.07
0.11
0.04
0.08
0.28
0.15
0.27
0.17
0.63
0.26
0.30
0.20
0.10
Anas Almukhtar 2016
Mean
(mm)
SD
(mm)
0.21
0.18
0.15
0.15
0.13
0.11
0.2
0.11
0.17
0.16
0.21
0.21
0.18
0.17
0.21
0.19
0.16
0.16
0.09
0.06
0.1
0.1
0.06
0.11
0.1
0.12
0.1
0.12
0.18
0.07
0.09
0.09
0.13
0.12
Continue
123
Chapter Two
Methodology
Landmark
number
Case number
Mean
(mm)
SD
(mm)
1
2
3
4
5
6
7
8
9
10
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
0.17
0.21
0.26
0.1
0.18
0.14
0.23
0.16
0.04
0.17
0.35
0.29
0.28
0.23
0.19
0.18
0.24
0.42
0.34
0.16
0.28
0.46
0.15
0.12
0.19
0.32
0.29
0.26
0.40
2.06
0.38
0.85
0.31
0.37
0.12
0.28
0.10
0.12
0.07
0.28
0.12
0.30
0.33
0.07
0.04
0.19
0.34
0.42
0.12
0.47
0.39
0.28
0.20
0.22
0.20
0.25
0.08
0.12
0.74
0.17
0.13
0.66
0.29
1.20
0.47
1.28
0.50
0.43
0.90
0.18
0.11
0.23
0.14
0.03
0.16
0.85
0.05
0.20
0.21
0.22
0.16
0.24
0.41
0.41
0.53
0.60
0.17
0.15
0.19
0.28
0.13
0.43
0.32
0.24
0.38
0.30
0.34
0.18
0.75
0.16
1.21
0.30
0.14
0.25
0.30
0.07
0.18
0.24
0.24
0.20
0.10
0.14
0.23
0.25
0.32
0.07
0.28
0.21
0.22
0.05
0.35
0.12
0.07
0.15
0.14
0.18
0.24
0.09
0.14
0.05
0.19
0.25
0.16
0.32
0.29
0.24
0.24
0.94
0.08
0.08
0.29
0.23
0.41
0.05
0.17
0.18
0.22
0.22
0.37
0.50
0.34
0.41
0.25
0.11
0.52
0.22
0.23
0.16
0.07
0.14
0.01
0.14
0.18
0.15
0.29
0.15
0.33
0.40
0.32
0.39
0.32
0.32
0.17
0.15
0.22
0.16
0.22
0.34
0.17
0.19
0.26
0.27
0.57
0.36
0.44
0.44
0.33
0.28
0.24
0.08
0.08
0.11
0.08
0.12
0.27
0.11
0.09
0.18
0.07
0.64
0.21
0.38
0.31
0.11
Overall
mean
0.19
0.37
0.19
0.34
0.24
0.32
0.17
0.16
0.25
0.23
0.23
0.11
Anas Almukhtar 2016
124
Chapter Two
Methodology
Table 17: Mean surface distance in millimetres
Lateral mandible
shift
Case
s
1
2
3
4
5
6
7
8
9
10
Abs
Mean
0.04
0.00
0.05
0.00
0.02
0.02
0.02
0.00
0.02
0.00
Mean
SD
0.01
0.00
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.84
0.03
0.10
0.03
0.03
0.02
0.04
0.01
0.17
0.02
Forward mandible
shift
Cheek puff
Abs
Mean
0.04
0.00
0.06
0.00
0.02
0.02
0.02
0.00
0.02
0.00
Mean
SD
0.01
0.00
0.01
0.00
0.00
0.00
0.00
0.00
0.01
0.01
0.07
0.07
0.10
0.01
0.01
0.01
0.03
0.02
0.12
0.01
Abs
Mean
0.04
0.00
0.06
0.00
0.02
0.02
0.02
0.00
0.00
0.00
Mean
SD
0.01
0.00
0.01
0.01
0.03
0.00
0.04
0.01
0.02
0.00
0.06
0.02
0.10
0.01
0.03
0.01
0.03
0.01
0.05
0.01
Anas Almukhtar 2016
Smile
Abs
Mean
0.04
0.00
0.06
0.00
0.02
0.02
0.02
0.00
0.00
0.00
Lip purse
Mean
SD
0.01
0.00
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.07
0.02
0.10
0.01
0.06
0.01
0.03
0.01
0.02
0.01
Abs
Mean
0.04
0.00
0.06
0.00
0.02
0.02
0.02
0.00
0.00
0.00
Mean
SD
0.01
0.00
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.09
0.03
0.11
0.02
0.04
0.01
0.03
0.03
0.05
0.01
125
Chapter Two
Methodology
Table 18: The mean Euclidean distances (mm) of the 19 corresponding landmarks between the conformed and original mesh for all facial expressions.
Landmark
Number
1
2
3
10
11
12
16
17
18
20
21
22
24
26
28
29
30
32
34
Overall
error
Mean
(mm)
SD
(mm)
1.58
1.05
1.17
0.52
0.56
0.59
1.18
1.16
1.01
0.93
0.98
0.52
0.41
0.56
1.32
1.38
1.15
0.86
0.36
1.38
0.83
1.29
1.30
1.31
1.21
1.14
1.28
1.24
1.85
1.49
0.83
0.82
0.81
1.10
1.53
1.53
1.01
1.04
0.68
0.65
0.74
0.98
0.83
0.88
0.75
0.86
0.74
1.36
0.86
0.72
0.58
0.49
1.10
1.02
0.76
0.7
1.75
0.91
1.21
0.28
CASE 1
CASE 2
CASE 3
CASE 4
CASE 5
CASE 6
CASE 7
CASE 8
CASE 9
CASE 10
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Mean
1.48
0.60
0.94
1.03
0.89
1.96
1.39
1.15
1.12
1.17
1.35
1.00
1.11
0.94
0.71
1.43
1.34
0.56
0.93
1.81
1.16
1.80
1.40
1.71
1.75
1.33
1.74
1.43
1.72
1.37
1.48
1.08
1.21
1.06
1.34
1.43
0.80
1.32
0.89
0.41
0.86
1.32
1.35
1.04
1.16
0.90
1.04
1.53
1.33
1.33
0.86
0.75
0.12
0.78
1.13
0.92
0.41
1.33
1.07
1.08
1.46
1.56
1.08
1.09
1.43
1.01
1.41
1.44
0.91
0.85
0.62
0.91
1.31
1.68
0.92
0.35
1.47
0.80
1.64
0.69
1.22
1.32
1.28
1.66
1.48
1.27
0.52
0.91
0.90
0.92
1.02
1.52
1.60
0.69
0.40
0.88
0.31
1.22
1.86
1.52
0.49
1.08
1.55
1.09
2.06
1.92
0.49
0.88
0.62
1.35
1.17
1.00
0.89
0.30
0.74
0.44
0.84
1.82
1.54
1.41
0.28
0.89
1.79
3.12
2.48
0.37
0.77
0.73
2.36
2.43
2.18
1.33
2.10
1.20
0.53
1.13
0.88
0.73
1.12
1.09
0.65
0.60
2.26
0.83
0.31
0.73
0.65
1.44
1.65
1.93
1.74
0.89
1.29
1.31
1.26
1.03
1.03
0.91
0.81
0.69
0.85
1.88
1.47
0.27
0.54
0.36
0.21
1.14
0.98
0.92
2.68
1.12
1.42
0.96
1.14
1.13
1.09
1.46
1.08
1.04
Anas Almukhtar 2016
126
Chapter Two
Methodology
Table 19: Mean Euclidean distance between the corresponding landmarks for each facial expression (mm).
Landmark
number
Landmarks
Abbr.
1
EBR
2
Gla
3
EBL
10
SbtrR 1/3
11
Sbtr R2/3
12
Ala-R
16
Ala-L
17
Sbtr L1/3
18
Sbtr L2/3
20
GoR
21
GoR 1/3
22
GoR 2/3
24
PhLR
26
PhLL
28
GoL 2/3
29
Go L1/3
30
GoL
32
Li
34
Pog
Overall mean
SD
Name
Eyebrows-R
Glabella
Eyebrows-L
Subtragion-R (1/3)
Subtragion-R (2/3)
Alar curvature-R
Alar curvature-L
Subtragion-L (1/3)
Subtragion-L (2/3)
Gonion-R
Gonion-R 1/3
Gonion-R 2/3
Philtrum crest-R
Philtrum crest-L
Gonion-L 1/3
Gonion-L 2/3
Gonion-L
Labial inferius
Pogonion
Lateral
mandible
shift
Mean
1.87
1.08
1.23
1.09
1.08
0.96
0.99
0.83
0.94
1.40
1.31
0.70
0.69
0.53
0.78
0.99
1.51
1.39
0.74
1.06
0.33
Cheek
puff
Mean
1.10
0.58
1.23
1.65
1.95
1.76
1.67
2.03
1.62
1.58
1.76
0.84
1.03
0.90
1.25
1.65
1.41
0.68
0.76
1.34
0.45
Forward
mandible
shift
Mean
1.25
1.01
1.43
0.82
0.78
0.98
0.66
0.88
0.80
2.49
0.97
0.77
0.72
0.66
1.07
1.38
1.35
1.16
1.89
1.11
0.46
Anas Almukhtar 2016
Smile
Mean
1.62
0.97
1.5
1.85
1.63
1.02
0.93
1.60
1.81
2.51
1.80
0.97
1.16
1.11
1.30
2.11
2.15
0.61
0.99
1.46
0.51
Lip
purse
Mean
1.07
0.50
1.08
1.11
1.10
1.35
1.44
1.04
1.01
1.27
1.62
0.87
0.87
0.83
1.08
1.53
1.23
1.19
0.82
1.11
0.27
Mean
(mm)
1.38
0.83
1.29
1.30
1.31
1.21
1.14
1.28
1.24
1.85
1.49
0.83
0.89
0.81
1.10
1.53
1.53
1.01
1.04
1.21
0.28
SD
(mm)
0.35
0.27
0.17
0.43
0.47
0.35
0.41
0.52
0.45
0.60
0.35
0.10
0.20
0.23
0.20
0.41
0.36
0.34
0.48
127
Chapter Two
Methodology
It would have been ideal to track every vertex on the conformed mesh to
evaluate the accuracy of the conformation process. However, this was not
possible to achieve. Instead, the positional changes of a set of pre-defined
landmarks that were digitized on the pre-conformed surface mesh were tracked
to detect any shift from their original anatomical position as a result of the
conformation process. Although this was not a comprehensive surface analysis,
its robustness was maximised by carefully selecting the set of landmarks to
represent various anatomical regions of the face. We have also considered two
methods of measuring the disparities between the two surface meshes, the
absolute distances and the Euclidean ones. The first method takes in
consideration the discrepancies between the two meshes in both directions and
produces
positive
and
negative
values.
Despite
the
fact
that
these
measurements are descriptive to the magnitude and the direction of the
conformation errors, the mean value of these measurements would be small as
the positive and the negative measurements would cancel each other. On the
other hand the Euclidean distances which measure the shortest distances
between corresponding points of the two surface meshes, do not consider the
directionality of the mismatch between the two surface meshes, therefore, the
average arithmetic value of this distances is more meaningful. The combination
of the two methods of measurements considered in this study provides the most
comprehensive analysis for the assessment of the accuracy of the conformation
process which is novel.
Stereophotogrammetry was the chosen facial capture method, in this study, to
avoid exposing the volunteers to harmful radiation. At the same time, the
method produced a three dimensional surface image of the face that contained
the texture information which facilitated landmark identification.
In order to simulate the variation associated with facial expression, five facial
expressions were captured for each participant in addition to the baseline
relaxed posture. The participants were asked to slide the bottom jaw forward to
resemble prognathic mandible, slide the mandible to the left to resemble
mandibular asymmetry, cheek puff, lip purse, and smile postures to test the
Anas Almukhtar 2016
128
Chapter Two
Methodology
accuracy of the conformation on variable facial features and with various
expressions.
This approach also allowed a comprehensive analysis of landmarking error which
impacts on the reliability of the conformation process. To reduce the effect of
landmarking errors rounded 2mm landmarks markers were placed on the
positions of the 34 anatomical landmarks on each participant face. The presence
of these markers significantly reduced the landmarking error and allowed the
conformation process to be analysed comprehensively by adding a reliable
landmarks on the peripheries of the face (245). The rounded shape of the
landmark
indicators
facilitated
the
accurate
in
landmarks
digitization,
(0.23mm±0.11mm).
Image registration was not necessary in this project. The initial step of the
conformation process involved the translation of the corresponding landmarks to
match their positions of the target image followed by the elastic deformation to
minimize bending energy (thin plate spline). This process included both shape
and positional change. The six facial postures were captured at the same
session. This provided a relatively close starting point for the conformation
process.
Despite the fact that only 10 volunteers participated in this study, each of the
facial postures was considered an individual case, therefore; the total number of
the images involved in the study was 50. A total of 15 landmarks were used to
execute the conformation procedure. To eliminate bias, these landmarks were
excluded from the analysis of the accuracy of the conformation procedure.
The results of this study confirmed that landmarks around the lips and nose were
associated with lower level of conformation error of facial surface meshes
compared to those around the borders of the image such as cheeks, gonial angle
regions. This might be due to the lack of distinguished surface topography upon
which the elastic deformation relied. However, these areas are of our least
concern as the regions around the lips, nose, and chin are the focus of facial
Anas Almukhtar 2016
129
Chapter Two
Methodology
analysis following orthognathic surgery and they showed a higher level of
accuracy.
Despite the minimal conformation error detected in both methods of analysis,
the results confirmed our assumption that there is a higher level of error in mesh
conformation that could not be detected using traditional closest surface
distance measurement which produced elusive high conformation accuracy. Two
main factors might attribute in the errors in the conformation process; first and
most important is the accuracy and reproducibility of the operator’s landmarks
digitization used in the initial conformation stage, and second is the deficiency
of the associated algorithm of conformation which can compensate for these
errors especially in the peripheral areas of the facial mesh (244).
2.3.1.6 Conclusions
The conformation procedure has an acceptable level of accuracy and could be
applied for the 3D dense correspondence analysis of facial morphology. This has
broad clinical applications including facial analysis, evaluation of the impact of
orthognathic surgery in changing facial morphology, and monitoring of facial
growth. The conformation accuracy is higher toward the centre of the face than
the peripheral regions.
Anas Almukhtar 2016
130
Chapter Two
Methodology
2.4 Section D: Analysis of skeletal and soft tissue
changes following orthognathic surgery.
This final section is aimed at relating the hard tissue changes of the maxilla and
mandible with the overlying soft tissue change in three-dimensions.
The
previous described methods and validations have been used to determine the
skeletal change and soft tissue changes.
To perform the analysis, three main steps were followed; these include; preanalysis 3D image preparation, Analysis of skeletal surgical displacement,
Analysis of soft tissue changes in response to surgery.
The
pre-analysis
3D
image
preparation
includes;
DICOM
image
superimposition; soft and hard tissue models segmentation and Soft
tissue image processing.
Analysis of skeletal
surgical
displacement
include;
DICOM
image
orientation and creation of reference planes; DICOM slice landmarking
and data analysis.
Analysis of soft tissue changes following surgery includes; construction of
generic mesh; conformation of the generic mesh to the full set of the
pre- and post- surgical soft tissue models; soft tissue model averaging
and generation of the dense correspondence analysis.
2.4.1 Pre- analysis 3D image preparation
2.4.1.1 Volumetric image superimposition using voxel based registration
Previously the add-on module for Maxilim was used but an alternative software
package, OnDemand3D (Cybermed, Seoul, South Korea) was used for this
purpose. The accuracy of the voxel based registration of this software has been
previously validated by Lee et al., 2012 (67). The process was however the
Anas Almukhtar 2016
131
Chapter Two
Methodology
same. For each patient both the pre- and post-operative images were loaded
into the software using the “fusion” module. The user interface of this module
showed three orthogonal slices of each image; the primary image “preoperative”, the secondary image “post-operative” in addition to a combined
view, (figure 36). The secondary image was the movable one or source whereas
the primary image was the target.
Four steps were involved in the registration procedure. Firstly using manual
registration, the post-operative image was moved manually in three dimensions
by translating and rotating the images to align it as close as possible to the
target pre-operative image.
Secondly the region of interest on which the
automated voxel based registration was to be performed was selected. The
region of interest was selected as a 3D box occupying the anterior cranial base
and extended to involve the forehead. Thirdly the automated registration was
completed. The combined view orthogonal slices window allowed visually
checking of the alignment between the two images. Fourthly the software
provided a re-slicing function by which the registered DICOM image could be
saved in the new position for further analysis. This procedure was applied to the
100 patient’s data sets.
2.4.1.2 Hard and soft tissue segmentation
Following voxel based registration and in order to perform the soft tissue
analysis, the 3D soft and hard tissue models of the superimposed pre- and postoperative images were segmented or extracted from the DICOM images. This was
achieved using Maxilim software as previously described and involved
segmenting the images at HU=276 and -967 for the hard and soft tissue models
respectively, (figure 37).
Anas Almukhtar 2016
132
Figure 36: Voxel based registration: showing the pre-operative( top row), post-operative (middle row) and combined
view (bottom row). The volume of interest was highlighted with a blue dotted box.
Chapter Two
Anas Almukhtar 2016
Methodology
133
Chapter Two
Methodology
2.4.1.3 Soft tissue model processing
Since the soft tissue models were segmented from CBCT scans, all the connected
tissue surface that shared the same HU value were segmented as one model. For
the soft tissue model, the facial surface was successfully segmented. However
the oro-nasal and pharyngeal soft tissue surfaces were also segmented as one
continuous model attached to the facial soft tissue. The presence of the internal
nasal soft tissue affected the accuracy of the generic mesh conformation and
had to be removed and holes in the facial mesh were successfully filled using
VRMesh. The process was accomplished by isolation of the undesirable tissue
surface mesh from the main facial surface mesh and deleting the unwanted part
using the function “delete floating models”. The holes in the mesh such as the
mouth opining and nostrils were mended using the function “Mend gaps” in
VRMesh software.
For future soft tissue analysis in the in-house developed software, the STL
models were converted to Virtual Reality Modelling Language (VRML) (.WRL) file
format using VRMesh software. The image coordinates and units were preserved
during the conversion process.
2.4.2 Measurement of hard tissue displacement following surgery
2.4.2.1.1 Image orientation and creation of reference planes
Maxilim software was used to segment the pre-operative hard tissue and a
horizontal (axial) reference plane was created by identifying the left and right
orbitale and left porion landmarks.
To establish the median (sagittal) reference planes, hard tissue nasion and sella
were identified. The median plane was oriented perpendicular to the horizontal
plane and passed through these nasion and sella. The vertical (coronal) was
automatically generated perpendicular to both planes passing through the sella
point, (figure 38).
Anas Almukhtar 2016
134
Chapter Two
Methodology
Table 20: Landmarks definitions used for the measurements of skeletal displacement
Landmark’s
name
Code Definition
The most posterior point of the incisive canal
Incisive
opining on the first axial slice showing the full
1
IF
foramen
form of the canal when scrolling coudo-cranially
through DICOM slices.
The most posterior point of the right greater
Greater
palatine foramen on the first axial slice showing
2
palatine
GPR
the full form of the foramen when scrolling
foramen/Right
coudo-cranially through DICOM slices.
The most posterior point of the right greater
Greater
palatine foramen on the first axial slice showing
3
palatine
GPL
the full form of the foramen when scrolling
foramen/Left
coudo-cranially through DICOM slices.
The lowest point of the right lingual foramen on
Lingual
the first coronal slice that shows the full form of
4
LR
foramen/Right
the foramen when scrolling antero-posteriorly
through DICOM slices.
The lowest point of the left lingual foramen on
Lingual
the first coronal slice that shows the full form of
5
LL
foramen/Left
the foramen when scrolling antero-posteriorly
through DICOM slices.
The outermost point of the right mental foramen
Mental
on the first axial slice that shows the full form of
6
MR
foramen/Right
the foramen when scrolling cranio-coudaly
through DICOM slices.
The outermost point of the left mental foramen
Mental
on the first axial slice that shows the full form of
7
ML
foramen/Left
the foramen when scrolling cranio-coudaly
through DICOM slices.
The tip of the lingual tubercle on the first
Lingual
coronal slice that showed the most posterior
8
Ling
tubercle
point on the lingual tubercle when scrolling
postero-anteriorly.
The most anterior point of the chin on the first
9
Pogonion*
Pog
slice that showed the first bony chin when
scrolling antero-posteriorly.
10 Menton*
Me
The lowest point of the chin on any coronal slice
The most anterior point of the Genioplasty plate
Genioplasty
11
Pt
on the coronal slice when scrolling anteroPlate*
posteriorly.
* Landmarks used with genioplasty cases only
Anas Almukhtar 2016
135
Chapter Two
Methodology
A
B
Figure 37: Segmented 3D models from the DICOM image using Maxilim software skeletal
(right) and soft tissue (left).
Figure 38: Reference planes: Showing the superimposed post-operative and the preoperative models. The reference planes were created and one couple of corresponding
landmarks were placed on the left mental foramen and were highlighted to show their
orthogonal distances measurements to the reference planes. the full set of measurements
was displayed on the left hand side of the screen
Anas Almukhtar 2016
136
Chapter Two
Methodology
2.4.2.2 Landmarking pre- and post-operative images
The skeletal changes achieved by orthognathic surgery were measured using the
newly developed method of directly landmarking the DICOM image slices.
Detailed descriptions of the measuring technique and validation process were
provided in section (2.2.2) the results of the validation study has been published
in peer reviewed scientific journal (246). For the current study, the same
technique was used. In summary, the technique measured the orthogonal
distance from selected anatomical landmarks placed on the DICOM slices,
following voxel based registration, of both the pre- and post-operative images to
common reference planes. The definition of each anatomical landmark was
given in table (20). The difference between the two measurements was a
measure of the surgical change of the maxilla or mandible.
Using a custom cephalometric analysis in Maxilim software, eleven landmarks
were identified on each of the pre- and post-operative CBCT images slice, (table
(20). The custom cephalometric analysis was based on the orthogonal distances
of 11 landmarks from each of the three reference planes. The measurements
were then exported into a cephalometric report and saved as an EXCEL file for
further analysis, (figure 38).
2.4.2.3 Data processing and analysis
Data analysis was accomplished using EXCEL software. Two sets of orthogonal
landmark distances relative to the reference planes were listed for each case
(one pre-operative and one post-operative). In the A-P dimension, since all the
landmarks were located anterior to the coronal plane, a positive measurement
meant anterior displacement of the landmark (away from the plane) while
negative means a posterior displacement. In the vertical dimension, since all the
landmarks are below the horizontal plane, negative sign measurement indicated
a downward displacement of the landmark (away from the plane) and positive
measurement meant upward displacement. In the x (medio-lateral) dimension,
since the landmark may be located on either side of the plane, calculations are
different. When both the pre- and post-operative landmarks had the same sign,
Anas Almukhtar 2016
137
Chapter Two
Methodology
the calculations follow the same patterns of the above calculations with the
same sign. In a few cases the pre- and post-operative landmarks carried
different signs. In this case, the displacement equals the sum of the two
measurements and in the direction of the post-operative measurement.
In this study, the main aim was to investigate the effect of skeletal movement
on the soft tissue; therefore the movement of the maxilla and mandible were
calculated considering the landmarks which were close to the soft tissue in the
direction of movements. For maxilla, the incisive foramen (IF) landmark was
used for A-P and lateral movements while in the vertical movement, the average
of the right and left greater palatine foramen (GPR and GPL) were considered in
addition to the measurements at the IF landmark. To assess mandibular
movement, the mean displacement of the left and right mental foramen (LM and
RM) and genial tubercle (Ling) were used. For cases with genioplasty, additional
measurements were made in the region of the chin. Menton (Me), Pogonion (Pog)
and Plate (Pt) were considered for the vertical and AP displacement, please
refer to table(20) for landmark definitions. Unfortunately, no reliable landmark
could be found to assess the lateral chin displacement.
2.4.2.4 Hard tissue landmarking error study
A landmarking error study was carried out on 30% of the total study sample. The
error study sample was selected using a systematic sampling technique where
the first one third of each group (Le Fort I, bimaxillary and BSSO) was selected.
A total of 30 cases were included in the study.
The full set of landmarks used in the main study was re-digitized by the same
researcher following a two week interval and the x, y and z coordinates were
recorded. The orthogonal distance in each dimension (the distance from each
landmark perpendicular to each of the three dimensional planes) was calculated
and a mean distance at each dimension was computed using EXCEL software.
The significance of the positional differences of the repeated landmark
digitization were analysed in the three dimensions (x, y and z) using a paired
sample t-test.
Euclidian distances of the repeated measurements were
Anas Almukhtar 2016
138
Chapter Two
Methodology
calculated in Excel using the 3D Pythagoras equation (Equation 2) and the mean
error was calculated.
Table 21: Landmarks used for generic mesh conformation.
1
2
3
4
5
6
Abbr.
Exc-R
End-R
Na
Exc-L
End-L
Ab-R
Landmarks
Exocanthion-R
Endocanthion-R
Nasion
Exocanthion-L
Endocanthion-L
Alar base-R
7
Prn
Pronasale
8
Ab-L
Alar base-L
9 Ch-R
10 FL-R
Cheilion-L
Philtrum crest-R
11 Ls
12 FL-L
Labial superius
philtrum crest-L
13 Ch-L
14 Li
15 UM-R
19 Lm
Cheilion-L
Labial inferius
Upper
lip
distance-R
Upper
lip
distance-L
Upper
lip
distance-R
Upper
lip
distance-L
Labio-Mental fold
20 Pog
Pogonion
16 UM-L
17 LM-R
18 LM-L
Anas Almukhtar 2016
Mid
Mid
Mid
Mid
Definition
Outer commissure of the eye fissure
Inner commissure of the eye fissure
Mid-point on the nasal bridge.
Outer commissure of the eye fissure
Inner commissure of the eye fissure
Junction between the right nostril and
upper lip
Most protruded point of the apex nasi (tip of
the nose)
Junction between the right nostril and
upper lip
Point located at lateral labial commissure
Tip of the right philtral ridge at the upper
lip vermilion border
Midpoint of the upper vermilion line
Tip of the right philtral ridge at the upper
lip vermilion border
Point located at lateral labial commissure
Mid-point of the lower vermilion line
Mid-point on the upper lip vermilion border
between Ch-R and FL-R
Mid-point on the upper lip vermilion border
between Ch-L and FL-L
Mid-point on the lower lip vermilion border
between Ch-R and Li
Mid-point on the lower lip vermilion border
between Ch-L and Li
Deepest median point on the curve between the
lower lip and chin
Most prominent Median point of the chin
139
Chapter Two
2.4.3 Analysis of soft tissue changes following surgery
Methodology
A dense correspondence analysis was used for the soft tissue analysis. In addition
to the Euclidean distance analysis, the positional changes of facial mesh nodes
were analysed in x, y and z dimensions individually. The analysis relied on the
pre-and post-operative conformed generic meshes to create the dense
correspondence. These meshes contain the same number of vertices and
triangles which helped in identifying the corresponding vertices of the compared
meshes. Based on this correlation, the translation of each vertex from the preoperative to the post-operative mesh was measured both as a Euclidian distance
and orthogonal (x, y and z) dimensions separately.
The full data sample of 100 patients was grouped according to surgical
procedure into six groups. These were; Le Fort I maxillary advancement; Le Fort
I maxillary advancement with genioplasty; BSSO mandibular advancement; BSSO
mandibular
advancement
with
genioplasty;
bimaxillary
advancement;
bimaxillary advancement with genioplasty. Only three groups were successful in
producing a satisfactory sample size (more than 10 samples). These were Le Fort
I maxillary advancement (33 samples); BSSO mandibular advancement (12
samples); and bimaxillary advancement groups (12 samples). The soft tissue
analysis was limited to these three groups.
For each of the selected groups, soft tissue model averaging was carried out.
This procedure produced pre-operative and post-operative average soft tissue
models which were used for analysing soft tissue changes following orthognathic
surgery within the group.
2.4.3.1 Construction of the generic mesh
The generic mesh used in this project was created by our research group in the
University of Glasgow. The facial polygonal mesh composed of approximately
1000 vertices distributed in almost equal distances from each other. Using
VRMesh further modifications of the original mesh were carried out; any holes in
Anas Almukhtar 2016
140
Chapter Two
Methodology
the mesh were filled i.e. nostrils, eyes and mouth and removal of any unwanted
areas of the mesh i.e. back of the head.
This created a generic mesh with uniform 2 - 3 mm sized triangles free from
defects extending from the upper hairline to the submental area vertically and
to the tragus of the ears bilaterally. The generic mesh carried no texture
information and was saved as a VRML (.wrl) file to be used later for the
conformation procedure.
Using VRMesh software the mesh was duplicated and segmented into 9
anatomical regions; forehead, nose, left and right paranasal areas, left and right
cheek areas, upper lip, lower lip and chin. The 9 regions were segmented but
remained at the same coordinates to preserve the generic mesh vertices index,
(figure 39).
2.4.3.2 Generic mesh conformation
The in-house developed software was used to perform the conformation (elastic
deformation) as previously described (Section 2.3.1.3.6.2). The same generic
mesh was conformed to the pre- and post-operative soft tissue models of the 100
patients included in the sample.
Either the pre- or the post-operative 3D models (one at a time), was loaded into
the software in the “scan” panel, (figure 40). The generic mesh was loaded to
the software in the “generic mesh” panel. A set of landmarks, (table 21) were
identified on both models. The sequence of landmarking was crucial for a
successful conformation since it represents the corresponding points “anchor
points” between the generic mesh and the facial 3D model mesh.
No
superimposition
(rigid
registration)
was
required
since
the
elastic
deformation process moved the generic mesh into the position of the facial
model mesh. The initial step was semi-automated and was guided by the
corresponding landmarks on both images combined with elastic deformation of
Anas Almukhtar 2016
141
Chapter Two
Methodology
the entire mesh based on bending energy reduction (thin plate spline). The next
step was fully automated performing refinement of the generic mesh surface
with elastic deformation based on localised geometrical feature matching.
The conformed mesh was then exported as a (.wrl) file. The x, y and z
coordinate of the conformed mesh were also exported. At the end of this
process, 200 conformed meshes (pre and post-operative meshes for each of the
100 patient) were saved with their respective coordinate data for further
analysis.
The complete generic mesh was used for the purpose of conformation (elastic
registration) of the per- and post-operative models and soft tissue analysis.
While the segmented mesh was used for the analysis of the final prediction
accuracy.
2.4.3.3 Facial Soft tissue model averaging
Following the conformation of the generic mesh onto the soft tissue models of
both the pre-operative and the post-operative scans. An average face for each
group, pre-operative and post-operative, was created. The face averaging
procedure
was
based
on
the
conformed
generic
meshes.
Since
the
correspondence was already established using the generic mesh index, partial
Procrustes Analysis (PPA) was performed on each of the groups where translation
and rotation was performed and the average position for each vertex was
calculated based on the generic mesh correspondence. The result of this
procedure was an average pre-operative and post-operative facial model. The
average face carried the same number and index of vertices as the original
generic mesh. The procedure of facial averaging was performed for the analysis
of the facial changes following Le Fort I advancement, BSSO advancement and
bimaxillary advancement surgeries.
Anas Almukhtar 2016
142
Chapter Two
Methodology
I
A
C
B
D
H
E
G
F
Figure 39: The generic mesh. Nine segments were detached (preserve the same 3D
coordinate). The nose (B), the right and left Paranasal areas (B,C), the upper lip (D), the
lower lip €, the chin (F), the right and left cheeks (G,H), and the forehead (A), the cheeks
and forehead areas were not involve in the analysis.
Anas Almukhtar 2016
143
Chapter Two
Methodology
A
B
C
Figure 40: Steps of generic mesh conformation: Landmarks placement
(A), initial elastic deformation (B), final elastic deformation(c)
Anas Almukhtar 2016
144
Chapter Two
Methodology
2.4.3.4 Anatomical dense correspondence analysis
Following facial averaging, the average pre-operative and post-operative faces
were superimposed on the eyes-nasal bridge region using surface rigid
registration. The process was accomplished using VRMesh software. An interim
template for registration was created as a copy of the preoperative model
cropped to the region of the eyes and nasal bridge regions only without changing
the 3D original position. The post-operative model was then registered on the
template using PPA followed by ICP surface rigid registration. The template was
then discarded. Since the template and the preoperative model have the same
3D position, the post-operative model is now registered to the pre-operative
model. The generic mesh index was applied to create the correspondence
between the superimposed average faces and the changes at the surgical sites
were
analyzed
using
in-house
software.
The
distance
between
the
corresponding vertices was displayed as a colour scale in millimeters ranging
from red colour (outward located vertices) to blue (inward located vertices)
whereas green colour represented no change. The intensity of the colour, on
both sides of the scale, represents the distance from zero location.
The differences in the corresponding vertices position were analyzed in the
individual dimension of space (x, y and z) using in-house software. The
difference in each dimension was displayed as a colour coded distance map,
(figure 41). Since a directional element was involved. A right hand coordinate
system was constantly adopted in the analysis. The colour map was specific for
each directional change. For the horizontal (medio-lateral) changes the points
which were displaced to the left of the face were shown in red whereas points
displaced to the right of the face were shown in blue. Non displaced points were
shown in green colour. The intensity of the colour represented the amount of
the displacement. The upper and lower limits of the colour scale were set at the
mean skeletal displacement in each of the analyzed surgical groups. The same
was applied to the rest of the directions with specific directional indicators, i.e.
in the y-dimension points which moved up were shown in red while those which
moved down in blue.
In the z-dimension points which moved towards the
observer were shown in red while those away from the observer in blue.
Anas Almukhtar 2016
145
Chapter Two
Methodology
2.4.4 Simulation of soft tissue following orthognathic surgery
2.4.4.1 Prediction of soft tissue changes following orthognathic surgery
The correlation between the soft and hard tissue displacements for Le Fort I
advancement surgery were analyzed. Only Le Fort I osteotomy cases (with no
genioplasty) were considered in this part of the study, with 30 pairs of pre-and
post-operative conformed generic meshes. The facial shapes were divided into
the regions of forehead, nose, left and right paranasal, upper lip, low lip and
chin regions. The displacements of soft tissues at these regions were calculated
from the post and pre-operative conformed generic meshes. Principal
component analysis (PCA) was used to analyze these displacements. The first 22
principal components (which represent 95% of variations of the displacements)
were used to reconstruct the displacements of an individual case. With pairs
data – loadings of 22 principal components versus the displacements of hard
tissues, the relationship between the loadings of the first 22 principal
components of the displacements of soft tissues and the displacements of hard
tissues were calculated. With the relationship established, the displacements of
hard tissues are now used to predict the 22 loadings of the principal
components; further the soft tissue changes are calculated from the predicted
loadings.
2.4.4.2 Validation of prediction accuracy.
Leave-one-out cross validation was applied to validate the approach of
prediction of soft tissue changes. In this validation, 29 of the 30 cases were used
as a training set to establish the relationship between the loadings of the first 22
principal components of the displacements of soft tissues and the displacements
of hard tissues, while the remaining one case was used for validation. It is an
iterative procedure so the next layer of analysis is to choose a different case for
the sample
Anas Almukhtar 2016
146
Chapter Two
Anas Almukhtar 2016
A
B
C
Figure 41: soft tissue changes in the three dimensions: (x) A, (y) B and (z) C.
Note the colour scale is different from the normal colour code used with closest surface distance measurement. In figure A (horizontal
changes) the blue side of the scale highlights the regions displaced to the left side while the red side of the scale highlights the region
that was displaced to the right side. In figure B (vertical changes) red highlight were given to region displaced upward while blue
highlights were given to the regions displaced downward. In figure C (Anteroposterior changes) the red colour highlights the
advancing regions while the blue colour highlights the setback regions of the facial soft tissue surface.
Methodology
147
Chapter Two
Methodology
as a leave out case and use the rest as training set. The results were 30 trials
providing the mean and standard deviation of the surface distance between the
predicted displacements and the actual displacements of the validation case at
each of the examined region. This procedure was applied to the regions of the
upper lip, lower lip, chin, nose, and the right/left paranasal areas.
The mean and standard deviation of the differences between the predicted and
the actual displacements of each region were calculated and a boxplot was used
to display the results. Colour coded maps showing the mean difference between
the predicted and the actual post-operative facial changes in the x, y, and z
dimensions were also displayed.
Anas Almukhtar 2016
148
3
Results
C
ontents
3.1 ERROR STUDY.................................................................................................................. 150
3.2
ANALYSIS OF SKELETAL SURGICAL MOVEMENTS ........................................................................ 155
3.3
ANALYSIS OF SOFT TISSUE CHANGES FOLLOWING SURGERY ......................................................... 156
3.3.1
SOFT TISSUE RESPONSE TO LE FORT I MAXILLARY ADVANCEMENT. .................................................. 156
3.3.2
SOFT TISSUE RESPONSE TO BSSO MANDIBULAR ADVANCEMENT. .................................................... 163
3.3.3
SOFT TISSUE CHANGES FOLLOWING BI-MAXILLARY ADVANCEMENT. ................................................. 167
3.4
PREDICTION OF FACIAL SOFT TISSUE CHANGES FOLLOWING LE FORT I ADVANCEMENT SURGERY. .......... 172
3.4.1
UPPER LIP ............................................................................................................................ 173
3.4.2
LOWER LIP ............................................................................................................................ 176
3.4.3
CHIN ................................................................................................................................... 179
3.4.4
NOSE .................................................................................................................................. 182
3.4.5
PARANASAL LEFT ................................................................................................................... 185
3.4.6
PARANASAL RIGHT ................................................................................................................. 188
Anas Almukhtar 2016
149
Chapter Three
Results
Introduction
A total of 137 pre-surgical and post-surgical CBCT scans were successfully
retrieved from the database at the radiography department at the dental
hospital. 37 cases were excluded from the sample for various reasons according
to our exclusion criteria, (Section 2.1.1). Image quality and ethnic background
were the two main reasons for exclusion. The remaining 100 cases were
successfully processed according to the research protocol.
3.1
Error study
Landmarking a CBCT derived soft tissue image has always been challenging due
to the relatively smooth surface and the lack of facial texture information.
Therefore intra observer landmarking error was investigated. Table (22) shows
the
measured
orthogonal
distances
between
the
coordinates
of
the
corresponding 20 landmarks. The digitization was repeated two weeks later.
These were the same landmarks employed for the conformation of the generic
mesh of 30 cases randomly selected from the study sample.
In the x dimension (medio-lateral), the landmarking error ranged from (0.27 mm
± 0.37 mm) to (0.72 mm ± 0.89 mm). The mean error of the repeated
digitization of the 20 landmarks was (0.45 mm ± 0.13 mm). In the z dimension
(the depth), landmarking error ranged from (0.12 mm ± 0.18mm) to (0.42 mm ±
0.5 mm), the mean error of the repeated digitization of the 20 landmarks was
(0.23 mm ± 0.17 mm). In the y dimension (Vertical), the landmarking cases
ranged from (-0.37 mm ± 0.25 mm) to (0.87 mm ± 1.15 mm), the mean error of
the repeated digitization of the 20 landmarks was (0.13 mm ± 0.19 mm). The
only landmark that showed a statistically significant difference in the
coordinates between repeated digitization which was limited to the x dimension
was the Inner canthus-Right (0.44 mm ± 0.63 mm, p=0.01). Paired sample
correlation
test
confirmed
a
high
correlation
between
the
repeated
measurements for all the landmarks.
Anas Almukhtar 2016
150
Chapter Three
Results
Table 22: Landmarking error (orthogonal distance) between the repeated digitization of landmarks used for generic mesh conformation.
No:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Mean x
0.27
0.32
0.45
0.44
0.34
0.28
0.42
0.35
0.33
0.51
0.42
0.44
0.50
0.53
0.68
0.51
0.48
0.49
0.58
0.72
Mean
SD
SD
P value
Corr.
Mean z
SD
P value
Corr.
Mean y
SD
P value
Corr.
0.37
0.76
0.64
0.63
0.55
0.38
0.61
0.58
0.53
0.59
0.55
0.58
0.77
0.62
0.87
0.65
0.61
0.65
0.75
0.89
0.89
0.10
0.21
0.01
0.2
0.41
0.98
0.28
0.74
0.11
0.81
0.74
0.88
0.07
0.95
0.79
0.33
0.71
0.45
0.88
0.45
0.13
0.997
0.990
0.994
0.992
0.994
0.996
0.994
0.993
0.994
0.991
0.993
0.993
0.986
0.991
0.984
0.991
0.993
0.991
0.987
0.983
0.27
0.17
0.16
0.12
0.4
0.26
0.22
0.33
0.13
0.16
0.12
0.2
0.18
0.12
0.42
0.33
0.39
0.39
0.08
0.12
0.53
0.3
0.23
0.18
0.56
0.38
0.65
0.57
0.2
0.22
0.21
0.28
0.27
0.17
0.5
0.43
0.53
0.49
0.12
0.18
0.23
0.17
0.16
0.21
0.27
0.93
0.57
0.12
0.34
0.85
0.68
0.16
0.84
0.26
0.72
0.43
0.67
0.78
0.75
0.97
0.54
0.67
0.997
0.999
0.999
1.000
0.997
0.999
0.996
0.997
1.000
1.000
1.000
0.999
0.999
1.000
0.998
0.999
0.997
0.998
1.000
1.000
0.13
0.13
0.55
0.19
0.21
0.19
0.62
0.22
0.16
0.36
0.34
0.29
0.23
0.61
0.47
0.48
0.48
0.48
0.42
0.87
0.2
0.19
0.75
0.28
0.3
0.26
0.85
0.31
0.22
0.46
0.46
0.38
0.36
0.78
0.58
0.6
0.61
0.63
0.59
1.15
0.37
0.25
0.87
0.06
0.78
0.15
0.20
0.29
0.75
0.10
0.86
0.30
0.40
0.45
0.93
0.53
0.58
0.32
0.65
0.53
0.16
0.54
1.000
1.000
0.999
1.000
1.000
1.000
0.998
1.000
1.000
0.999
0.999
0.999
0.999
0.998
0.998
0.998
0.998
0.998
0.999
0.995
Anas Almukhtar 2016
151
Chapter Three
Results
The Euclidean distance landmarking errors were calculated. Table (23) shows
the mean and standard deviation of the repeated measurements at each of the
20 landmarks. The error ranged from (0.45 mm ± 0.53 mm) to (1.26 mm ± 0.71
mm) with a mean of (0.72 mm ± 0.22 mm).
Anas Almukhtar 2016
152
Chapter Three
Results
Table 23: Landmarking error (Euclidean distance) between the repeated digitisations of landmarks used for generic mesh conformation.
Exc-R
End-R
1
1.98
0.81
2
0.22
3
Na
End-L
Exc-L
Ab-R
Prn
Ab-L
Ch-R
FL-R
Ls
FL-L
Ch-L
2.21
1.96
0.81
0.79
1.19
0.46
0.52
0.18
0.77
0.47
0.34
0.12
1.20
0.62
0.03
1.51
0.08
0.42
0.26
0.75
0.52
0.67
0.23
0.22
0.36
0.39
0.61
1.48
0.69
0.05
0.21
0.21
1.73
4
0.52
0.19
0.71
0.33
0.71
0.43
1
0.73
0.55
1.24
5
0.30
1.33
0.58
0.8
0.07
0.21
1.69
0.28
0.69
6
0.26
0.23
0.52
0.39
1.23
0.27
0.92
1.05
7
0.34
0.07
1.56
0.64
0.12
0.78
0.48
8
0.51
0.19
1.21
0.31
0.75
0.22
9
0.18
0.12
1.26
0.80
0.30
0
2.42
0.25
1.06
2.57
11
0.17
4.15
2.40
12
0.57
0.15
13
1.05
14
Li
UM-R
UM-L
LM-R
LM-L
Lm
Pog
0.86
1.03
1.80
0.44
0.24
0.60
1.12
0.55
1.05
1.06
0.60
0.35
1.59
1.74
2.59
0.65
0.65
1.05
0.82
1.14
0.87
0.39
0.22
2.33
0.51
1.71
0.12
0.89
0.64
0.74
1.00
0.57
0.71
0.25
0.68
1.68
0.91
0.27
1.20
1.31
1.02
1.98
1.97
0.50
1.88
2.01
0.34
0.56
0.58
0.72
0.58
0.59
1.39
0.51
0.24
0.16
1.44
0.4
0.12
0.94
0.22
0.58
0.38
1.27
0.25
1.68
1.15
1.13
0.10
0.86
1.26
0.16
0.34
0.38
0.65
0.71
0.06
1.03
0.42
0.61
0.65
0.57
0.52
1.27
1.02
1.27
0.97
0.19
0.48
0.13
0.41
0.64
0.65
2.30
0.81
0.42
0.39
0.28
2.89
1.45
0.14
2.17
2.18
0.04
0.09
0.43
0.34
0.04
1.13
0.68
0.90
0.34
2.24
0.40
2.23
0.7
0.36
0.94
0.60
0.35
0.64
1.06
0.32
0.35
0.72
0.81
1.22
0.48
0.92
0.54
0.79
2.19
1.14
0.08
0.11
0.28
4.14
0.05
0.15
0.71
0.82
0.63
0.93
1.61
0.37
1.37
0.28
1.90
0.20
1.80
0.08
1.19
0.15
0.45
0.25
0.63
0.14
1.27
1.19
0.29
0.14
2.03
0.41
2.16
0.82
0.33
0.98
0.16
1.48
0.79
0.28
1.11
0.83
0.45
0.43
0.59
0.64
0.56
0.91
0.11
0.60
1.53
0.12
1.15
0.55
0.52
0.97
0.49
0.76
15
0.02
0.37
0.66
0.17
0.08
0.15
1.63
0.13
0.67
0.14
0.71
0.73
0.17
0.76
1.73
0.78
0.41
0.36
0.64
0.85
16
0.58
0.89
1.00
1.14
1.15
0.27
2.00
0.37
0.00
0.91
0.61
0.20
0.27
0.84
1.30
1.14
0.40
0.36
0.55
1.21
17
0.16
0.76
0.4
0.83
0.58
0.36
0.21
0.35
0.13
0.57
0.23
0.58
0.29
0.39
0.49
0.38
0.69
0.08
1.36
0.84
18
0.28
0.28
0.46
0.33
0.22
0.17
0.42
3.18
0.6
0.85
0.93
0.30
0.33
2.14
0.89
0.51
1.12
0.75
1.09
0.65
19
0.12
0.28
0.24
0.03
0.40
0.51
0.27
0.44
0.08
0.48
0.26
0.38
0.41
0.45
0.74
0.86
0.52
1.46
0.99
0.11
Continue
Anas Almukhtar 2016
153
Chapter Three
Results
Exc-R
End-R
20
0.27
0.15
21
0.36
22
Na
End-L
Exc-L
Ab-R
UM-R
UM-L
LM-R
LM-L
0.60
0.23
0.17
0.17
0.18
0.21
0.16
0.94
0.9
0.42
0.12
0.78
0.17
1.50
1.23
0.78
0.76
0.50
0.06
0.63
0.24
2.9
0.55
0.30
0.35
0.37
0.69
0.50
0.55
0.45
0.84
0.65
0.12
0.69
0.47
0.97
0.62
0.30
0.16
0.33
1.04
0.39
0.23
0.79
0.32
0.08
1.21
1.30
0.56
0.22
0.18
0.95
0.33
0.64
1.16
1.16
1.38
23
0.38
0.45
0.85
0.09
0.14
0.67
0.65
1.56
0.24
1.21
0.80
0.51
0.19
0.28
0.78
1.19
0.58
0.49
2.05
1.39
24
0.30
0.05
0.23
0.24
0.34
0.05
0.28
0.58
0.14
0.91
0.29
0.19
1.93
1.85
0.36
0.18
0.80
1.42
1.64
1.36
25
0.12
0.34
0.22
0.18
0.50
0.77
0.07
0.34
0.52
1.34
0.95
1.21
0.16
0.75
1.86
0.05
1.57
0.54
1.94
1.95
26
0.22
0.27
0.43
0.16
0.57
0.16
0.75
0.58
0.38
0.48
0.39
0.35
1.75
1.48
1.02
0.93
1.12
1.27
1.65
1.28
27
0.07
0.08
1.54
0.45
0.98
0.27
0.80
0.13
1.00
0.30
0.39
1.63
0.61
0.91
0.71
1.20
1.91
0.94
0.37
0.22
28
0.04
0.06
0.18
0.12
1.58
0.46
1.33
0.12
0.53
0.63
1.02
0.33
0.20
0.57
2.17
0.75
0.71
0.33
0.60
0.80
29
0.26
0.38
0.22
0.15
0.24
0.30
0.70
0.39
0.35
1.09
0.32
1.00
2.20
1.69
0.97
0.75
1.52
0.77
0.95
0.83
30
0.43
0.17
0.11
mea
n
0.09
0.45
0.08
0.14
0.33
0.13
0.42
0.26
1.27
0.50
0.52
1.01
1.45
2.10
0.86
0.23
0.83
0.45
0.43
0.82
0.54
0.60
0.46
0.91
0.58
0.43
0.71
0.62
0.63
0.63
0.90
0.99
0.87
0.86
0.86
0.79
1.26
SD
0.53
0.76
0.58
0.57
0.60
0.38
0.83
0.67
0.42
0.36
0.42
0.39
0.62
0.49
0.57
0.45
0.52
0.56
0.56
0.71
Overall Mean
0.72
Overall SD
0.22
Prn
Ab-L
Ch-R
FL-R
Ls
Anas Almukhtar 2016
FL-L
Ch-L
Li
Lm
Pog
154
Chapter Three
3.2 Analysis of skeletal surgical movements
Results
Table (24) shows the mean displacement of the maxilla and mandible in 3D
following orthognathic surgery. Three groups were analysed; Le Fort I maxillary
advancement, Bilateral sagittal split osteotomy (BSSO) mandibular advancement
and Bimaxillary advancement groups. The measurements were based on
landmarks at specific anatomical locations and recorded in the x, y and z
dimensions (Section 2.4.1.2).
Le Fort I group expressed mainly a forward maxillary displacement of 5.95 mm.
This was combined with minimal lateral maxillary displacement of 0.05 mm,
anterior and posterior upward maxillary impaction of 0.14 mm and 0.6 mm
respectively. The effect of Maxillary advancement extended to involve the
mandible where 2.79 mm advancement and 2.13 mm upward displacement were
recorded at the mandibular symphysis.
In the mandibular advancement group, forward and downward displacements of
the mandible of 3.51 mm and -2.12 mm with minimal lateral displacement of 0.09mm were recorded at the mandibular symphysis.
In the bimaxillary advancement group, both the maxilla and mandible were
displaced anteriorly by 5.51mm and 4.58mm respectively. The mandible was
displaced 5mm upward at the symphysis while the maxilla was simultaneously
impacted anteriorly and posteriorly by 2.74 mm and 2.36 mm respectively.
Minimal lateral displacement was detected in both jaws at 0.19 mm and 0.4 mm
at the maxilla and mandible respectively.
Table 24: Measurements (mm) of skeletal displacement following orthognathic surgery
Skeletal displacement
Anterior maxillary displacement
Anterior mandibular displacement
Lateral maxillary displacement
Lateral mandibular displacement
Vertical maxillary displacement anteriorly
Vertical maxillary displacement posteriorly
Vertical mandibular displacement
Anas Almukhtar 2016
Le Fort I
Mean
5.95
2.79
0.05
0.21
0.14
0.6
2.13
SD
1.79
2.09
0.95
0.97
1.72
1.48
1.86
BSSO
Mean
-----3.51
------0.09
-----------2.12
SD
-----2.64
-----1.24
----------2.71
Bimax
Mean
5.51
4.58
0.19
0.47
2.74
2.36
5.00
STDV
2.79
3.28
1.50
3.18
2.59
1.38
3.68
155
Chapter Three
Results
3.3 Analysis of soft tissue changes following surgery
3.3.1 Soft tissue response to Le Fort I maxillary advancement.
3.3.1.1 Euclidean distance measurements
The average face models of the preoperative and post-operative facial meshes
were superimposed on the eye region and the corresponding distances between
the vertices on both models were presented in a colour coded map. The colour
pattern represented the net Euclidian distance between the corresponding
vertices, (figure 42). Considering the geometric centre of the facial mesh, the
mesh vertices which were displaced in an outward direction due to Le Fort I
surgery were highlighted in red whereas these displaced in an inward direction
due to Le Fort I surgery were highlighted in blue . The colour spectrum ranging
from red (maximum outward) to blue (maximum inward) assigned to each vertex
of the mesh represented the variation in positional change of each vertex. The
green colour represents the zero change and located at the centre of the
spectral colour scale. The colour scale was displayed on the side of the image
with the displacement values been assigned to each colour segment. The results
are subcategorised into the following anatomical regions:
1. The nose: The nasal bridge area showed no changes due to surgery,
whereas the difference gradually increased toward the nasal cartilage and
the tip of the nose. The most obvious changes in response to surgery were
observed at the alar cartilage which was highlighted in red, the intensity
of the red colour increased close to the nasolabial groove. This indicates
widening of the nostrils following Le Fort I surgery. The region at the
columella and the lower part of the nasal tip was highlighted in blue
which indicates an inward movement toward the geometric centre of the
face. This combined with the faint orange patches on the nasal tip
indicates an minimal upward displacement of the nasal tip increases
posteriorly at the columella secondary to Le Fort I osteotomy.
2. The upper lip: The upper lip region is bounded by the nasolabial junction
superiorly, extends to the inferior end of the vermillion border and to the
oral commissures bilaterally. The majority of the region showed a uniform
dark red highlight which indicates a marked anterior displacement in the
Anas Almukhtar 2016
156
Chapter Three
Results
upper lip region in response to surgery. A change in colour to light green
at the commissures was observed which indicated the limited effect of
the maxillary advancement on this region.
3. The lower lip: the region of the lower lip is anatomically bounded by the
upper border of the lower lip vermilion border superiorly, extends to the
labiomental fold inferiorly and to the oral commissures laterally. The
majority of this region displayed a light red highlight due to the slight
forward displacement at the lower lip. A colour change into a light blue at
the area below the vermilion border extending to the labiomental fold
downward and about 50% of the lip width was observed which indicates a
backward and upward displacement of the lower lip at this region.
4. The chin: The chin area extends from the labiomental fold to the lower
border of the face and extends laterally to the area marked by an
imaginary vertical lines descending from the oral commissures. The
majority of the area was highlighted in red which appeared darker
centrally at the pogonion region. A blue strip at the lower border of the
chin was also observed, (figure 2,A). This together with the red colour at
'pogonion' point indicates a minor combined upward and forward change
at chin region.
5. Paranasal regions: The red colour display extended bilaterally to the
nose and limited laterally to a vertical line from the outer canthi.
Superiorly, there was a display of the red colour immediately below the
malar eminence which extended down to the level of the oral
commissures.
6. The cheeks: The light red colour highlight in this area indicates a minor
change in response to maxillary surgery. A change to the blue colour at
the lower border was noted. This was continuous with the strip of blue
colour below the chin area. These findings indicated a minor upward
movement of the soft tissue at the lower border of the mandible
associated with the shortening of the lower facial height.
Anas Almukhtar 2016
157
Chapter Three
Results
A
B
Figure 42: Dense anatomical correspondence (Euclidean ) showing soft tissue changes
following Le Fort I advancement surgery in Front view (A) and 45º view (B)
In summary, the results shows generalised forward movements at the mid and
lower face regions associated with widening of the nostrils and a limited
shortening of the lower facial height.
3.3.1.2 Directional soft tissue changes
The soft tissue changes following Le Fort I osteotomy were analysed for each of
the x, y and z dimensions separately. Therefore, different colour spectrum was
assigned for this part of the analysis. The change at surface vertices due to
surgery to the right or the top sides of the screen, or toward the observer, was
considered a positive change and highlighted in red. While the change to the left
or to the bottom of the screen or away from the observer was considered
negative and were highlighted in blue. Green colour was for the regions where
no changes were detected. The intensity of the colour indicated the magnitude
of the change at the examined area.
Anas Almukhtar 2016
158
Chapter Three
Results
The colour scale was displayed on the left side of the face, (figure 43). The
higher and lower extremes of the spectral colour scale were limited to 5.5mm
which was recorded as the mean A-P skeletal displacement in Le Fort I maxillary
advancement group.
3.3.1.2.1 Soft tissue changes in x dimension (medio-Lateral).
Figure (43,A) shows the soft tissue surface changes in x dimension only. Minimal
changes were noted around the eye regions and the nasal bridge was highlighted
mostly in green which indicates a minimal change in the transverse direction as
a result of surgery. The colour changed to yellow toward the nasal tip. A yellow
colour patch on the right side at the lower part of the dorsum of the nose and a
blue colour patch on the right side indicated a minor narrowing of the nose at
this area. The opposite was observed at the alar region, the right ala of the nose
displayed a blue colour while the left ala of the nose displayed a red colour
which is an indication of widening of the nostrils.
Most of the soft tissue changes in response to Le Fort I osteotomy were in the
reign of the upper lip and paranasal area. This was demonstrated by the welldefined distribution of the orange colour on the left side and the blue colour on
the right side of these two regions which indicates a tendency toward a lateral
expansion along these anatomical regions. The changes were limited to the
anatomical boundaries of the paranasal region and upper lip. The changes at the
commissures at the corners of the mouth, the lower lip and the chin were
minimal.
Anas Almukhtar 2016
159
Chapter Three
Anas Almukhtar 2016
A
B
C
Figure 43: corresponding (directional ) soft tissue changes in: X (Medio-lateral) dimension (A); Y (vertical) dimension and Z( antero-posterior)
dimension following Le Fort I advancement surgery.
Results
160
Chapter Three
Results
3.3.1.2.2 Soft tissue changes in y dimension (Vertical change)
Figure (43,B) shows the vertical component of soft tissue changes only in y
dimension in response to Le Fort I osteotomy.
Most of the region around the eyes and the nasal bridge displayed green colour
which confirms the trivial vertical changes of these areas. The dorsum of the
nose showed minimal changes and displayed a green strip which extended
toward the tip of the nose. The tip of the nose, the right and left alar curvatures
and the columella showed a mild upward movement which was displayed as a
yellowish orange colour.
The paranasal area showed a homogenous but mild upward movement coinciding
with the adjacent cheek areas which displayed predominantly in a yellowish
green colour. The upper lip displayed mostly green colour with yellow/green
patches closer to the nostrils. This indicates there was a minimal vertical
displacement of the upper lip following Le Fort I maxillary advancement.
Vertically, the lower lip showed two types of response to Le Fort I maxillary
advancement separated by the vermillion border. The region above the
vermilion border showed minimal vertical changes highlighted in green yellow,
the region below the vermillion border showed a clear vertical change displayed
in orange manly at the central region. This indicated an upward movement in
this region and perhaps a minimal shortening of the lower lip length. The chin
area on the other hand showed a predominant orange colour with the highest
intensity centrally then decreased posteriorly toward the cheek areas on both
sides. This was an indication of an upward movement of the chin area.
3.3.1.2.3 Soft tissue changes in z dimension (Antero-posterior changes)
Figure (43,C) shows the depth component (A-P) of the corresponding soft tissue
changes in response to surgery. The region around the eyes and the nasal bridge
showed a minimal change which was displayed as generalised green colour. The
green colour continued on the dorsum of the nose toward the nasal tip with no
Anas Almukhtar 2016
161
Chapter Three
Results
marked change in colour. The columella together with the right and left alae of
the nose showed orange colour indicating a marked forward movement of this
region.
A well-defined dark red region covered the upper lip and the paranasal Regions
which indicates a marked forward displacement. The colour was confined to the
anatomical boundaries of these two regions. The effect of surgery on the
commeasures at the corner of the mouth was minimal, displayed as a yellowish
green colour. Green colour was displayed on the lower lip. The chin showed a
relatively homogenous mostly orange colour which indicates a moderate forward
displacement secondary to surgery. Similar changes were evident on the lower
part of the cheeks where the colour intensity reduced posteriorly.
Summary
In summary, there were distinctive forward soft tissue movements combined
with a marked lateral expansion at the upper lip and paranasal regions, these
changes were limited by the anatomical boundaries of these regions. The
commeasures at the corners of the mouth showed minor changes in all
directions. The effect extended to the chin area where a marked vertical and
to a lesser extent forward displacement were evident. The lower lip showed
minimal changes in all dimensions except a marked vertical movement at the
region between the vermilion border and the labiomental fold. The effect on
the nose was limited to the widening and advancement of the nostrils and
the same was noted at the base of the nose (Subnasale and alar base) region
with a minimal forward change at the nasal tip. The changes at the cheeks
were to a lesser extent with minimal but consistent upward and forward
displacement of the soft tissue in response to Le Fort I advancement
osteotomy.
Anas Almukhtar 2016
162
Chapter Three
Results
3.3.2 Soft tissue response to BSSO mandibular advancement.
3.3.2.1 Euclidean distance measurements
The pre and post-operative average facial models were superimposed and dense
anatomical correspondence was created following the same protocol (Section
3.2.1). The results were displayed as a colour coded map, (figure 44).
A
B
Figure 44: Dense anatomical correspondence (Euclidean) showing soft tissue changes
following BSSO advancement surgery in Front view (A) and 45º view (B)
Unlike Le Fort I osteotomy, the soft tissue changes were confined to the lower
third of the face corresponding to mandibular advancement. No marked changes
were observed on the nose, paranasal areas, and upper lip regions where green
colour was predominant. The lower lip showed a mild change at the vermilion
border near the midline demonstrating the roll-in effect of the surgery on the
lower lip. The marked colour change started to show just below the vermilion
border and increase in intensity towards the chin region. The red colour
continued backward along the mandibular border towards the gonial angle.
Anas Almukhtar 2016
163
Chapter Three
Results
3.3.2.2 Directional changes measurements
In order to investigate the changes in more details, soft tissue analysis was
segmented into the x, y, and z dimensions where changes at each of the three
dimensions were analysed independently.
3.3.2.2.1 Soft tissue changes in x dimension (medio-Lateral)
Figure (45,A) shows the soft tissue surface changes in x dimension only. The
changes around the eye regions and the nose were minimal, displayed mostly in
green colour. The changes around the midface region were observed as a mild to
moderate narrowing of the upper lip mainly at the sides near the commeasures
of the mouth displayed as a yellow/orange colour on the right side and light blue
on the left side of the face, whereas the commissures of the mouth expressed a
minimal changes which displayed in green colour. Lateral to the Upper lip the
same colour pattern extended slightly both sides. The lower lip demonstrated a
tendency toward minimal to mild lateral expansion marked by an orange colour
on the left side and central regions combined with a bluish green colour on the
right side of the lip. The results also highlighted a mild shift of the chin to the
left side which coincided with the recorded skeletal lateral shifting of the
mandible.
3.3.2.2.2 Soft tissue changes in y dimension (vertical change)
Figure (45,B) shows the vertical component of soft tissue changes only in y
dimension in response to BSSO mandibular advancement surgery.
Most of the regions around the eyes, nose and paranasal region displayed green
colour which confirms the trivial vertical changes. The upper lip demonstrated a
tendency toward downward movement which is displayed in blue colour, this
diminished laterally to a minimal change at the commeasures of the mouth. The
blue colour on the upper lip combined with the minimal change at the nose
highlighted a mild tendency toward lengthening of the upper lip.
Anas Almukhtar 2016
164
Chapter Three
Anas Almukhtar 2016
A
B
C
Figure 45: Corresponding (directional) soft tissue changes in: X (Medio-lateral) dimension (A); Y (vertical) dimension and Z( antero-posterior)
dimension following Mandibular advancement surgery.
Results
165
Chapter Three
Results
The vermilion border of the lower lip showed upward movement displayed as a
yellow/orange patches mainly at the centre and diminished laterally toward the
commeasures of the mouth. This change highlighted the inward and upward
rolling of the lower lip as it moved forward and freed from the classical lip trap
under the palatal surface of upper incisors associated with mandibular
deficiency. Downward movement was noted at the region below the vermilion
border of the lower lip, the chin and laterally extended to the cheeks region.
3.3.2.2.3 Soft tissue changes in z dimension (antero-posterior changes)
Figure (45,C) shows the depth component of the corresponding soft tissue
changes in z dimension.
The region around the eyes, nose, paranasal regions and major part of the upper
lip showed a minimal change which was displayed as a generalised green colour.
The main changes were evident at the lower face and lateral cheeks region. A
generalized and homogenous red colour covering these regions indicates a
marked forward movement. The vermilion border of the lower lip showed a
minimal change in the A-P dimension following BSSO advancement surgery. The
same was noted at the commeasures of the mouth indicating a minimal change
due to surgery. The vermilion region of the upper lip showed a mild backward
movement displayed as a blue colour and extends laterally where the colour
changed to green indicating minimal changes at the commissures.
Summary
There was a marked forward movement of the chin and lower lip regions
extended posteriorly toward the gonial angle. A marked narrowing of the
upper lip and widening of the lower lip were also observed. There was a
generalised downward movement of the chin, lower lip and extended
posteriorly toward the cheeks.
Anas Almukhtar 2016
166
Chapter Three
Results
3.3.3 Soft tissue changes following bimaxillary advancement.
3.3.3.1 Euclidean distance measurements
The pre and post-operative average facial models were superimposed and dense
anatomical correspondence was created. The results were displayed as a colours
coded distance map, (figure 46).
A
B
Figure 46: Dense anatomical correspondence (Euclidean) showing soft tissue changes
following Bimaxillary advancement surgery in Front view (A) and 45º view (B).
Soft tissue changes resembled to some extent the combined effect of
advancement Le Fort I osteotomy and BSSO osteotomy. Changes at the eyes
level and nasal bridge were minimal indicating good registration accuracy. The
yellow colour at the tip of the nose was due to a mild forward and upward
movement. The columella showed a mild upward displacement marked in light
blue while the alar cartilages flared laterally and were displayed in orange. The
upper lip and paranasal regions showed the major changes at the midface
marked by a darker red colour in a relatively similar fashion to cases which had
had a Le Fort I osteotomy only, (Section 3.2.1). However, in bimaxillary
Anas Almukhtar 2016
167
Chapter Three
Results
advancement the red colour extended to cross the boundaries of the upper lip
and paranasal regions laterally and inferiorly as it merged with the changes
associated with the lower face as a result of the associated mandibular
advancement. A moderate changes were observed at the oral commissures and
lower lip vermilion. The changes started to increase toward the chin area
marked by a darker red colour. Major changes were only observed at the chin
area where the red colour started to fade into lighter shades laterally towards
the gonial angles.
3.3.3.2 Directional changes measurements
In order to investigate the changes in more details, soft tissue analysis was
separated into the x, y, and z dimensions where changes at each of the three
dimensions were analysed independently.
3.3.3.2.1 Soft tissue changes in x dimension ( medio-Lateral)
Figure (47,A) shows the soft tissue surface changes in x dimension only. Minimal
changes were noted around the eye regions and the nasal bridge which were
highlighted in green. The changes around the midface region are close to the
changes observed following Le Fort I maxillary advancement, (section 3.2.1.2.1).
The region around the nasal bridge and nasal tip demonstrated a green colour
which indicates minimal changes at these regions. The right ala of the nose
displayed a blue colour while the left ala of the nose displayed a yellowish-red
colour, an indication of widening of the nostrils.
The upper lip and paranasal area showed a clear tendency toward lateral
expansion demonstrated as a yellow/red colour on the left side of the face and
the blue colour on the right side of these two regions. The changes were limited
to the anatomical boundaries of the paranasal region and upper lip and was less
than that observed with Le Fort I maxillary advancement cases (Section
3.2.1.2.1). The commissures at the corners of the mouth showed minimal
tendency toward narrowing as demonstrated by a blue colour on the left
commissure and yellow colour on the right side. The lower lip showed a similar
Anas Almukhtar 2016
168
Chapter Three
Results
tendency toward narrowing as demonstrated by a blue colour on the left side
and yellow colour on the right side of the lip. The chin region showed minimal
change displayed in green colour.
3.3.3.2.2 Soft tissue changes in y dimension (vertical change)
Figure (47,B) shows the vertical component of soft tissue changes in response to
Bi-maxillary advancement surgery.
Most of the region around the eyes and the nasal bridge displayed green colour
which confirms the trivial vertical changes of the soft tissue of this region to bimaxillary osteotomy. The dorsum of the nose showed minimal changes and was
displayed in green/yellow colour. The tip of the nose, the right and left alar
curvatures and the columella showed a marked upward movement which was
displayed in orange/ red colour.
The paranasal area showed a homogenous but mild upward movement. The
upper lip displayed yellowish green colour with patches of orange colours. This
indicates there was a minimal vertical displacement of the upper lip following
Bi-maxillary advancement surgery in the vertical dimension. The lower lip
showed a compound response to surgery. The vermillion border showed minimal
vertical changes highlighted in green, the region below the vermillion border
demonstrated more vertical change displayed in red/orange colour with a
central dark red colour patch. This indicated an upward movement in this region
and perhaps a minimal shortening of the lower lip length. An upward movement
was observed at the chin area which was marked by a well-defined red patch at
the chin region.
Anas Almukhtar 2016
169
Chapter Three
Anas Almukhtar 2016
A
B
C
Figure 47: Corresponding (directional) soft tissue changes in: X (Medio-lateral) dimension (A); Y (vertical) dimension and Z( antero-posterior)
dimension following Bimaxillary advancement surgery
Results
170
Chapter Three
Results
3.3.3.2.3 Soft tissue changes in z dimension (antero-posterior changes)
Figure (47,C) shows the depth component of the corresponding soft tissue
changes in z dimension.
The region around the eyes and the nasal bridge showed a minimal change which
was displayed as a generalised green colour. The green colour on the dorsum of
the nose extended toward the nasal tip with no marked change in colour. A
change to red colour was evident at the nasal tip. The columella together with
the right and left alas of the nose showed a well-defined dark red colour
indicating a marked forward movement of this region. The forward displacement
at this region coincided with a generalised forward displacement of most of the
mid and lower face regions including the paranasal region, the upper lip and oral
commeasures, the lower lip, the chin and extended posteriorly to involve most
of the cheeks regions which were marked with a uniform dark red colour.
Summary
There was distinctive forward soft tissue movements in the mid and lower
face combined with a marked lateral expansion at the nostrils, upper lip and
paranasal regions. The expansion was limited by the anatomical boundaries
of these three regions whereas a milled narrowing of the oral commissures
and lower lip was observed. Marked vertical upward changes were evident at
the tip of the nose, part of the lower lip and the chin regions. A relatively
mild upward movement was observed at the upper lip and paranasal regions.
Anas Almukhtar 2016
171
Chapter Three
Results
3.4 Prediction of facial soft tissue changes following Le
Fort I advancement surgery.
The correlation between the soft and hard tissue was calculated for each of the
facial anatomical regions, these included: The upper lip, lower lip, chin, nose,
right paranasal and left paranasal regions. Soft tissue prediction and validation
has been carried out using the statistical test “Leave-One-Out-Cross-Validation”.
Details about this test have been provided in the related methodology section.
Thirty cases were included in the process, 29 of them were used as a training
group and the resultant simulation was tested by applying the simulation
algorithm on the preoperative image of the remaining one case. The difference
between the actual post-operative change and the prediction was calculated.
The process was then repeated 30 times to predict a case by case.
The results are visually displayed with a colour coded map showing the
differences (errors of prediction) in the three dimensions between the predicted
and the actual post-operative changes for each of the anatomical regions. The
mean shape difference of the 30 trials is presented in each of the three
dimensions (x, y and z). A Box-and-whisker plot was produced showing the
median and the range of the 30 trials and the 25 percentile for each of the
individual facial regions also in x, y, and z dimensions separately.
The mean values ranged from as low as (0.01 mm ± 0.582 mm) at the nose region
to (-0.03 mm ± 2.1996 mm) at the chin region. Although the mean values were
relatively small, there was a general tendency of over estimation of the
predicted change of soft tissue in response to Le Fort I advancement osteotomy
in the x dimension (medio-lateral) and y dimension (depth) while an
underestimation of the vertical changes was noted. This was demonstrated on
the colour coded map.
Anas Almukhtar 2016
172
Chapter Three
3.4.1 Upper lip
Results
The results of the upper lip prediction are shown in figures (48 and 49). The
difference between the predicted and the actual post-operative changes were
individually analysed in the three dimensions.
The x dimension (Medio-lateral) showed the lowest mean ± standard deviation
among other dimensions (-0.02 mm ± 1.04 mm) followed by the z dimension (the
depth) and the y dimension (vertical) dimension with a mean value of (0.12 mm
± 0.96 mm) and (-0.07 mm ± 1.08 mm) respectively. In general, the total mean ±
standard deviation in all dimensions was lower than 1.2 mm.
Figure (48) shows the difference in the three dimensions (x, y, and z) Box plot.
The results showed that 50% of the trials showed less than 0.5 mm prediction
errors. The size of the boxes showed a relatively low variation among the trials
which was comparable in the three dimensions and equally distributed around
the zero point. The range of the data represented by the whiskers showed a
relatively lower level of variations in z dimension.
Figure (49) shows the colour map of the differences between the predicted and
the actual post-operative soft tissue changes. There was a general tendency
toward an overestimation (over prediction) of the changes at the upper lip
region of around in the z (depth) dimensions. Whereas an under estimation of
the same values were expressed in the y (vertical) dimensions as displayed on
the associated colour scale.
Anas Almukhtar 2016
173
Chapter Three
Results
Figure 48: Accuracy of prediction at the upper lip region. Illustrates the median, the three
quartiles and the whiskers of the errors associated with the prediction of soft tissue change
following Le Fort I maxillary advancement at the Upper lip region in the three dimensions.
Anas Almukhtar 2016
174
Chapter Three
Anas Almukhtar 2016
A
B
C
Figure 49: Accuracy of the prediction at the upper lip region. Colour map showing the (directional ) mean error magnitude between the predicted
Upper lip region and the real post-operative changes in the three dimensions: X (Medio-lateral) dimension (A); Y (vertical) dimension and Z(
antero-posterior) dimension.
Results
175
Chapter Three
Results
3.4.2 Lower lip
The results of the Lower lip prediction are shown in figures (50 and 51). The
difference between the predicted and the actual post-operative changes were
individually analysed in the three dimensions.
The x dimension (medio-lateral) showed the lowest mean ± standard deviation
among other dimensions (-0.02 mm ± 1.09 mm) followed by the y dimension
(vertical) dimension and the z dimension (the depth) with a mean value of (-0.07
mm ± 1.81 mm) and (-0.16 mm ± 1.27 mm) respectively. In general, the total
mean ± standard deviation in all dimensions was lower than 1.44 mm.
Figure (50) shows the difference in the three dimensions (x, y and z) Box plot.
The results showed that 50% of the trials showed less than 1 mm of error in
prediction. The size of the boxes showed a relatively low variation among the
trials which was higher in the y dimension than the other dimensions. All of the
boxed were equally distributed around the zero point. The range of the data
represented by the whiskers showed a higher level of variations in y dimension.
Figure (51) shows the colour map of the differences between the predicted and
the actual post-operative soft tissue changes. There was a general tendency
toward an overestimation (over prediction) of the changes at the lower lip
region in the z (depth) dimension. Whereas an under estimation were expressed
in the y (vertical) dimension as displayed on the associated colour scale.
Anas Almukhtar 2016
176
Chapter Three
Results
Figure 50: Accuracy of the prediction at the lower lip region. Illustrates the median, the three
quartiles and the whiskers of the errors associated with the prediction of soft tissue change
following Le Fort I maxillary advancement at the Lower lip region in the three dimensions.
Anas Almukhtar 2016
177
Chapter Three
Anas Almukhtar 2016
A
B
C
Figure 51: Accuracy of the prediction at the lower lip region. Colour map showing the mean (directional) error magnitude between the
predicted Lower lip region and the real post-operative changes in the three dimensions; X (Medio-lateral) dimension (A); Y (vertical)
dimension and Z( antero-posterior) dimension.
Results
178
Chapter Three
3.4.3 Chin
Results
The results of the chin prediction are shown in figures (52 and 53). The
difference between the predicted and the actual post-operative changes were
individually analysed in the three dimensions.
The z dimension (the depth) showed the lowest mean and standard deviation
among other dimensions (-0.12 mm ± 1.16 mm) followed by the x dimension
(Medio-lateral) and the y dimension (vertical) dimension with a mean value of
(0.07 mm ± 1.76 mm) and (-0.03 mm ± 2.19 mm) respectively. In general, the
total mean ± standard deviation in all dimensions was lower than 2.24mm.
Figure (52) shows the difference in the three dimensions (x, y, and z) box plot.
The results showed that 50% of the trials showed less than 1 mm of error in
prediction. The size of the boxes showed a relatively low variation among the
trials which was higher in the y dimension than the other dimensions. All of the
boxed were equally distributed around the zero point. The range of the data
represented by the whiskers showed a higher level of variations in y dimension.
Figure (53) shows the colour map of the differences between the predicted and
the actual post-operative soft tissue changes. There was a general tendency
toward an overestimation (over prediction) of the changes in all dimensions.
However, a mixed blue and yellow colour patches were noticed over the chin
region as displayed on the associated colour scale.
Anas Almukhtar 2016
179
Chapter Three
Results
Figure 52: Accuracy of the prediction at the chin region. Illustrates the median, the three
quartiles and the whiskers of the errors associated with the prediction of soft tissue change
following Le Fort I maxillary advancement at the Chin region in the three dimensions.
Anas Almukhtar 2016
180
Chapter Three
Anas Almukhtar 2016
A
B
C
Figure 53: Accuracy of the prediction at the chin region. Colour map showing the mean (directional) error magnitude between the predicted Chin
region and the real post-operative changes in the three dimensions; x (Medio-lateral) dimension (A); Y (vertical) dimension and Z( antero-posterior)
dimension.
Results
181
Chapter Three
Results
3.4.4 Nose
The results of the nose prediction are shown in figures (54 and 55). The
difference between the predicted and the actual post-operative changes were
individually analysed in the three dimensions.
The x dimension (medio-lateral) showed the lowest mean ± standard deviation
among other dimensions (0.00 mm ± 0.58 mm) followed by the y dimension
(vertical) dimension and the z dimension (the depth) with a mean value of (-0.04
mm ± 0.69 mm) and (-0.02 mm ± 0.75 mm) respectively. In general, the total
mean ± standard deviation in all dimensions was lower than 0.79 mm.
Figure (54) shows the difference in the three dimensions (x, y, and z) box plot.
The results showed that 50% of the trials showed less than 0.5 mm of error in
prediction. The size of the boxes showed a relatively low variation among the
trials which was higher in the z dimension than the other dimensions. All of the
boxed were reasonably distributed around the zero point. The range of the data
represented by the whiskers showed a lower level of variations in z dimension.
Figure (55) shows the colour map of the differences between the predicted and
the actual post-operative soft tissue changes. Although marginal differences
were recorded, there was a general tendency toward an underestimation (under
prediction) at different regions and dimensions including the flaring of the
nostrils region in the x dimension (medeo-lateral) and the nasal tip displacement
was also underestimated in both the Y (vertical) and z (depth) dimensions as
displayed on the associated colour scale.
Anas Almukhtar 2016
182
Chapter Three
Results
Figure 54: Accuracy of the prediction at the nose region. Illustrates the median, the three
quartiles and the whiskers of the errors associated with the prediction of soft tissue change
following Le Fort I maxillary advancement at the Nose region in the three dimensions.
Anas Almukhtar 2016
183
Chapter Three
Anas Almukhtar 2016
A
B
C
Figure 55: Accuracy pf the prediction at the nose region. Colour map showing the mean (directional) error magnitude between the
predicted Nose region and the real post-operative changes in the three dimensions; X (Medio-lateral) dimension (A); Y (vertical)
dimension and Z( antero-posterior) dimension.
Results
184
Chapter Three
3.4.5 Paranasal Left
Results
The results of the left paranasal region prediction are shown in figures (56 and
57). The difference between the predicted and the actual post-operative
changes were individually analysed in the three dimensions.
The z dimension (the depth) showed the lowest mean and standard deviation
among other dimensions (0.09 mm ± 1.00 mm) followed by the y dimension
(vertical) dimension and the x dimension (Medio-lateral) with a mean value of
(0.02mm ± 1.08 mm) and (0.08 mm ± 1.19 mm) respectively. In general, the
total mean ± standard deviation in all dimensions was lower than 1.26 mm.
Figure (56) shows the difference in the three dimensions (x, y, and z) box plot.
The results showed that 50% of the trials showed less than 1 mm of error in
prediction. The size of the boxes showed a relatively low variation among the
trials which was comparable in all dimensions. All of the boxed were equally
distributed around the zero point. The range of the data represented by the
whiskers showed a comparatively higher level of variations in x dimension.
Figure (57) shows the colour map of the differences between the predicted and
the actual post-operative soft tissue changes. There was a general tendency
toward an overestimation (over prediction) of the changes in all dimensions as
displayed on the associated colour scale.
Anas Almukhtar 2016
185
Chapter Three
Results
Figure 56: Accuracy pf the prediction at the left paranasal region. Illustrates the median, the
three quartiles and the whiskers of the errors associated with the prediction of soft tissue
change following Le Fort I maxillary advancement at the Paranasal Left region in the three
dimensions.
Anas Almukhtar 2016
186
Chapter Three
Anas Almukhtar 2016
A
B
C
Figure 57: Accuracy of the prediction at the left paranasal region. Colour map showing the mean (directional) error magnitude
between the predicted Left Paranasal region and the real post-operative changes in the three dimensions; X (Medio-lateral) dimension
(A); Y (vertical) dimension and Z( antero-posterior) dimension.
Results
187
Chapter Three
3.4.6 Paranasal Right
Results
The results of the right paranasal region prediction are shown in figures (58 and
59). The difference between the predicted and the actual post-operative
changes were individually analysed in the three dimensions.
The z dimension (the depth) showed the lowest mean and standard deviation
among other dimensions (0.019 mm ± 0.69 mm) followed by the x dimension
(medio-lateral) and the y dimension (vertical dimension) with a mean value of (0.05 mm ± 0.99 mm) and (-0.02 mm ± 1.1832 mm) respectively. In general, the
total mean ± standard deviation in all dimensions was lower than 2.2 mm.
Figure (58) shows the difference in the three dimensions (x, y, and z) Box plot.
The results showed that 50% of the trials showed less than 1 mm of error in
prediction. The size of the boxes showed a relatively low variation among the
trials which was comparable in all dimensions. All of the boxed were relatively
equally distributed around the zero point. The range of the data represented by
the whiskers showed a higher level of variations in y dimension and a lower
variation in the z dimension.
Figure (59) shows the colour map of the differences between the predicted and
the actual post-operative soft tissue changes. There was a general tendency
toward an underestimation (under prediction) of the changes in the y and z
dimensions but over estimation in x dimension. However, a mixed blue and
yellow colour patches were noticed over the right paranasal region close to the
corner of the mouth as displayed on the associated colour scale.
Anas Almukhtar 2016
188
Chapter Three
Results
Figure 58: Accuracy of the prediction at the right paranasal region. Illustrates the median,
the three quartiles and the whiskers of the errors associated with the prediction of soft
tissue change following Le Fort I maxillary advancement at the Paranasal Right region in
the three dimensions.
Anas Almukhtar 2016
189
Chapter Three
Anas Almukhtar 2016
A
B
C
Figure 59: Accuracy of prediction at the right paranasal region. Colour map showing the mean (directional) error magnitude between the
predicted Right Paranasal and the real post-operative changes in the three dimensions; X (Medio-lateral) dimension (A); Y (vertical)
dimension and Z( antero-posterior) dimension.
Results
190
Chapter Three
Results
Summary
The differences between the predicted and the actual results were
minimal. Relatively large variation in the prediction were observed at the
vertical dimension represented by a higher than average standard
deviations in addition to wider range (whiskers) in most of the compared
boxplots. There is no specific pattern of under and over estimation of the
predicted
results
as
the
means
were
divided
equally
between
underestimation and overestimation of the changes distributed among
different regions and dimensions.
Anas Almukhtar 2016
191
4
Discussion
C
ontents
4.1
STUDY SAMPLE................................................................................................................ 193
4.2
DICOM IMAGE PROCESSING AND ANALYSIS ........................................................................... 197
4.2.1
IMAGE SUPERIMPOSITION ....................................................................................................... 197
4.3
MEASUREMENT OF SKELETAL DISPLACEMENT ......................................................................... 198
4.4
MEASUREMENT OF SOFT TISSUE CHANGES IN RESPONSE TO ORTHOGNATHIC SURGERY. ...................... 201
4.4.1
THE GENERATION OF THE “AVERAGE FACE” ................................................................................ 203
4.4.2
THE CORRESPONDING 3D FACIAL SOFT TISSUE CHANGES IN RESPONSE TO LE FORT I MAXILLARY
ADVANCEMENT. ................................................................................................................................ 205
4.4.3
THE CORRESPONDING 3D FACIAL SOFT TISSUE CHANGES IN RESPONSE TO BSSO MANDIBULAR
ADVANCEMENT. ................................................................................................................................ 210
4.4.4
4.5
THE CORRESPONDING 3D FACIAL SOFT TISSUE CHANGES IN RESPONSE TO BIMAXILLARY ADVANCEMENT. 213
SOFT TISSUE PREDICTION ................................................................................................... 218
4.5.1
MASS SPRING MODEL (MSM) ................................................................................................. 218
4.5.2
FINITE ELEMENT MODEL ( FEM) ............................................................................................... 219
4.5.3
MASS TENSOR MODEL (MTM) ................................................................................................ 221
4.5.4
MEASUREMENT OF PREDICTION ACCURACY ................................................................................ 224
Anas Almukhtar 2016
192
Chapter Four
4.1 Study Sample
Discussion
The analysis of pre- and post-operative records to investigate surgical changes
following
orthognathic
surgery
has
been
previously
reported
(107).
Superimposition of the pre-operative and post-operative images provide valuable
information to quantify the magnitude of surgical movements and the overlying
soft tissue response (247). Three main reasons justified the choice of CBCT
imaging as the main source of data for this study. Firstly, the 3D surface analysis
applied in this research necessitates 3D image of the skeletal and soft tissue
facial structures. 3D data cannot be obtained from a lateral cephalogram alone
and therefore the choice was between CT and CBCT. Cone beam computerized
tomography (CBCT) images provide three-dimensional volumetric data with a
lower exposure to radiation compared to conventional CT but more than a plain
film (248,44,45) more details were provided in the literature review chapter.
CBCT facial scans were validated for their accuracy in capturing facial images
(248–253). Farman et al., 2006 reported that the soft tissue definition of the
CBCT was sufficient to determine air/soft boundaries (252). Moerenhout et al.,
2009 used mannequin head to determine the 3D surface accuracy of soft tissue
acquired from a CBCT scan. They found that the 3D surface of the facial soft
tissues segmented from a CBCT scan were accurate and comparable to laser
surface scan (253). Kau et al., 2005 stated that “The CBCT is excellent for
imaging hard tissues structures and most soft tissue components” (248).
Secondly, the aim of this research was to investigate the surface changes of the
soft tissue following orthognathic surgery. A CBCT scan captures the hard and
soft tissues simultaneously and preserves the relation between them. Finally,
the availability of the pre-surgical and post-surgical CBCT scans routinely
captured for orthognathic patients.
Ethical approval was obtained from the West of Scotland Research Ethics Service
on 24th May 2012 (Rec Reference 12/WS/0133). The approval limited the study to
a retrospective analysis of the pre-existing data, hence there was limited control
with respect to the technical parameters of the CBCT scans, as the radiography
technicians standardised the method of CBCT scans for orthognathic surgery
patients (Section 2.1.2). Ethical approval allowed the analysis of other types of
images such as facial stereophotogrammetry, to validate the methodology.
Anas Almukhtar 2016
193
Chapter Four
Discussion
A total of 137 pre-surgical and post-surgical CBCT scans were successfully
retrieved from the database in the radiography department of the Glasgow
Dental Hospital; 37 cases were excluded from the final sample as they failed to
meet the inclusion criteria. Image quality and ethnic background were the two
main reasons for exclusion. The remaining 100 cases were successfully processed
according to the research protocol mentioned earlier, (Section 2.4.1).
The sample size was based on previous studies which investigated the accuracy
of orthognathic surgical simulation and prediction planning (table 25). The range
of the sample size of previous studies varied from 1 to 100 patients. The cases in
most of the studies were mixtures of maxillary and mandibular surgical
procedures.
Table 25: Sample size reported in previous studies.
Author
Year
Type of simulation
Sample
size
and 10
Marchetti et. al., (170)
2011
Marchetti et. al., (255)
2006
Tucker et. al., (256)
2010
Rihan Ullah (236)
2012
Mathematical
modelling
numerical simulation
Mathematical
modelling
numerical simulation
Mathematical
modelling
numerical simulation
Mass spring model
Shafi et. al., (158)
2013
Mass tensor model
13
Lieberegts et al., (312)
2015
Mass tensor model
100
Beldie et. al., (257)
2010
Finite element analysis
1
Meehan et. al., (234)
2003
mass spring model
1
Deuflhard et. al.,(258)
2006
Mathematical
modelling
numerical simulation
and 25
and 20
13
and 4
The need for homogeneous samples when studying soft- and hard-tissue
behaviour following orthognathic surgery is essential. Soft tissue behaviour
following orthognathic surgery proved to be variable in the various regions of the
face (223,254). Therefore the study sample was sub-grouped according to the
type of surgical intervention, i.e. cases which had maxillary surgery only (52
patients), cases which had mandibular surgery only (20 patients) and cases which
had bimaxillary surgery (28 patients).
Anas Almukhtar 2016
194
Chapter Four
Discussion
Several factors can be attributed to the predominance of maxillary surgery in
the current sample, as compared to other studies conducted in the UK(259).
These include the prevalence of Class III deformities in the West of Scotland
which is greater than other regions in the United Kingdom (260); In addition, the
prevalent attitude is to postpone the treatment of mild to moderate skeletal
Class III jaw deformities until the completion of growth for a number of reasons,
including
the
limited
application
of
mid-face
protraction
and
growth
modification (261,262).
Variation in facial soft tissue characteristics among different ethnic groups has
been documented (167,188,215,221). Soft tissue thickness, consistency and
facial features were the most common ethnic variations that may affect
orthognathic surgery planning (215,221). To establish a reasonable homogeneity,
the study sample was confined to Caucasian subjects.
Patients who had a history of previous operations in the orofacial region were
excluded to avoid the effect of previous facial scaring on soft tissue response to
orthognathic surgery (185,220); this includes cleft lip and palate, facial trauma
and surgical treatment of facial pathologies.
No age restrictions were applied; however, the lowest age included was 17 years
and 8 months. This excluded any significant growth related facial changes during
the treatment period.
The timing for CBCT scan capture was a purely clinical decision. However,
patients who had their scans captured within a month prior to surgery (preoperative) and between 6-12 months after surgery (post-operative) were
included in the study. This pre-operative time limit was set to avoid non-surgical
soft tissue changes that might occur preoperatively. The post-operative scan
limits were implemented to avoid the short-term effect of maxillofacial surgery
on
postoperative
swelling
(oedema
from
retraction,
irritations,
and
inflammation), which has been well documented (186). It has been reported that
facial morphology recovers to approximately 90% within 3 months after
bimaxillary surgery (263). Dolce et al., 2003 (264) showed that the swelling
caused by BSSO for mandibular advancement began to resolve by 8 weeks and
Anas Almukhtar 2016
195
Chapter Four
Discussion
was fully resolved by 6 months. Although the post-surgical relapse rate is
multifactorial (including the technique and extent of the performed surgery),
most studies evaluating surgical relapse suggested one year as a cut-off point for
the collection of follow-up records. It was suggested that the majority of
surgical relapse was evident during the first few months after surgery, following
bone fragment consolidation, and the relapse rate was fairly minimal after one
year (265–269). Therefore, in this study, the post-operative CBCT scans included
the combined effect of orthognathic surgery and early relapse without affecting
the robustness of the methodology. The exact bony movements were extracted
by superimposing post-operative radiographs on the pre-operative CBCT scans
and the changes of the soft tissue where related to the calculated skeletal
movement.
Despite the noticeable technological advancement, CBCT scan is still a time
consuming procedure. A full head 22 cm extended field of view (EFOV) scan
takes from 17.5 to 36 seconds (248,270). The (i-Cat) CBCT scanner at the
Glasgow Dental Hospital takes 20 seconds for facial capture with the radiation
gantry rotating twice around the head. Patients were asked to remain still
throughout the procedure, but minor movements during the scan are sometimes
unavoidable. Depending on the extent of these movements, the distortion of the
captured image may range from blurred image to a clear stepwise regional shift
in the image. The retrospective nature of the research sample and the policy of
the Glasgow Dental Hospital regarding CBCT scans made it impossible to repeat
the scan for these cases. Errors associated with image capture were the main
source of exclusion of cases in this study.
The surgical technique was another important exclusion criterion. The majority
of the cases had Le Fort I maxillary osteotomy, bilateral sagittal split mandibular
osteotomy or a combination of both. However, few cases were treated with
different surgical techniques including total body otectomy, vertical sub sigmoid
osteotomy and multi-piece segmental Le Fort I maxillary osteotomy. These few
cases were excluded to preserve the homogeneity of the study sample.
Anas Almukhtar 2016
196
Chapter Four
4.2 DICOM image processing and analysis
Discussion
4.2.1 Image superimposition
Image superimposition on a relatively stable region which was unaffected by the
procedure was the classical approach for assessment of the outcome of
orthognathic surgery. A variety of superimposition techniques were applied
(Section 1.2.2). The choice for voxel based registration in this project was
evidence based. A published study by our group (70) compared two common
types of automated 3D image superimposition techniques (surface based
registration and voxel based registration). The results revealed that there was
no statistically significant difference between the accuracy of the two
techniques in terms of the mean distance of the superimposed meshes. However
the results also showed that the correlation between the soft and hard tissue
models has been lost in the surface based registration group since each model
was superimposed separately. This was not the case with the voxel based
registration where the two tissues (soft and hard tissue) were registered
simultaneously. The conclusions suggested that studies which aim to investigate
the soft to hard tissue correlation should use the voxel based registration
whenever it is applicable.
Around
40%
of
image
superimpositions
were
carried
out
using
the
superimposition module of the Maxilim software package. This module was
offered from the manufacturing company for a limited time to help with part of
this study. The rest of the images were superimposed using another software
package (OnDemand3D), which provided the same quality of 3D image
superimposition. This software have been validated for its image registration
accuracy (67) and hence was used for the rest of the sample in this study.
Only one DICOM image registration was carried out for each patient at the start
of the analyses. Both soft and skeletal tissue analyses were carried out on the
same image superimposition. This eliminated potential error of the repeated
registration process.
Upon successful superimposition, both soft and hard tissues were segmented
from each of the preoperative and postoperative DICOM images using Maxilim
Anas Almukhtar 2016
197
Chapter Four
Discussion
software. This software applies the marching cube algorithm (61) to segment
each tissue boundary according to its associated Hounsfield unit (HU) value. The
optimal HU value for each tissue varied between patients. However it was not
possible to estimate the exact value for each patient; this might have introduced
another variable that could have affected the credibility of the analysis. For that
reason, a standard HU value of (276) for skeletal tissue and (-976) for soft tissue
segmentation were adopted for all cases. These were the default values built
into Maxilim software.
When analysing the superimposed models, the skeletal tissue appears to align to
a clinically acceptable level on the anterior cranial base, which is the region of
registration, whereas the soft tissue models seem to have some inaccuracy on
the forehead. Details of these inaccuracies were previously discussed (Section
2.2.1). This could be explained by the fact that voxel based registration relies on
the grey scale values of the voxels of the selected region of interest and not on
the surface of each tissue. Minor changes in soft tissue related to different facial
expression during scan or weight gain or loss does not affect the registration
since the surface has a limited effect on this variable. This, on the other hand,
provided an accurate description of the real change and preserved the relation
between the hard and soft tissue models rather than losing this relation to
produce a better soft tissue superimposition which was achieved by surface
based registration.
To address the aim of the project, that of finding the correlation between soft
and hard tissue changes, three distinct steps of the analysis had to be followed;
firstly to analyse the skeletal displacement; secondly to analyse soft tissue
deformation and finally to find the correlation between the two changes.
4.3 Measurement of Skeletal Displacement
The measurement implemented to assess skeletal displacement was a novel
approach developed and validated within this PhD project. The method was
simple, reproducible and yet highly accurate in analysing the skeletal
displacement in three dimensions of space. Previous methods to analyse skeletal
displacement were less successful in achieving these goals without a higher
magnitude of errors (75,100,210). Despite the problems with 2D lateral
Anas Almukhtar 2016
198
Chapter Four
Discussion
cephalometric analysis with regard to its landmarking errors, with lack of
recording of bilateral structures and with magnification issues (75,100,271–273),
it is still considered by other researchers (274,275). The reason for this might
have been the unavailability of the 3D CBCT scans. This project benefited from
the availability of the CBCT scans in the data base to carry out the 3D analysis
presented in this study.
A variety of 3D image analyses have been considered for the evaluation of soft
and hard tissue changes following surgery. The most commonly used are the
anatomical landmarking (3D cephalometry), Euclidean surface distance (mean
distance on a surface patch or distance at specific points), and, less frequently
volumetric analysis (110). Nevertheless, each of these methods has its
drawbacks which affect its credibility on certain aspects. Details of shortcomings
of each method have been explained earlier in the literature review.
The method applied in this study was based on creating an anatomically based
local 3D axis (three orthogonal planes) to track the differential change of the
jaw bones at different areas where the linear (distances) and the angular (pitch,
role, and yaw) displacements could be measured. Each plain required at least
three points to be established. The challenge was in finding valid and
reproducible landmarks on the smooth bony surface of the jaw bones, especially
the maxilla, to establish this plane. The three foramina on the hard palate
surface (incisive foramen and bilateral greater palatine foramina) and five
anatomical locations at the mandible (bilateral lingual foramina, bilateral
mental foramina and lingual tubercle) were used. Although their boundaries
were clearly visible on the DICOM slices, the relatively thin bony plate
surrounding the foramina were not accurately segmented into the 3D model and
in the majority of the cases there were noticeable differences in size and shape
between the pre and post-operative images especially around the greater
palatine foramina, (figure 60). This was in addition to the fact that digitizing
landmarks on DICOM image slices were more accurate and reproducible than
identifying them on the three dimensional model surface (96). Maxilim software
provides the ability for DICOM slices landmarking; therefore, the decision was
taken to place the landmarks on the shadow of the foramen boundary on the
DICOM slices with a distinctive anatomical definition of each landmark.
Anas Almukhtar 2016
199
Chapter Four
Discussion
Validation of the protocol of this method has been established by our research
group (246)
Figure 60: Difference between the 3D model and the radiographic shadow in representing
the contour of the greater palatine foramina.
Only the linear measurements were considered in this project. Angular
measurements were excluded from the study. The newly developed PCA
simulation algorithm has not reached the level to allow incorporation of the
rotational elements (pitch, role and yaw) and yet these were thought to be a
source of noise to the statistical soft tissue analysis and it was believed to have
a limited effect on the overlying soft tissue behaviour compared to the linear
measurements. Future studies may enhance the overall result of the prediction
planning by the addition of the rotational element.
The measurements in this study were established as the orthogonal distance
from the selected landmarks to three orthogonal reference planes. A detailed
Anas Almukhtar 2016
200
Chapter Four
Discussion
description of the measurements was previously provided (Section 2.2.2). The
orthogonal distance from the same reference planes to each landmark on the
pre- and post-operative images were measured following voxel based
registration. This eliminates errors that could have developed as a result of the
establishment of reference planes for the pre- and post-operative images
separately.
4.4 Measurement of soft tissue changes in response to
orthognathic surgery.
Soft tissue analysis was based on the concept of dense correspondence between
the pre- and the post-operative images. Previous studies reported other methods
for soft tissue analysis, including anatomical landmarking (cephalometric
analysis), colour coded surface distance and volumetric analysis (110,147,276).
Each of these methods has its own shortcomings and could not be used to
address the aim of our research (details were provided in Section 1.3.2).
The application of the corresponding surface analysis is relatively new to the
field of orthognathic surgery. Mao et al., 2006 (244)published their work on a
similar concept which they named "Anatomical dense correspondence analysis"
and showed that it could be applied to study human faces. Since then, limited
studies have been carried out on this method. Claes et al., 2011-2012 and
Walters et al., 2013 (152,175,277) described the use of an "the anthropometric
facial mask ", which followed the same concept. The aim of their studies was to
describe the soft tissue changes due to growth and after correction of
asymmetric faces by orthognathic surgery. However, their analysis did not
progress further and the two studies were concluded as a visual description of
the vectors connecting the corresponding vertices on the pre and post-operative
facial meshes or at two time intervals. A number of reasons might have
contributed to the limited number of publications including the complicated
procedure associated with generic mesh construction and conformation; limited
software packages providing this type of analysis (usually coded specifically for
this purpose within the research group) and lastly the lack of evidence to
support any superiority of the method.
Anas Almukhtar 2016
201
Chapter Four
Discussion
The advantage of this method over other surface or landmarks based analyses, is
in the preservation of the anatomical correspondence while a comprehensive
facial surface analysis is carried out. The technique however, is sensitive and a
high level of accuracy is required in all the stages of the analysis, including
image
capture,
generic
mesh
construction,
landmark
digitization
and
conformation procedure.
In this project, an investigation was carried out to ensure the highest possible
accuracy of all stages of the analysis. The results showed an acceptable level of
accuracy and formed a solid base for a comprehensive analysis, (Section 2.3.1).
Although the magnitude of surgical displacement was variable among the cases,
the surgical technique and the direction of movement were standardised to
reduce the variability within each group. This variation might have a negative
effect on the analysis. However, the clear correlation between the results of the
analysis and the clinical observation support its validity. A larger sample size is
recommended for future studies to reduce this margin of error.
Dense correspondence was the selected method of the analysis in this project
for two main reasons: Firstly, it provided surface information data hence it
overcame the shortcomings associated with the landmarks based analysis.
Secondly, it provided an anatomical correspondence between the pre- and postoperative facial meshes, thus, it avoids the limitations associated with the
traditional closest distance and coloured coded map. The availability of the
database and the in house developed software packages allowed the application
and validation of the method.
In order to apply the concept of dense correspondence, a facial generic mesh
conformation was used. The original generic mesh was previously generated by
our research group for a similar purpose. VRMesh software was employed to
reduce the number of surface mesh triangles to 1000, closing mesh holes and
removing duplicate triangles. The resultant generic mesh has a uniformly
distributed approximately 3mm apart nodes with around 1000 triangles.
The process of conformation (elastic deformation) of the generic mesh to the
pre- and post-operative soft tissue models was carried out using in house
Anas Almukhtar 2016
202
Chapter Four
Discussion
developed software (Section 2.3.1). Validation study was carried out to confirm
the software package accuracy and to test the accuracy of the conformation
process. Errors, including mesh sliding and poor elastic deformation, were
examined. The results showed high accuracy of the conformation with a mean
error of 1.12±0.23 mm.
The procedure of conformation was previously described, following the
placement of 20 anatomical landmarks, the conformation performed in two
steps: firstly was the “spline” step which is the application of the ‘thin plate
spline’ algorithm; secondly was the “conformation” which is a localised
automated elastic deformation algorithm. The first step relied on 20 landmarks
manually placed on both the generic mesh and the target mesh (pre- or postoperative images). The CBCT scan produces one of the most challenging
landmarking placement soft tissue images due to the smooth and texture-less
surface (248). Landmarking error study was carried out and results showed a
reasonable level of accuracy for clinical application and soft tissue analysis.
Although the landmarking error was below the clinical significance level, two
additional measures were carried out to reduce the range of errors: Firstly the
generic mesh was landmarked only once for the purpose of conformation on all
cases; This reduced the landmarking errors to 50% by excluding the potential
errors associated with repeated generic mesh landmarking for each case.
Secondly, the pre- and post-operative meshes were landmarked simultaneously
on a multi view window on the same screen which facilitated and improved the
accuracy of landmark identification. Digitizing the landmarks on the exact
position on both images was the key for successful facial dense correspondence
analysis.
4.4.1 The generation of the “average face”
The use of the “average face” to study human facial features has been
previously reported (162). This method provided a possible solution to process a
large data set of facial meshes by averaging facial shapes into a single mesh. A
single visual outcome of the facial analysis was achieved for the entire sample;
this could not have been possible through individual case analysis. However, the
accuracy of this type of analysis is highly dependent on the magnitude of case
variations and the sample size to generate the “average face”. In this study, the
Anas Almukhtar 2016
203
Chapter Four
Discussion
variations were reduced through strict case selection criteria, excluding cases
with additional genioplasty procedures and subdividing the study sample into
five groups (Le Fort I advancement, BSSO advancement, BSSO setback,
bimaxillary
(Maxillary
advancement-Mandibular
setback)
and
bimaxillary
advancement. Despite the large number of collected sample (100 cases), only
three of the five groups achieved a sufficient sample size to be considered in the
analysis. These were Le Fort I advancement (33 cases); BSSO advancement (12
cases) and Bimaxillary advancement (12 cases).
The method applied for facial averaging was relatively new to the field of facial
analysis. Limited work has been published on this topic, (Hutton et al., 2003 and
Hammond et al., 2004). Both studies applied the average facial meshes to study
facial morphology. Their methods were based on a similar concept of
establishing a closest point dense correspondence between the two averaged
meshes. The selection of the initial facial mesh was critical in their study, as it
has a direct effect on the generation of an average mesh which was the result of
subsequent superimposition and averaging of all individual meshes of the study
sample. The starting mesh (template), therefore, should be selected at the
middle of the shape variation across the sample. This was achieved by
landmarking the entire study sample with a minimal set of facial points,
followed by the application of full Procrustes analysis. Principal component
analysis was then used to extract the facial averaging template (initial mesh).
The closest point correspondence has been used by many researchers to assess
facial changes and was found to be more accurate than other methods (177),
however, it still carries a considerable amount of error, especially at the
peripheral regions of the mesh.
The method of facial averaging used in our project is more accurate, and yet
considerably less complicated. A detailed description has been previously
provided (Section 2.4.1.1.7). The averaging procedure was based on using the
conformed generic meshes utilizing the common 3D index of the generic mesh
vertices to establish the dense correspondence among the facial meshes of the
whole study sample. Facial average was generated by applying partial Procrustes
analysis (PPA) on the entire sample. Each vertex was considered as a landmark
and participated in the (translation and rotation) procedure of the PPA based on
Anas Almukhtar 2016
204
Chapter Four
Discussion
the generic index. The mean position of each vertex across the full sample was
considered as the average position of all vertices that carries the same index.
The resultant face was an average facial mesh at the centre of the facial
variations across the study sample.
This method appeared to be more reliable than the previously published data
(278). Increasing the study sample would be highly advantageous for future
studies.
4.4.2 The corresponding 3D facial soft tissue changes in
response to Le Fort I maxillary advancement.
4.4.2.1 Euclidean distance measurements
The average pre-operative and post-operative faces were superimposed on the
eye and nasal bridge regions. The forehead region was not used in this analysis
because it was deficient in some of the averaged facial meshes which affected
the position of the forehead in the average face. Figure (42) showed good
accuracy of superimposition by expressing a predominant green colour on the
region of the superimposition and the surrounding areas. Since the average face
was constructed by averaging conformed facial meshes which belong to the same
generic mesh, the vertices indexing were still valid and the two meshes had an
identical number of triangles. The colour assignment was generated by crossmatching the vertices indices of the two images creating a corresponding
distance colour map. Some of the areas on the mesh surface were assigned with
the colour red which indicated a significant movement, even though they were
extremely close to the corresponding mesh surface. This was due to the
detection of mesh sliding movement which was completely overlooked in the
classical colour surface analysis. Figure (61) shows the difference between the
two analyses, the closest point correspondence and the anatomical dense
correspondence, of cases which had had a Le Fort I advancement osteotomy. It
is clear that a major area around the cheeks, nose and chin were overlooked or
misinterpreted when the closest surface distance, (figure 61 A) was applied.
Figure (61 B) shows the corresponding distance which reveals an area of
additional changes involving chin, cheeks and nose region. However the classical
inner and outer surface colour scale assignment (inner blue, outer red) was still
considered. Using the Euclidean distances obscured the directionality of the
Anas Almukhtar 2016
205
Chapter Four
Discussion
change, the red colour at the chin may be due to an upward, downward, right or
left movement of soft tissue following surgery. To solve this problem, a separate
analysis of each dimension was developed.
A
B
Figure 61: Comparison between classical colour map A, and the corresponding analysis B.
4.4.2.2 Directional changes measurements
Surface distance, in a form of a colour map, is commonly applied for a
descriptive analysis of surface changes (276), however, it lacks the directionality
of the change. The classic landmark based analysis provided some information
regarding the direction of the displacement at selected regions (17–20). Verdenik
et al., 2014 modified this approach to overcome this issue, four horizontal and
two vertical planes were established anatomically to segment the facial surface
into ten regions. The method provided additional regional information without
the need for measuring the change at certain landmark. However, the
segmentation procedure was not representative of the anatomical regions,
especially in the paranasal region. It was also assumed that most of the changes
were in the A-P direction. The authors commented that the changes around the
lower lip and chin areas showed minimal change. However, they also noted
significant change in the submandibular region in an upward direction. This
contradiction was due to the fact that the method used for the analysis (closest
surface distance) did not detect the upward sliding of the facial soft tissue mesh
that occurred at the lower lip and chin area, unlike the submandibular region
Anas Almukhtar 2016
206
Chapter Four
Discussion
where the upward surgical movement set the two surfaces apart, therefore, the
changes were detected only in the submandibular region. With our approach,
anatomical correspondence (dense correspondence analysis) has the ability to
detect mesh sliding and to produce an accurate analysis of the changes at this
region.
In this study, the application of the dense correspondence analysis combined the
advantage of anatomical landmarking and the comprehensive surface analysis
which is desirable and innovative.
The results presented by the corresponding colour map in the x (medio-lateral)
dimension were novel, (figure 43). Most of previous studies focused on the A-P
changes of facial morphology at the mid-facial region following Le Fort I
osteotomy (147,187,201,281–284); other directional changes were largely
overlooked. In addition to the well-defined forward displacement of the mid
face soft tissue, the four striking features observed in this analysis were the
lateral stretching of the upper lip and paranasal areas; the limited effect of
surgery on the lower lip vermilion and oral commeasures; the vertical and
forward chin movement and lastly the shortening of the nostril height associated
with flaring of the alar cartilages.
The detected changes In the x (medio-lateral) dimension (figure 43 A) on the
nose including widening of the nostrils, were in agreement with the vast
majority of previous published studies (147,161,259,285). However, this effect
was evident at the alar curvature more than the alar base. This may be due to
the fact that the cases in this study had a nasal cinching stitch which might have
reduced the alar base width. This observation was in agreement with Metzler et
al., 2014 (282). However, their measurements were based on linear distances
between anatomical landmarks whereas in this study the total alar base region
and alar curvature surfaces were analysed more comprehensively.
Keep in mind that Le Fort I maxillary osteotomy is performed to address the
underlying maxillary hypoplasia with its characteristic soft tissue facial
configuration. One of these soft tissue characteristics is the narrow alar base.
Therefore, it is not unusual to allow some widening of the alar base to occur
following surgery to restore normal facial appearance. It is difficult, however, to
Anas Almukhtar 2016
207
Chapter Four
Discussion
report on the effect of cinching stitch on the width of the nostrils and whether
this effect was intended.
The main detected effect of the Le Fort I osteotomy on the mid face region was
the lateral widening of the upper lip and the paranasal areas. To date, no
previous studies have reported on similar findings. Metzler et al., 2014(282)
reported on an increase in the upper lip philtrum width following Le Fort I
osteotomy. On the other hand they showed a limited change of the upper labial
width. This was not in agreement with our study. This disagreement was due to
the fact that the labial width was measured as the linear distance between the
two oral commeasures. We demonstrated that the expansion of the upper lip
was evident at a higher level outlining the oral commissures as anatomical
boundaries of that expansion. van Loon et al., 2015 (286) reported an increase in
the upper lip volume following Le Fort I advancement. Our conclusion was that
the widening of the upper lip and paranasal regions were a result of an increase
in fullness in these regions secondary to surgery and the changes were limited to
their anatomical boundaries.
Vertical skeletal maxillary displacement was minimal in the study sample. Soft
tissue vertical displacement was only observed at the nostrils and chin (figure 43
B). No marked vertical changes were observed at the upper lip region (0mm0.5mm). This was combined with a relatively higher upward displacement
observed at the naso-labial junction and sub-alar area. This led to the conclusion
of an increase in lip length following Le Fort I advancement surgery. This result
was in agreement with previous studies(281). The lower lip on the other hand
showed a generalised upward displacement in response to the Le Fort I
osteotomy. However, the lower lip vermilion border was relatively stable within
(0mm-0.5mm) of the pre-operative position; this observation led to the
assumption that the lower lip had shortened vertically secondary to Le Fort I
osteotomy. At the chin area, major upward displacement had occurred which
was relatively confined to the chin region and faded away laterally. Although the
mean vertical maxillary displacement was minimal, the changes were observed
at the chin secondary to the mandibular autorotation as a result of posterior
maxillary impaction.
Anas Almukhtar 2016
208
Chapter Four
Discussion
The red colour at the nostrils and nasal tip suggested an upward movement of
this region secondary to Le Fort I advancement osteotomy. This result was in
agreement with previous studies (281,285).
In the antero-posterior z dimension (figure 43 C), there was an obvious forward
movement of the mid face region at the upper lip, paranasal areas and nostrils
with a less extent at the chin region. These results were predictable with
advancement of the maxilla at Le Fort I level and were in agreement with the
previous studies (161,259).
Interestingly, the tip of the nose did not show significant change in the A-P
direction; this was associated with a significant forward displacement of the alar
cartilages and columella of the nose. The combined effect in x, y and z
dimension could be described as compression of the nostrils in the A-P dimension
which were expressed as shortening of the columella and widening of the alar
cartilage curvature. This result was in partial agreement with previous studies
which suggested shortening of the nostrils and increase of the alar
width(281,287). However, the majority of these studies reported a significant
nasal tip protrusion following Le Fort I osteotomy(281,285,287). The reason
behind this disagreement might have been the minimal maxillary impaction in
our study, the removal of the anterior nasal spine and the reduction in the
inferior borders of the pyriform aperture during Le Fort I osteotomy.
The relationship between the soft tissue surface changes and the underlying
anatomical structures was clear. The orientation of the muscles and their
attachments played a marked role in the expression of the overlying soft tissue
changes, (figure 62). The bilateral group of muscles of facial expression which
originated
from
the
side
of
the
bridge
of
the
nose
including
the
levator labii superioris alaeque nasi and from the zygomatic bone including
zygomatico major, zygomatico minor and levator labi superioris muscles which
are inserted in the facia of the upper lip and the muscle fibre of orbicularis oris
muscle. This group of muscles were stretched by the advancement of the
maxilla, which contributed to the augmentation of the relatively depressed area
at the paranasal region. These effects were limited superiorly by the origin of
these muscles and laterally by the anterior superficial fibres of masseter muscle
which were not affected by the surgery. This was in agreement with previous
Anas Almukhtar 2016
209
Chapter Four
Discussion
studies which proposed similar boundaries of the soft tissue changes at these
regions of the face(279).
Figure 62: Anatomy of mid-face muscle) quoted from Grant's atlas of anatomy)
4.4.3 The corresponding 3D facial soft tissue changes in
response to BSSO mandibular advancement.
4.4.3.1 Euclidean distance measurements
Figure (63) shows the difference between two methods of analyses of the BSSO
mandibular advancement group, the closest point correspondence and the
anatomical dense correspondence analysis. It is clear that changes around the
cheeks, nose and chin were completely overlooked or underestimated when the
closest surface distance, figure (63-A), was applied. Figure (63-B) showed the
corresponding distance which revealed more extensive but homogenous pattern
Anas Almukhtar 2016
210
Chapter Four
Discussion
of changes involving the chin, lower lip and extended back to the cheeks and
gonial angle regions. Colour coded analysis obscures the directionality of the
change. This means the red noted colour at the chin may be an upward,
downward, right or left direction. To solve this problem, a separate analysis of
each dimension was again used.
A
B
Figure 63: Comparison between classical colour map A, and the corresponding analysis B.
4.4.3.2 Directional changes measurements
The current study attempts to clarify the three dimensional changes of the soft
tissue following mandibular advancement surgery in a comprehensive and
anatomically guided surface analysis. Every effort has been made to standardise
the multi stage method of assessment. The small sample size (12 cases) was due
to the strict exclusion criteria. BSSO mandibular advancement with minimal
lateral shift surgery with rigid fixation was selected and cases with adjunctive
genioplasty were excluded.
The colour map was stratified to reveal the directional changes in each of the
three dimensions. This helped in providing a comprehensive 3D description of
the soft tissue changes following mandibular advancement. The results of this
study clarified the areas of controversy and introduced further details of the soft
tissue changes following BSSO mandibular advancement surgery.
Anas Almukhtar 2016
211
Chapter Four
Discussion
Facial soft tissue changes following BSSO mandibular advancement has been
investigated previously either by profile assessment using 2D cephalometric
analysis (288–290) or by the 3D soft tissue analysis (159,288).
Apart from the frequently reported direct forward movement of the chin soft
tissue, the lower lip showed less forward movement, this displayed as a lower
intensity of the colour red when compared to the chin region, while the upper
lip showed a backward movement marked by a light blue colour, (figure 45). This
was in agreement with most of the published studies, (223). In addition, a wider
lower lip and narrower upper lip were also demonstrated, (figure 45), as a result
of mandibular advancement.
Mobarak et al., 2001 (222) showed that the lip thickness decreased as the
mandible moved forward. This could be logically clarified by the tendency of the
soft tissue, on both sides of the lip, to move laterally, stretching the lip and
reducing its thickness, (figure 45). Stretching the lower lip laterally gave the
mandible a broader aspect, which was in agreement with previous studies(291).
Reduction of lower lip eversion following mandibular advancement was reported
(289,291). This in agreement of our findings and was clearly demonstrated by
the minimal change in both the vertical and A-P position of the vermilion part of
the lower lip when compared to the marked forward and downward
displacement of the cutaneous part (figure 45).
Minimal vertical change in the vermilion part, combined with a marked
downward displacement of the cutaneous part of the lip was observed (figure
45). This was in agreement with Raschke et al., 2013 (291), who showed a
marked Lengthening of the lower lip.
In general, the change of the facial soft tissue in response to mandibular
advancement surgery was confined to the anatomical borders of the underlying
skeletal structures. Muscle orientation and attachments played a marked role in
controlling the expression of the overlying soft tissue changes as shown in figure
(64). The firm control of the supra labial muscle group kept the changes at the
commissures of the mouth to a minimum and the sharp colour change on the
cheeks in both A-P and vertical dimensions appeared to be limited superiorly to
Anas Almukhtar 2016
212
Chapter Four
the position of the teeth. Stretching of the platysma
Discussion
muscle in the A-P
dimension, in addition to the change in the orientation of the masseter muscle
following mandibular advancement, (292) was clearly responsible for the welldefined expression of the displacement of the overlying soft tissue especially at
the cheeks regions.
Figure 64: Anatomy of mandibular muscles (quoted from Grant’s atlas of anatomy)
4.4.4 The corresponding 3D facial soft tissue changes in
response to bimaxillary advancement.
4.4.4.1 Euclidean distance measurements
Figure (65) shows the difference between the two analyses, the closest point
correspondence and the anatomical dense correspondence analysis, of the
bimaxillary advancement group. It is clear that major areas around the cheeks,
Anas Almukhtar 2016
213
Chapter Four
nose,
Discussion
mandibular
borders
and
chin
were
completely
overlooked
or
misinterpreted when the closest surface distance, (figure 65,A), was applied.
Figure (65-B) shows the corresponding distance analysis which revealed a more
extensive displacement which extended back to the cheeks and gonial angle
regions. A separate analysis of each of the three dimensions (x, y and z) was
carried out to investigate the changes in detail.
A
B
Figure 65: Comparison between classical colour map A, and the corresponding analysis B.
4.4.4.2 Directional changes measurements
The method used in this study for the analysis of soft tissue changes is unique,
which enabled a comprehensive 3D description of the soft tissue changes
following bi-maxillary advancement surgery. Limited publications were found
focusing on soft tissue changes following bimaxillary advancement surgery; this
might be due to the fact that the surgical procedure was applied mainly for
treatment of obstructive sleep apnoea cases with most of the publication
focussed on its effect on the pharyngeal airspace rather than facial soft tissue
changes. However, bi-maxillary advancement surgery has been used for the
correction of high angle class II cases (274).
Anas Almukhtar 2016
214
Chapter Four
Discussion
The earlier 2D and 3D landmark based analysis helped in identifying the
direction of the displacement at selected anatomical sites on the mid and lower
face regions following maxillo-mandibular advancement surgery (134,274,293)
but the extent of these changes were not examined. Surface distance, in a form
of colour coded map, was commonly applied as a descriptive analysis (276).
However, the lack of anatomical correspondence was its main limitations.
In this study, the application of the anatomical dense correspondence analysis
combined the advantages of the anatomical landmarking and the comprehensive
surface analysis. Since the average pre and post-surgery facial surface meshes
have the same number and index of the vertices, the correspondence between
the two meshes was built on the anatomical bases and the displacement of each
vertex from the pre-operative to the post-operative position was calculated. The
results of the all measurements of all the vertices on the facial mesh were
displayed as a colour coded map to simplify the analysis. A similar method was
previously applied, (152,294), to monitor facial growth and to evaluate facial
asymmetry correction where each point on the facial mesh was tracked to its
displaced position, the displacement was displayed as a coloured line joining the
two positions. However, in their analysis, it was not possible to determine the
direction of change since the classical method of the outward-inward colour map
configuration was applied. In our analysis, soft tissue changes in each the x, y
and z dimensions were analysed separately with a unique colour coding display.
This enabled the current analysis to investigate the changes in a more
comprehensive manner and enabled the analysis of the extent of lateral
expansion of the midface in response to maxillary advancement which was not
previously reported.
The use of the average pre and post-surgical face mesh for soft tissue analysis
was reported (295). Gerbino et al., 2014 (295) reported on the use of average
pre- and post-surgery facial meshes to analyse soft tissue changes following
maxillo-mandibular advancement surgery for treatment of obstructive sleep
apnoea (OSAS). The sample size was close to our study (10 cases) and the results
were in partial agreement. They also reported minimal widening of alar
curvatures. However, their study also showed widening at the commissure of the
mouth which was in disagreement with our study. The reason behind the
disagreement might be due to the fact that their measurements were based on
Anas Almukhtar 2016
215
Chapter Four
Discussion
the linear distance between the right and left chilion points passing through
subnasali and pogonion points (Chil r- Sn- Chil l). Their measurements indicated
an increase in fullness of the upper lip but not necessarily widening of the intercommissural distance. Our findings suggested widening of the upper lip at a
higher level whereas the inter-commissural distance remained constant.
The colour distribution on the lower lip showed narrowing of the sub
commissural regions. The surgical movements involved advancement of the
mandible with an element of vertical movement, especially in open bite cases,
which might have contributed to the noted soft tissue behaviour of the lower lip.
The response of the lower lip to these skeletal movements is complex. Three
factors might have contributed to the noted soft tissue response to surgery: The
vertical element of mandibular and/or maxillary surgical displacement; the
magnitude of maxilla-mandibular advancement and the achievement of lip seal
following bimaxillary advancement surgery.
The extents of the soft tissue changes were confined to the anatomical
boundaries of the affected regions. The medio-lateral changes at the upper lip
and paranasal regions were limited to the muscular attachments around the
osteotomy site where as the effect of the BSSO advancement was localized to
the lower lip and commissures of the mouth, which are the most affected
regions by surgery. It is difficult to separate the effect of the two osteotomies
since the soft tissue acts as a single unit and changes in one region may be
affected by changes in other regions. This contributed to the reason for the
difference in the response of the lower lip noted in Mandibular advancement
cases in the x dimension and not seen with a combined maxilla-mandibular
advancement.
In the A-P dimension, the effect of the bi-maxillary advancement on the midface
extended posteriorly more than what was observed with a Le Fort I maxillary
advancement alone. The reason might be the loss of the posterior boundaries of
this region as a result of the change of the masseter muscle orientation following
BSSO mandibular advancement (292).
The vertical changes were less obvious than those in the other dimensions. This
was in agreement with previous studies (293) Vertically, shortening of the lower
Anas Almukhtar 2016
216
Chapter Four
Discussion
lip was observed as a combined display of a red colour at the cutaneous part of
the lower lip down to point B and yellowish green colour at the vermillion part,
whereas minimal lengthening of the upper lip was observed with green colour on
the upper lip and orange colour in the naso-labial junction and supra
commissures regions. These two observations were in agreement with the
previous studies of Gerbino et al., 2014 (295) and Conley and Boyd 2007 (293).
Anas Almukhtar 2016
217
Chapter Four
4.5 Soft tissue prediction
Discussion
The relationship between hard and soft tissue changes has been extensively
investigated over the past decades using 3D imaging tools (170,235,296,297).
Novel approaches were applied for prediction of soft tissue changes following
orthognathic surgery, which included three main approaches; Mass Spring model
(MSM); Linear/Non Linear Finite Element Model (FEM) and Mass Tensor Models
(MTM). In general, the volume tissue in-between the skin surface and the bone
surface is occupied by mathematical dense tetrahedron prisms or individual
connectors (298,299). Each prism or connector has a fixed side at the bone
surface and a free side at the skin side. Movement of the jaw bones introduces a
stress on the overlying soft tissue volume, which in turn results in an internal
strain. The associated algorithm reacts to this by minimizing the energy through
displacing the soft tissue side to re-establish equilibrium state (227). The
magnitude of skin surface displacement is directly dependent on the function
built into the algorithm that defines the stiffness at each connection site. Some
of these methods were found to be more accurate and more reliable than the
others but there was an agreement on the need for improved soft tissue
prediction following orthognathic surgery.
The approach in this study is unique, with no reliance on the inter-surface
connectors or on the stress strain correlation. Instead, a statistical model was
built on the actual soft tissue surface changes associated with the skeletal tissue
displacement.
4.5.1 Mass spring model (MSM)
Spring mass model (MSM) originated in the computer animation world (Keeve et
al., 1998, Teschner 2001). The application of this method for prediction of facial
soft tissue changes in response to orthognathic surgery was based on quoting
constants from pure engineering contexts of materials addressing stress-strain
properties and applying them for the prediction of soft tissue changes. The
prediction algorithm of this approach relies on a dense mesh of connectors
between each vertex on the soft tissue surface mesh and the nearest point on
the hard tissue surface. These connections were described as springs and each
Anas Almukhtar 2016
218
Chapter Four
Discussion
three springs at the corner of a mesh triangle formed a tetrahedral prism. The
stiffness of these springs is determined by a constant representing the stressstrain (stiffness) value which is only an approximation to the soft tissue stiffness
and has no biomechanical characteristics.
The weakness in this approach is two folds; first is the assumption of the
similarity of the behaviour of soft tissue to that of artificial material, and second
is the assumption of a linear stress-strain relationship. It is well known from
previous literature that biological tissue is anisotropic, inhomogeneous and has a
non-linear stress-strain relationship (300). Human soft tissue is composed of a
multi layered volume of tissues which contains variable percentages of various
tissues types, including muscle fascia and fat, each of these had its own stress
strain curve which is not necessary a linear correlation. There is no way to really
control the volume conservation to resemble biological tissue behaviour during
simulation, even when the model is extended with extra ‘volume spring’ as
described by Molleman et al., 2003 (227).
Keeve et al., 1998 (230) proposed an approach to solving these problems by
introducing a multi layered spring with different stiffness values in addition to a
non-linear stress-strain relationship. The total spring stiffness value is the
summation of its layers. This improves the quality of the prediction of soft tissue
changes due to surgery. However, the choice of the value was still based on
pure physical, non-biological material properties. The results of the experiment
showed lower accuracy compared to the finite element analysis. However, a
considerable gain in computational time was achieved with this method (230).
3dMD vultus software is one of the surgical prediction planning software
packages that used this algorithm.
4.5.2 Finite element model ( FEM)
The analysis of facial soft tissue changes using finite element analysis is based on
a similar concept to MSM in terms of correlating each vertex on the soft tissue
mesh to a single point on the skeletal surface. However, the connections in this
method are treated individually rather than the tetrahedral prisms structure of
Anas Almukhtar 2016
219
Chapter Four
Discussion
the MSM and their stiffness is a function of stress-strain values based on
biological tissue properties. These values were obtained mainly from the online
source “Diffpack” (301). For a linear FEM, the assumption of the linearity of
facial biomechanical behaviour was based on the work of Gladilin et al., 2001
(302) who described the biomechanical behaviour of soft biological tissue as an
isotropic, homogenous and linear elastic continuum. However, it is well known
from previous studies that biological tissue is anisotropic, inhomogeneous and
has a non-linear stress-strain relationship (300). The stiffness values of FEM are
obtained from different tissues, mainly thigh muscles and forehead tissues. The
results showed better prediction accuracy when compared to the MSM approach
(299). However, the assumption of a linear stress-strain correlation was a major
source of deficiency of this method in addition to the inability to separate each
connection into layers as in the MSM approach, rendered it short of an accurate
soft tissue prediction. Moreover, the high computational power and long
computational time (up to 10 minutes) made it difficult to use in a regular
clinical settings (299).
Ulusoy et al., 2010 (303) proposed a solution to the computational demand and
the high processing time named “Dynamic volume spline”. This approach was
based on condensing the analysis into number of selected vertices scattered on
the facial mesh surface then joining them with a continuous spline (303). This
approach reduced the processing time but introduced another source of error in
the areas between the selected points.
In an attempt to improve the prediction accuracy of the soft tissue changes
following orthognathic surgery, Delingette et al., 1998 (304) applied a
biomechanical FEM that has non-linear stress-strain correlation. This was found
to be more accurate in predicting facial changes (304). However the difference
between the two models was non-significant.
In FEA, the tetrahedron prisms of the FEM are not segmented into layers as with
the MSM. To overcome this shortcoming, Delingette et al., 1998 and Chabanas et
al., 2003 (231,304) structured the skull, muscle and skin in the patient’s face
with three to four connected 3D mesh layers to reconstruct patient’s realistic
Anas Almukhtar 2016
220
Chapter Four
Discussion
facial anatomy. The muscles of the face were individually modelled at their
anatomical position with interconnected meshes allowing each mesh to deform
individually under the stress generated by surgical skeletal displacement. This
approach required a higher computational power; it is individualized to each
patient and an operator input is necessary to model each muscle.
Although the FEM in its linear and non-linear types has a higher level of accuracy
compared to the mathematical mass spring model (227,299), the marked higher
computational time was a disadvantage. Molleman et al., 2007 highlighted the
fact that there is always a trade between the accuracy of prediction and the
processing time (227).
VISU simulation model as a novel modification of the classic FEM was introduced
by Sarti et al., 1999 (232). The CT image was segmented in three components;
skin surface, bone surface and the bulk of tissue in-between. The simulation
algorithm applied the finite differences on the acquired CT grid directly and
utilised the bulk of soft tissue volume rather than the bone and soft tissue
surface meshes thereby reducing the computational cost without losing details
of the anatomic structures. The algorithm utilized the modified linear elasticity
equation system(305) based on the in vivo mechanical properties of skin and
muscles described by Black and Hastings 1989, which assumed that soft tissue
responded as a linear, elastic, isotropic material (306).
This algorithm was later incorporated in commercially available software
(SurgiCaseCMF®). In a 10-patient series, Marchetti et al., 2006 (255) reported
80% reliability and reproducibility for VISU. Bianchi et al., 2010 (296) reported
0.94 mm ±0.9mm absolute error with an average 86.8% of the errors remaining
below 2mm.
4.5.3 Mass tensor model (MTM)
This method could be seen as a combination of the MSM and the FEM. On one
hand the model preserves the easy architecture of the MSM and on the other
hand the model has the bio-mechanical relevance of FEM. The original MTM was
Anas Almukhtar 2016
221
Chapter Four
Discussion
introduced by Cotin et al., 1999 (235), modified by Schwartz et al., 2005 and
Molleman et al., 2007 (300)(227) to be applied for soft tissue prediction. The
algorithm has been incorporated in commonly used software packages including
Maxilim prediction software.
In MTM the modelled object is subdivided into
tetrahedral mesh prisms. Inside every tetrahedron, the field is defined by a
linear interpolation of the displacement vectors of the four vertices as defined
by the finite element theory using biomechanical elastic constants.
During simulation of the surgical facial changes, the displacement at one side of
the prism is fixed to represent the skeletal displacement. Vertices at the other
side of the prism are then free to move affected by the vectors of the generated
elastic forces. The new rest position of these free points is established when
total elastic force in all dimensions is zero while the object is at rest.
Using this novel algorithm, Molleman et al., 2007 (227) were able to predict soft
tissue changes in response to orthognathic surgery at up to 0.6 mm accuracy,
with 90% of the errors remained below 1.5 mm. However, the linear stress/strain
model, on which this method is based, might be the cause of constant
overcorrection of specific regions on the face including the upper lip (307).
All previous methods share the common concept of employing engineering
oriented mathematical algorithms, which were originally applied to simulate
mechanical behaviour of a stressed raw materials or alloys based on their pure
physical properties, and modifying this to simulate the biomechanical behaviour
of a biological tissue under stress. The methods could be divided into two main
groups; the first group are those who appreciated the differences between
biological tissue and raw material properties such as the non-linear FEM by
Delingette et al., 1998 (304) and the anatomy based FEM by Keeve et al., 1996
(301), with their algorithms built on a non-linear stress-strain correlation. Those
researchers were faced with individual tissue variations in addition to human
differences, which rendered their analysis to be of a limited accuracy with a
high need for computation power. These might be the reason that none of these
methods had broad clinical applications.
Anas Almukhtar 2016
222
Chapter Four
Discussion
The second group of prediction methods are those which ignored the difference
between the biological tissues and the raw materials and built the analysis on
linear stress- strain correlation hypotheses such as MTM (227), Linear FEM (299)
and VISU prediction (232,255).The algorithm expressed relatively lower
simulation accuracy (299). However, they showed a reduced need for
computational power with a higher level of reproducibility (255).
All the methods were not fully successful in bridging the gap between the raw
materials and biological tissue to develop a reliable prediction of soft tissue
changes following orthognathic surgery. The algorithm that was developed in
this study is based on decomposing the actual soft tissue changes following
orthognathic surgery into its principal components of the three-dimensional
shape variation and finding the correlation of each to the associated skeletal
displacement. This was accomplished by applying the principal component
analysis (PCA) to the CBCT scans of a homogenous cohort of cases to develop the
prediction algorithm. The identified principal components would predict the
facial shape changes in response to orthognathic surgery for future cases.
In this method, the principal components of the shape difference for soft tissue
prediction were derived from the actual soft tissue changes which are
biologically reliable. The simple statistical equation of PCA and the relatively
low required computational power allowed an instant prediction.
The
segmentation of the face into 6 anatomical regions (nose, upper lip, lower lip,
chin, right paranasal region and left paranasal region) facilitated region specific
analysis and produced an anatomically guided prediction.
The estimation of the soft tissue changes is based on the pre analysed PCA
values where the skeletal surgical movement is simply a one numerical value in
each of the x, y and z dimensions. This enabled the algorithm to predict soft
tissue changes without the need for the skeletal surface model similar to the
FEM and MTM prediction methods. In other words, prediction of soft tissue
changes
could
be
applied
to
any
facial
soft
tissue
model
such
as
stereophotogrammetry image or laser scanned image simply by providing the
numerical three-dimensional skeletal surgical movements. The overall accuracy
Anas Almukhtar 2016
223
Chapter Four
Discussion
of the prediction was comparable to previous methods. However, prediction of
the changes at specific regions such as the upper and lower lip showed better
results when compared individually. The method could be applied to a variety of
facial deformities including cleft lip and palate, facial paralysis and facial
reconstruction surgery without the need for CBCT scan. The prediction of various
facial motions could also be achieved with the appropriate application of a 4D
training set.
4.5.4 Measurement of prediction accuracy
The aims of the prediction algorithms are to simulate the surgical outcome of
facial soft tissue changes which will help in patient education and, more
importantly, in finalising a surgical treatment plan. Thus, the accuracy of
prediction is crucial.
Three general criteria should be considered when establishing a method to
evaluate the prediction accuracy. Firstly, the measurements should disclose
errors at all the vertices of the 3D mesh rather than at a few selected points;
secondly the measurements of prediction accuracy using 3D surface mesh should
not be based on the mean value of the whole face as this will camouflage errors
at specific facial regions; and lastly a reliable form of correspondence should
exist to allow tracking of these vertices.
Most of the previously published studies which evaluated the accuracy of soft
tissue prediction following orthognathic surgery were either based on surface
analysis or evaluation of the changes at certain landmarks. Surface based
analysis is the most common approach(158,170,227,296,303,308). The analysis is
based on calculating the square root of the distances of all vertices on the
predicted facial meshes to the nearest points on the real post-operative facial
surface mesh. The main drawback of this method is the lack of a true
correspondence between the two meshes and more important is the reliance in
some studies (227,303)on the whole facial mesh for estimation of the mean
value of the prediction accuracy. The inclusion of anatomical areas away from
the surgical site, including the eye region which has a low error values, could
Anas Almukhtar 2016
224
Chapter Four
Discussion
camouflage the relatively higher errors in specific regions of the face. Bianchi et
al., 2010 and Marchetti et al., 2011 (170,296), tried to overcome this problem
by including the lower and midface regions only in the analysis in addition to
setting a threshold for the acceptable error and calculated the percentage of
the vertices which lies below that level. Their method gave a better description
of the prediction accuracy (86.8% of vertices lower than 2mm of error) using
SurgiCase® software. However, this was not region specific, in addition to the
classical lack of the anatomical surface correspondence. Further refinement of
the analysis was carried out by Shafi et al., 2012 (158) and Khambay et al.,
2014(309). Both studies added the segmentation of the facial meshes into
various anatomical regions to measure the relative accuracy at each area using
Maxilim and 3dMD Vultus software packages respectively. Although this method
provided a more detailed analysis of regional error, it still suffers from the
fundamental lack of anatomical correspondences in addition to the inevitable
subjectivity in the segmentation of the facial regions. However, to avoid the
latter problem, 90th percentile of the mesh surface was considered in the
analysis.
On the other hand, landmark based analysis was applied to investigate the
accuracy of soft tissue change prediction following orthognathic surgery.
Hemelen et al., 2015 (310) applied landmark based analysis to investigate the
accuracy of 3D soft tissue prediction. This approach has the advantage of
establishing anatomical correspondence between landmarks on the predicted
and post-surgical facial meshes. However, using few landmarks on the face does
not represent the surface changes comprehensively.
In an attempt to overcome the deficiencies associated with both methods
(surface and landmarks based analyses), some studies were undertaken to
incorporate the two methods in the analysis (237,311). The combined analysis
benefited from the advantages of both of them but it also included their
associated limitations. Neither the limited reliability of the closest point
correspondence associated with the surface distance measurements, nor the
deficient surface representation of the landmarks based analysis were
addressed.
Anas Almukhtar 2016
225
Chapter Four
Discussion
In this study, the accuracy measurement was based on the dense anatomical
correspondence provided by the conformed generic mesh index. This approach
combined the comprehensiveness of surface analysis with the reliability of
anatomical correspondence.
“Leave-One-Out cross-validation” is a well-known statistical method which was
applied in this study to overcome the problem of the relatively small sample
size. The method considers one of the total 30 samples included in this study as
a tester while the remaining 29 samples as training set for the PCA. The result of
the training set is then applied to predict the surgical results of the tester case
and compared against its postoperative image. The next step is to choose the
next case in the 30 cases sample as a tester and follow the same root, and so on.
This produced 30 measurements for the analysis of the accuracy prediction
method of soft tissue changes to orthognathic surgery. The procedure was
applied six times one for each of the segmented facial region.
The application of the dense correspondence analysis in facial anthropometry
and specifically for orthognathic surgery, was first introduced by Mao et al.,
2006 (176). The approach not only combines the advantages of the landmarks
based analysis and surface based analysis, but also addresses most of their
deficiencies. The applied generic mesh in this study consisted of over 1000
vertices, the use of the generic mesh index made every single mesh vertex as an
actual corresponding landmark between the analysed images. This addressed
both issues of the lack of actual correspondence and lack of surface
representation associated with the previous methods.
The main advantages of our analysis could be summarised in the following
points:
1. The use of the generic mesh index provided a comprehensive and
accurate method for analysis of the facial soft tissue changes in response
to orthognathic surgery.
Anas Almukhtar 2016
226
Chapter Four
Discussion
2. The adoption of the novel directional change analysis, based on the
conformed generic mesh index enabled the analysis of errors in each of
the x, y and z dimensions. This provided more detailed information on the
prediction accuracy in 3D.
3. The segmentation of the face into anatomical regions of interest lends
itself to a comprehensive and specific analysis of the errors of soft tissue
prediction in response to orthognathic surgery.
4. The adoption of the true mean and standard deviation, rather than the
absolute mean used in previous studies (158,309), augments the
credibility of the results, as it shows the variation of the measurements
around the mean. In addition, it enables the differentiation between over
and under estimation of facial soft tissue changes in response to surgery
which was not possible using the absolute mean distance measurements.
However it is essential to consider the standard deviation for the
interpretation of the mean values.
In general, the mean±SD of the difference between the predicted and actual
changes of facial soft tissue in the A-P (y dimension) was equal or lower than the
lowest previously recorded value of 2 mm (158,170,296,309) at the lower lip,
nose and right and left paranasal regions. However, a slightly higher error was
recorded for the upper lip (0.12 mm ± 0.95 mm) and chin (-0.03 mm ± 2.19 mm)
regions.
A general tendency of overestimation at the upper lip change in response to Le
Fort I osteotomy was observed. This was in agreement with Shafi et at., (158)
and Ullah et al., 2014 (236), who recorded similar findings using the MTM and
MSM algorithm respectively.
The prediction approach is a promising prospect. However, as with any other
approach, some limitations do exist. Nevertheless, most of these are
Anas Almukhtar 2016
227
Chapter Four
Discussion
manageable and could be addressed through further research. The main
weaknesses can be summarised as follows:
1. The relatively high cumulative baseline errors associated with the
preparatory stage up to the final prediction. Errors associated with
generic conformation including landmarking accuracy which imposes a
total of around 1.5mm baseline error. Further improvement of the
landmarking accuracy and conformation algorithm could address most
of these errors.
2. The low sample size used as training set for the prediction algorithm.
Although the level of the prediction accuracy is close to the baseline
error values, increasing the number of the training set samples will
improve the accuracy of prediction.
3. Increase the homogeneity in terms of phenotype by sub phenotypes.
Anas Almukhtar 2016
228
5
C
Conclusions & Suggestions
ontents
5.1
CONCLUSIONS ................................................................................................................. 230
5.2
SUGGESTIONS FOR FUTURE STUDIES ...................................................................................... 231
5.3
POSSIBLE APPLICATIONS .................................................................................................... 232
Anas Almukhtar 2016
229
Chapter Five
5.1 Conclusions
Conclusions and suggestions
Voxel based registration and surface based registration are both valid methods
of 3D images superimposition with a comparable registration accuracy level.
Voxel based registration; however, has the advantage of preserving the link
between the skeletal and soft tissue structures during the registration process.
Direct DICOM slice landmarking is clinically applicable measurement method
with an acceptable level of accuracy in measuring skeletal tissue displacement
following orthognathic surgery in 3D with low intra- and inter-examiner
landmarking errors.
Generic mesh conformation has an acceptable level of accuracy. Variable range
of errors at different areas of the face were recorded, peripherally more than
centrally, which were all within an acceptable range for clinical application.
The introduced dense correspondence analysis (directional analysis) was novel.
New range of information regarding soft tissue changes following orthognathic
surgery.
The novel statistical algorithm for prediction of soft tissue changes following Le
Fort I advancement surgery was proved to be applicable, practical and produced
low prediction errors.
Anas Almukhtar 2016
230
Chapter Five
5.2 Suggestions for future studies
Conclusions and suggestions
1. A larger sample size would be beneficial since more variation in
surgical procedures could be specifically analysed. This could be
achieved by establishing a substantial database, probably through
multicentre collaboration.
2. The prediction algorithm in its current state predicts soft tissue
changes following orthognathic surgery based on regional analysis.
Further
improvement
on
the
prediction
algorithm,
including
connection between facial regions to produce a full face prediction,
will be the next step.
Anas Almukhtar 2016
231
Chapter Five
5.3 Possible applications
Conclusions and suggestions
The promising approach of dense correspondence analysis opens the doors
toward a simpler and yet more reliable tool of a multi-disciplinary applications
including
cleft
lip
and
palate,
facial
paralysis,
variable
craniofacial
reconstruction procedures and breast cancer reconstructive surgery. The novel
approach of producing a reliable soft tissue prediction without the need for 3D
radiography will be a significant addition to the current diagnoses and treatment
planning protocol.
Anas Almukhtar 2016
232
6
References
Chapter Six
References
1.
Moorrees CFA, Lebret LM. The mesh Diagram and Cephalometrics. Angle
Orthodontics; 1962. p. 214–31.
2.
Moorrees CFA, Venrooij ME van, Tandarts, Lebret LM 1., Carlton B. Glatky
MA, Jr RK. New norms for the mesh diagram analysis. Am J Orthod.
1976;69(1):57–71.
3.
Ferrario VF, Sforza C, Schmitz JH, Miani A. ORIGINAL ARTICLE A threedimensional computerized mesh diagram analysis and its application in soft
tissue facial morphometry. Am J Orthod Dentofac Orthop. 1998;114:404–
13.
4.
Lovesey EJ. The devet t of a 3-dimensional anthrop measuring technique.
Appl Ergon. 1974;5(1):36–41.
5.
Motoyoshi M, Arai HY, Ridge P. A three-dimensional measuring system for
the human face using three-directional photography. Am J Orthod
Dentofac Orthop. 1992;101:431–40.
6.
Wu J, Tillett R, Mcfarlane N, Ju X, Siebert JP, Schofield P. Extracting the
three-dimensional shape of live pigs using stereo photogrammetry. Comput
Electron Agric. 2004;44:203–22.
7.
Winder RJ, Darvann TA, Mcknight W, Magee JDM, Ramsay-baggs P.
Technical validation of the Di3D stereophotogrammetry surface imaging
system. Br J Oral Maxillofac Surg. 2008;46:33–7.
8.
Burke P., Beard HF. Stereophotogranwnetry of the face. Am J Orthod.
1967;53(10):769–82.
9.
Tzou C-HJ, Artner NM, Pona I, Hold A, Placheta E, Kropatsch WG.
Comparison of three-dimensional surface-imaging systems. J Plast Reconstr
Aesthet Surg. 2014;67(4):489–97.
10.
Ras F, Habets LLMH, van Ginkel FC, Prahl-Andersen B. Quantification of
facial morphology using stereophotogrammetry — demonstration of a new
concept. J Dent. 1996;24(5):369–74.
11.
Hajeer MY. Changes Psychosocial and Following Orthognathic Surgery
Mohammad Younis Hajeer Thesis submitted to the University of Glasgow
for the degree of PhD in Orthodontics Faculty. 2003.
12.
Khambay B, Nairn N, Bell A, Miller J, Bowman A, Ayoub AF. Validation and
reproducibility of a high-resolution three-dimensional facial imaging
system. 2008;46:27–32.
13.
Pfister A, West AM, Bronner S, Noah JA. Comparative abilities of Microsoft
Kinect and Vicon 3D motion capture for gait analysis. J Med Eng Technol.
2014;38(5):274–80.
14.
Macleod C a, Conway B a, Allan DB, Galen SS. Development and validation
of a low-cost, portable and wireless gait assessment tool. Med Eng Phys.
2014;36(4):541–6.
Anas Almukhtar 2016
234
Chapter Six
References
15.
Mishima K, Yamada T, Ohura A, Sugahara T. Production of a range image
for facial motion analysis: a method for analyzing lip motion. Comput Med
Imaging Graph. 2006;30(1):53–9.
16.
Sjögreen L, Lohmander a, Kiliaridis S. Exploring quantitative methods for
evaluation of lip function. J Oral Rehabil. 2011;38(6):410–22.
17.
Al-Anezi T, Khambay B, Peng MJ, O’Leary E, Ju X, Ayoub a. A new method
for automatic tracking of facial landmarks in 3D motion captured images
(4D). Int J Oral Maxillofac Surg. 2013;42(1):9–18.
18.
Ju X, O'Leary E, Peng M, Al-Anezi T, Ayoub A, Khambay B. Evaluation of
the Reproducibility of Nonverbal Facial Expressions Using a 3D Motion
Capture System. Cleft Palate Craniofac J. 2014 Dec 22. [Epub ahead of
print] .
19.
Shujaat S, Khambay BS, Ju X, Devine JC, McMahon JD, Wales C. The
clinical application of three-dimensional motion capture (4D): a novel
approach to quantify the dynamics of facial animations. Int J Oral
Maxillofac Surg. 2014;43(7):907–16.
20.
Popat H, Zhurov AI, Richmond S, Marshall D, Rosin PL. Determining normal
and abnormal lip shapes during movement for use as a surgical outcome
measure. J Oral Rehabil. 2013;40(5):348-57.
21.
McCance A, Moss J, Wright W. A three-dimensional soft tissue analysis of
16 skeletal class III patients following bimaxillary surgery. Br J Oral
Maxillofac Surg. 1992;30:221–32.
22.
Moss JP, Mccance AM, Fright WR, Linney AD. A three-dimensional soft
tissue analysis of fifteen patients with Class II , Division I malocclusions
after bimaxillary surgery. Am J Orthod Dentofacial
Orthop. 1994;105(5):430-7.
23.
Kau, C.H., Zhurov, A.I., Knox, J., Chestnutt, I., Playle, R., Hartles, F.R.,
and Richmond, S. Reliability of measuring facial morphology. Am J Orthod
Dentofacial Orthop 2005;128:(3)424-430.
24.
Kau CH, Richmond S, Zhurov AI, Knox J, Chestnutt I, Hartles F. Reliability
of measuring facial morphology with a 3-dimensional laser scanning
system. Am J Orthod Dentofacial Orthop. 2005;128(4)424–30.
25.
Ovsenik M, Perinetti G, Zhurov A, Richmond S, Primozic J. Threedimensional assessment of facial asymmetry among pre-pubertal class III
subjects: a controlled study. Eur J Orthod. 2014;36(4):431-5.
26.
Perinetti G, Zhurov A, Richmond S, Ovsenik M. Assessment of facial
asymmetry in growing subjects with a three- dimensional laser scanning
system. Orthod Craniofac Res. 2012;15(4):237-44..
27.
Zhurov A, Playle R, Richmond S. S Richmond A three-dimensional look for
facial differences between males and females in a British-Caucasian
sample aged 15½ years old. Orthod Craniofacial Res. 2008;11:180–5.
Anas Almukhtar 2016
235
Chapter Six
References
28.
Stephen RMBCHK, Ovsenik NIHAZMUSMMO. Facial Morphology of Slovenian
and Welsh White Populations Using 3-Dimensional Imaging. Angle Orthod.
2009;79:640–5.
29.
Kau CH, Richmond S. Three-dimensional analysis of facial morphology
surface changes in untreated children from 12 to 14 years of age. Am J
Orthod Dentofac Orthop. 2008;134(6):751–60.
30.
Guest E, Berry E, Morris D. Clinical paper : Orthognathic surgery Novel
methods for quantifying soft tissue changes after orthognathic surgery. Int
J Oral Maxillofac Surg. 2001;30:484–9.
31.
Kau C, Richmond S. Three‐dimensional surface acquisition systems for the
study of facial morphology and their application to maxillofacial surgery.
Int J Med Robot. 2007;3(2):97–110.
32.
Tartaglia GM, Grandi G, Mian F, Sforza C, Ferrario VF. Non-Invasive 3D
Facial Analysis And Surface Electromyography During Functional PreOrthodontic Treatment. J Appl Oral Sci. 2009;17(5):487–94.
33.
Sforza C, Grandi G, De Menezes M, Tartaglia GM, Ferrario VF. Age- and
sex-related changes in the normal human external nose. Forensic Sci Int.
2011;204(1-3):205.e1–9.
34.
Sforza C, Peretta R, Grandi G, Ferronato G, Ferrario VF. Three-dimensional
facial morphometry in skeletal Class III patients. A non-invasive study of
soft-tissue changes before and after orthognathic surgery. Br J Oral
Maxillofac Surg. 2007;45(2):138–44.
35.
Kalender W a. X-ray computed tomography. Phys Med Biol.
2006;51(13):R29–43.
36.
Guerrero ME, Jacobs R, Loubele M, Schutyser F, Suetens P, van
Steenberghe D. State-of-the-art on cone beam CT imaging for preoperative
planning of implant placement. Clin Oral Investig. 2006;10(1):1–7.
37.
White SC, Pharoah MJ. The evolution and application of dental
maxillofacial imaging modalities. Dent Clin North Am. 2008;52(4):689–705.
38.
Mah JK, Danforth R a., Bumann A, Hatcher D. Radiation absorbed in
maxillofacial imaging with a new dental computed tomography device.
Oral Surgery, Oral Med Oral Pathol Oral Radiol Endodontology.
2003;96(4):508–13.
39.
Park W, Kim K. Reduction of metal artifact in three-dimensional computed
tomography (3D CT) with dental impression materials. Conf Proc IEEE Eng
Med Biol Soc. 2007;2007:3496-9.
40.
Nandini S, Velmurugan N, Kandaswamy D. Calcific healing of a crown root
fracture of a maxillary central incisor evaluated with spiral computed
tomography and hounsfield units: a case report. Dent Traumatol.
2008;24(6):e96–100.
Anas Almukhtar 2016
236
Chapter Six
References
41.
Gray a. D, Marks JM, Stone EE, Butler MC, Skubic M, Sherman SL.
Validation of the Microsoft Kinect as a Portable and Inexpensive Screening
Tool for Identifying ACL Injury Risk. Orthop J Sport Med. 2014; 2(7)(suppl
2).
42.
Sohmura T, Hojoh H. A novel method of removing artifacts because of
metallic dental restorations in 3‐D CT images of jaw bone. Clin oral
Implant Res. 2005;16:728–35.
43.
Nkenke E, Zachow S, Benz M, Maier T, Veit K, Kramer M. Fusion of
computed tomography data and optical 3D images of the dentition for
streak artefact correction in the simulation of orthognathic surgery.
Dentomaxillofac Radiol. 2004;33(4):226–32.
44.
Scarfe WC, Farman AG, Sukovic P. Clinical applications of cone-beam
computed tomography in dental practice. J Can Dent Assoc. 2006;72(1):75–
80.
45.
Mah J, Hatcher D. Three-dimensional craniofacial imaging. Am J Orthod
Dentofac Orthop. 2004;126(3):308–9.
46.
Palomo JM, Kau CH, Palomo LB, Hans MG. Three-dimensional cone beam
computerized tomography in dentistry. Dent Today. 2006;25(11):130, 1325.
47.
Swennen GRJ, Mollemans W, De Clercq C, Abeloos J, Lamoral P, Lippens F.
A cone-beam computed tomography triple scan procedure to obtain a
three-dimensional augmented virtual skull model appropriate for
orthognathic surgery planning. J Craniofac Surg. 2009;20(2):297–307.
48.
Lewis EL, Dolwick MF, Abramowicz S, Reeder SL. Contemporary imaging of
the temporomandibular joint. Dent Clin North Am. 2008;52(4):875–90, viii.
49.
Bearcroft PWP. Imaging modalities in the evaluation of soft tissue
complaints. Best Pract Res Clin Rheumatol. 2007;21(2):245–59.
50.
Strauss R a, Burgoyne CC. Diagnostic imaging and sleep medicine. Dent
Clin North Am. 2008;52(4):891–915, viii.
51.
Tasaki MM, Westeson P-L. Temporomandibular joint: diagnostic accuracy
with sagittal and coronal MR imaging. Radiology. 1993;186:723–9.
52.
Ayoub a F, Siebert P, Moos KF, Wray D, Urquhart C, Niblett TB. A visionbased three-dimensional capture system for maxillofacial assessment and
surgical planning. Br J Oral Maxillofac Surg. 1998;36(5):353–7.
53.
Eggers G, Rieker M, Kress B, Fiebach J, Dickhaus H, Hassfeld S. Artefacts in
magnetic resonance imaging caused by dental material. MAGMA.
2005;18(2):103–11.
54.
Hell B. 3D Sonography. Int J Oral Maxillofac Surg. 1995;24(1):84–9.
Anas Almukhtar 2016
237
Chapter Six
References
55.
Nonnast-Daniel B, Martin RP, Lindert O, Mügge A, Schaeffer J, vd Lieth H,
Söchtig E, Galanski M, Koch KM, Daniel WG. Colour doppler ultrasound
assessment of arteriovenous haemodialysis fistulas Satisfactory function of
the arteriovenous fistula. Lancet. 1992;339(8786):143-5.
56.
Lomka PIJS, Andel JOM, Owney DOD, Enster AAF. Evaluation of voxel-based
registration of 3-D power Doppler ultrasound and 3-D magnetic resonance
angiographic images of carotid arteries. Ultrasound Med Biol.
2001;27(7):945–55.
57.
Akizuki H, Yoshida H, Michi K. Ultrasonographic evaluation during
reduction of zygomatic arch fractures. J Cranio-Maxillofacial Surg. 1990
;18(6):263–6.
58.
McCann P, Brocklebank L, Ayoub A. Assessment of zygomatico-orbital
complex fractures using ultrasonography. Br J Oral Maxillofac Surg.
2000;38(5):525-9.
59.
Khambay B, Nebel J, Bowman J, Walker F, Hadley D, Ayoub A. 3D
stereophotogrammetric image superimposition onto 3D CT scan images the
future of orthognathic surgery A pilot study. Int J Adult Orthodon
Orthognath Surg. 2002;17(4):331–41.
60.
Hajeer M, Ayoub A, Millett D, MitchumBock, Siebert P. Three-dimensional
imaging in orthognathic surgery The clinical application of a new method.
Int Adult Orthod Orthognath Surg. 2002;17(4):318–30.
61.
Lorensen W, Cline H. Marching cubes: A high resolution 3D surface
construction algorithm. ACM siggraph Comput Graph. 1987;21(4):163–9.
62.
Udupa J. Interactive segmentation and boundary surface formation for 3-D
digital images. Comput Graph Image Process. 1982;18(3):213–35.
63.
Roth SD. Ray casting for modeling solids. Comput Graph Image Process.
1982;18(2):109–44.
64.
Jayaratne YSN, Zwahlen R a, Lo J, Cheung LK. Three-dimensional color
maps: a novel tool for assessing craniofacial changes. Surg Innov.
2010;17(3):198–205.
65.
Hoefert CS, Bacher M, Herberts T, Krimmel M, Reinert S, Hoefert S, Göz G.
Implementing a superimposition and measurement model for 3D sagittal
analysis of therapy-induced changes in facial soft tissue: a pilot study. J
Orofac Orthop. 2010;71(3):221–34.
66.
Maal TJ, Verhamme LM, van Loon B, Plooij JM, Rangel FA, Kho A,
Bronkhorst EM, Bergé SJ. Variation of the face in rest using 3D
stereophotogrammetry. Int J Oral Maxillofac Surg. 2011;40(11):1252–7.
67.
Lee J-H, Kim M-J, Kim S-M, Kwon O-H, Kim Y-K. The 3D CT superimposition
method using image fusion based on the maximum mutual information
algorithm for the assessment of oral and maxillofacial surgery treatment
results. Oral Surg Oral Med Oral Pathol Oral Radiol. 2012;114(2):167–74.
Anas Almukhtar 2016
238
Chapter Six
References
68.
Lee J, Kim M, Kim S, Kwon O, Kim Y. The 3D CT superimposition method
using image fusion based on the maximum mutual information algorithm
for the assessment of oral and maxillofacial surgery. Oral Surg Oral Med
Oral Pathol Oral Radiol. 2012;114(2):167–74.
69.
Nada RM, Maal TJJ, Breuning KH, Bergé SJ, Mostafa Y a, Kuijpers-Jagtman
AM. Accuracy and reproducibility of voxel based superimposition of cone
beam computed tomography models on the anterior cranial base and the
zygomatic arches. PLoS One. 2011;6(2):e16520.
70.
Almukhtar A, Ju X, Khambay B, McDonald J, Ayoub A. Comparison of the
accuracy of voxel based registration and surface based registration for 3D
assessment of surgical change following orthognathic surgery. PLoS One.
2014;9(4):e93402.
71.
Vinchon M, Pellerin P, Pertuzon B, Fénart R, Dhellemmes P. Vestibular
orientation for craniofacial surgery: application to the management of
unicoronal synostosis. Childs Nerv Syst. 2007;23(12):1403–9.
72.
Baumrind S, Frantz RC. The reliability of head film measurements: 1.
Landmark identification. Am J Orthod. 1971;60(2):111–27.
73.
Baumrind S, Frantz RC. The reliability of head Jilm measurements: 2.
Conventional angular and linear measures. Am J Orthod. 1971;60(5):555–
517.
74.
Baumrind S, Miller D, Molthen R. The reliability of head film
measurements: 3. Tracing superimposition. Am J Orthod. 1976;70(6):617–
44.
75.
Adams GL, Gansky S a., Miller AJ, Harrell WE, Hatcher DC. Comparison
between traditional 2-dimensional cephalometry and a 3-dimensional
approach on human dry skulls. Am J Orthod Dentofac Orthop.
2004;126(4):397–409.
76.
Van Vlijmen OJC, Maal T, Bergé SJ, Bronkhorst EM, Katsaros C, KuijpersJagtman a M. A comparison between 2D and 3D cephalometry on CBCT
scans of human skulls. Int J Oral Maxillofac Surg. 2010;39(2):156–60.
77.
William B. Downs. Analysis of facial profile. Angle Orthod. 1956;26(4):191–
212.
78.
Downs WB, Aurora Ill. Variations in Facial Relationships: Their Significance
in Treatment and Prognosis. Angle Orthod. 1948;14(3):812–40.
79.
Steiner CC. Cephalometrics for you and me. Am J Orthod. 1953;39(10):720–
55.
80.
Steiner CC. Cephalometrics in clinical practice. Angle Orthod.
1959;29(1):8–20.
Anas Almukhtar 2016
239
Chapter Six
References
81.
Satrom KD, Sinclair PM, Wolford LM. The stability of double jaw surgery: a
comparison of rigid versus wire fixation. Am J Orthod Dentofacial Orthop.
1991;99(6):550–63.
82.
Burston CJ. The integumental profile. Am J Orthod. 1958;44(1):1–25.
83.
Sassouni V. A classification of skeletal facial types. Am J Orthod.
1969;55(2):109–23.
84.
Phillips JG. Photo-cephalometric analysis in treatment planning for
surgical correction of facial disharmonies. J Maxillofac Surg.
1978;6(3):174–9.
85.
Ricketts R. Cephalometric analysis and synthesis. Angle Orthod.
1961;31(3):141–56.
86.
Hwang H-S, Kim W-S, McNamara J a. Ethnic differences in the soft tissue
profile of Korean and European-American adults with normal occlusions
and well-balanced faces. Angle Orthod. 2002;72(1):72–80.
87.
Sutter R, Turley P. Soft tissue evaluation of contemporary Caucasian and
African American female facial profiles. Angle Orthod. 1988;68(6):487–95.
88.
Di Paolo RJ, Philip C, Maganzini a L, Hirce JD. The quadrilateral analysis:
an individualized skeletal assessment. Am J Orthod. 1983;83(1):19–32.
89.
Kurt Butow. A lateral photometric analysis for aesthetic-orthognathic
treatmen. J Maxillofac Surg. 1984;12(5):201–7.
90.
Swennen GRJ, Schutyser F, Barth E-L, De Groeve P, De Mey A. A new
method of 3-D cephalometry Part I: the anatomic Cartesian 3-D reference
system. J Craniofac Surg. 2006;17(2):314–25.
91.
Fuhrmann RAW. Three-dimensional cephalometry and three-dimensional
skull models in orthodontic/surgical diagnosis and treatment planning.
Semin Orthod. 2002;8(1):17–22.
92.
G.R.J. Swennen, F. Schutyser, E.-L. Barth, A. Lemaitre, C. Malevez, A. De
Mey. Presentation and validation of a voxel-based three-dimensional (3-D)
hard and soft tissue cephalometric analysis. Int J Oral Maxillofac Surg.
2005;34(sup1):72-86.
93.
Olszewski R, Cosnard G, Macq B, Mahy P, Reychler H. 3D CT-based
cephalometric analysis: 3D cephalometric theoretical concept and
software. 2006;48(11):853–62.
94.
Olszewski R, Zech F, Cosnard G, Nicolas V, Macq B, Reychler H. Threedimensional computed tomography cephalometric craniofacial analysis:
experimental validation in vitro. Int J Oral Maxillofac Surg. 2007;36(9):828–
33.
95.
Lee S-H, Kil T-J, Park K-R, Kim BC, Kim J-G, Piao Z, Corre P. Threedimensional architectural and structural analysis--a transition in concept
Anas Almukhtar 2016
240
Chapter Six
References
and design from Delaire's cephalometric analysis. Int J Oral Maxillofac
Surg. 2014;43(9):1154–60.
96.
De Oliveira AEF, Cevidanes LHS, Phillips C, Motta A, Burke B, Tyndall D.
Observer reliability of three-dimensional cephalometric landmark
identification on cone-beam computerized tomography. Oral surg oral med
oral pathol oral radiol Endod. 2009;107(2):256–65.
97.
Lou L, Lagravere MO, Compton S, Major PW, Flores-Mir C. Accuracy of
measurements and reliability of landmark identification with computed
tomography (CT) techniques in the maxillofacial area: a systematic review.
Oral Surg Oral Med Oral Pathol Oral Radiol Endod [Internet]. 2007 Sep
[cited 2014 Apr 29];104(3):402–11. Available from:
http://www.ncbi.nlm.nih.gov/pubmed/17709072
98.
Damstra J, Fourie Z, Huddleston Slater JJR, Ren Y. Reliability and the
smallest detectable difference of measurements on 3-dimensional conebeam computed tomography images. Am j orthod Dentofac orthop.
2011;140(3):e107–14.
99.
Lamichane M, Anderson NK, Rigali PH, Seldin EB, Will L a. Accuracy of
reconstructed images from cone-beam computed tomography scans. Am J
Orthod Dentofac Orthop. 2009;136(2):156.e1–156.e6.
100. Ludlow JB, Gubler M, Cevidanes L, Mol A. Precision of cephalometric
landmark identification: cone-beam computed tomography vs conventional
cephalometric views. Am J Orthod Dentofacial Orthop.
2009;136(3):312.e1–10; discussion 312–3.
101. Damstra J, Fourie Z, Ren Y. Comparison between two-dimensional and
midsagittal three-dimensional cephalometric measurements of dry human
skulls. Br J Oral Maxillofac Surg. 2011;49(5):392–5.
102. Katsumata A, Fujishita M, Maeda M, Ariji Y, Ariji E, Langlais RP. 3D-CT
evaluation of facial asymmetry. Oral Surg Oral Med Oral Pathol Oral Radiol
Endod. 2005;99(2):212–20.
103. Gaia BF, Pinheiro LR, Umetsubo OS, Santos O, Costa FF, Cavalcanti MGP.
Accuracy and reliability of linear measurements using 3-dimensional
computed tomographic imaging software for Le Fort I Osteotomy. Br J Oral
Maxillofac Surg. 2014;52(3):258–63.
104. Rossini G, Cavallini C, Cassetta M, Ph D, Barbato E. 3D cephalometric
analysis obtained from computed tomography. Review of the literature.
Ann Stomatol (Roma). 2011;2(3-4):31–9.
105. Pittayapat P, Limchaichana-Bolstad N, Willems G, Jacobs R. Threedimensional cephalometric analysis in orthodontics: a systematic review.
Orthod Craniofac Res. 2014;17(2):69–91.
106. Almeida RC, Cevidanes LHS, Carvalho FAR, Motta AT, Almeida MAO, Styner
M, T. Turveye, W.R. Proffitb, C. Phillipsb. Soft tissue response to
Anas Almukhtar 2016
241
Chapter Six
References
mandibular advancement using 3D CBCT scanning. Int J Oral Maxillofac
Surg. 2011;40(4):353–9.
107. Cevidanes LHS, Bailey LJ, Tucker SF, Styner M a, Mol A, Phillips CL, William
R. Proffit, and Timothy Turveyh. Three-dimensional cone-beam computed
tomography for assessment of mandibular changes after orthognathic
surgery. Am j orthod Dentofac orthop. 2007;131(1):44–50.
108. De Assis Ribeiro Carvalho F, Cevidanes LHS, da Motta ATS, de Oliveira
Almeida MA, Phillips C. Three-dimensional assessment of mandibular
advancement 1 year after surgery. Am J Orthod Dentofac Orthop.
2010;137(4):S53.e1–S53.e12.
109. Cevidanes LH, Franco AA, Gerig G, Proffit WR, Slice DE, Enlow DH,
Yamashita HK, Kim YJ, Scanavini MA, Vigorito JW. Assessment of
mandibular growth and response to orthopedic treatment with 3dimensional magnetic resonance images. Am J Orthod Dentofac Orthop.
2005;128(1):16–26.
110. Maal TJJ, de Koning MJJ, Plooij JM, Verhamme LM, Rangel FA, Bergé SJ,
W.A. Borstlap. One year postoperative hard and soft tissue volumetric
changes after a BSSO mandibular advancement. Int J Oral Maxillofac Surg.
2012;41(9):1137–45.
111. Park K-R, Park H-S, Piao Z, Kim M-K, Yu H-S, Seo JK, Sang-Hwy Lee. Threedimensional vector analysis of mandibular structural asymmetry. J craniomaxillo-facial Surg. 2013;41(4):338–44.
112. Farkas LG, Bryson W, Klotz J. Is photogrammetry of the face reliable? Plast
Reconstr Surg. Plastic and reconstructive surgery; 1980;66(3):346–55.
113. DiBernardo BE, Adams RL, Krause J, Fiorillo M a, Gheradini G.
Photographic standards in plastic surgery. Plast Reconstr Surg.
1998;102(2):559–68.
114. Ettorre G, Weber M, Schaaf H, Lowry JC, Mommaerts MY, Howaldt H-P.
Standards for digital photography in cranio-maxillo-facial surgery - Part I:
Basic views and guidelines. J cranio-maxillo-facial Surg. 2006;34(2):65–73.
115. Schaaf H, Streckbein P, Ettorre G, Lowry JC, Mommaerts MY, Howaldt H-P.
Standards for digital photography in cranio-maxillo-facial surgery – Part II:
Additional picture sets and avoiding common mistakes. J CranioMaxillofacial Surg. 2006;34(7):444–55.
116. Proffit WR and Fields HW. Contemporary Orthodontics. St. Louis: MosbyYear Book, Inc., 2003: P 233.
117. Farkas LG. Accuracy of anthropometric measurements: past, present, and
future. Cleft palate-craniofacial J. 1996;33(1):10–8; discussion 19–22.
118. Driessen JP, Vuyk H, Borgstein J. New insights into facial anthropometry in
digital photographs using iris dependent calibration. Int J Pediatr
Otorhinolaryngol. 2011;75(4):579–84.
Anas Almukhtar 2016
242
Chapter Six
References
119. DiSaia JP, Ptak JJ, Achauer BM. Digital photography for the plastic
surgeon. Plast Reconstr Surg. 2000;105(7):2636.
120. Farkas LG, Sohm P, Kolar JC, Katic MJ, Munro IR. Inclinations of the facial
profile: Art versus Reality. Plast Reconstr Surg. 1985;75(4):509–19.
121. Hajeer MY, Millett DT, Ayoub a F, Siebert JP. Applications of 3D imaging in
orthodontics: part II. J Orthod. 2004;31(2):154–62.
122. Hwang H-S, Yuan D, Jeong K-H, Uhm G-S, Cho J-H, Yoon S-J. Threedimensional soft tissue analysis for the evaluation of facial asymmetry in
normal occlusion individuals. Korean J Orthod. 2012;42(2):56–63.
123. Devlin MF, Ray A, Raine P, Bowman A, Ayoub AF. Facial symmetry in
unilateral cleft lip and palate following alar base augmentation with bone
graft: a three-dimensional assessment. Cleft palate-craniofacial J.
2007;44(4):391–5.
124. Woo J, Yeol J, Oh T, Man S, Joon S. Frontal soft tissue analysis using a 3
dimensional camera following two-jaw rotational orthognathic surgery in
skeletal class III patients. J Cranio-Maxillofacial Surg. 2014;42(3):220–6.
125. Galantucci L, Percoco G, Gioia E Di. Photogrammetric 3D digitization of
human faces based on landmarks. In: Proceedings of the International
MultiConference of Engineering and computer science. 2009;1:978–88.
126. Honrado CP, Lee S, Bloomquist DS, Larrabee WF. Quantitative assessment
of nasal changes after maxillomandibular surgery using a 3-dimensional
digital imaging system. Arch Facial Plast Surg. 2014;8(1):26–35.
127. Menezes M De, Sforza C. Three-dimensional face morphometry. Dental
Press J Orthod. 2010;15(1):13–5.
128. Sforza C, Peretta R, Grandi G, Ferronato G, Ferrario VF. Soft tissue facial
volumes and shape in skeletal Class III patients before and after
orthognathic surgery treatment. J Plast Reconstr aesthetic Surg.
2007;60(2):130–8.
129. Terajima M, Yanagita N, Ozeki K, Hoshino Y, Mori N, Goto TK, Tokumori K,
Aoki Y, Nakasima A. Three-dimensional analysis system for orthognathic
surgery patients with jaw deformities. Am J Orthod Dentofac Orthop.
2008;134(1):100–11.
130. Kim Y-I, Park S-B, Son W-S, Hwang D-S. Midfacial soft-tissue changes after
advancement of maxilla with Le Fort I osteotomy and mandibular setback
surgery: comparison of conventional and high Le Fort I osteotomies by
superimposition of cone-beam computed tomography volumes. J oral
Maxillofac Surg. 2011;69(6):e225–33.
131. Park S, Kim Y, Hwang D, Lee J. Midfacial soft-tissue changes after
mandibular setback surgery with or without paranasal augmentation:
Cone-beam computed tomography ( CBCT ) volume superimposition. J
Cranio-Maxillofacial Surg. 2013;41(2):119–23.
Anas Almukhtar 2016
243
Chapter Six
References
132. Sforza C, Grandi G, Pisoni L, Diablo C, Gondolfini M, Ferrario V. Soft tissue
facial morphometry in subjects with Moebius syndrome. Eur J Oral Sci.
2009;117:695–703.
133. Sforza C, Laino A, Grandi G, Pisoni L, Ferrario V. Three-dimensional facial
asymmetry in attractive and normal people from childhood to young
adulthood. Symmetry (Basel). 2010;2:1925–44.
134. Ryckman MS, Harrison S, Oliver D, Sander C, Boryor AA, Hohmann AA, Kilic
F, Kim KB. Soft-tissue changes after maxillomandibular advancement
surgery assessed with cone-beam computed tomography. Am J Orthod
Dentofac Orthop. 2010;137(4 Suppl):S86–93.
135. Schwenzer-Zimmerer K. Systematic contact-free 3D topometry of the soft
tissue profile in cleft lips. Cleft palate-craniofacial J. 2008;45(6):607–13.
136. Park S-B, Yoon J-K, Kim Y-I, Hwang D-S, Cho B-H, Son W-S. The evaluation
of the nasal morphologic changes after bimaxillary surgery in skeletal class
III maloccusion by using the superimposition of cone-beam computed
tomography (CBCT) volumes. J cranio-maxillo-facial Surg. 2012;40(4):e87–
92.
137. Terajima M, Furuichi Y, Aoki Y, Goto TK, Tokumori K, Nakasima A. A 3dimensional method for analyzing facial soft-tissue morphology of patients
with jaw deformities. Am J Orthod Dentofac Orthop. 2009;135(6):715–22.
138. Schimmel M, Christou P, Houstis O, Herrmann FR, Kiliaridis S, Müller F.
Distances between facial landmarks can be measured accurately with a
new digital 3-dimensional video system. Am J Orthod Dentofac Orthop.
2010;137(5):580.e1–580.e10; discussion 580–1.
139. Vezzetti E, Marcolin F. Geometry-based 3D face morphology analysis: softtissue landmark formalization. Multimed Tools Appl. 2014;68(3):895–929.
140. Fourie Z, Damstra J, Gerrits PO, Ren Y. Evaluation of anthropometric
accuracy and reliability using different three-dimensional scanning
systems. Forensic Sci Int. 2011;207:127–34.
141. Plooij J, Swennen G, Rangel F. Evaluation of reproducibility and reliability
of 3D soft tissue analysis using 3D stereophotogrammetry. Int J Oral
Maxillofac Surg. 2009;38(3):267–73.
142. Othman SA, Ahmad R, Jamaludin AFMM. Reproducibility of facial soft
tissue landmarks on facial images captured on a 3D camera. Aust Orthod J.
2013;29(1):58–66.
143. Gwilliam JR, Cunningham SJ, Hutton T. Reproducibility of soft tissue
landmarks on three-dimensional facial scans. Eur J Orthod.
2006;28(5):408–15.
144. Toma a M, Zhurov a, Playle R, Ong E, Richmond S. Reproducibility of facial
soft tissue landmarks on 3D laser-scanned facial images. Orthod Craniofac
Res. 2009;12(1):33–42.
Anas Almukhtar 2016
244
Chapter Six
References
145. Nakamura N, Okawachi T, Nozoe E, Nishihara K, Matsunaga K. Threedimensional analyses of nasal forms after secondary treatment of bilateral
cleft lip-nose deformity in comparison to those of healthy young adults. J
oral Maxillofac Surg. 2011;69(11):e469–81.
146. Schwenzer-Zimmerer K, Chaitidis D, Boerner I, Kovacs L, Schwenzer NF,
Holberg C, Zeilhofer HF. Systematic contact-free 3D topometry of the soft
tissue profile in cleft lips. Cleft palate-craniofacial J. 2008;45(6):607–13.
147. Ubaya T, Sherriff a, Ayoub a, Khambay B. Soft tissue morphology of the
naso-maxillary complex following surgical correction of maxillary
hypoplasia. Int J Oral Maxillofac Surg. 2012;41(6):727–32.
148. Chen C, Lai S, Lee H, Chen K, Hsu K. Soft-tissue profile changes after
orthognathic surgery of mandibular prognathism. Kaohsiung J Med Sci.
2012;28(4):216–9.
149. Verzé L, Nasi A, Quaranta F, Vasino V, Prini V, Ramieri G. Quantification of
facial movements by surface laser scanning. J Craniofac Surg.
2011;22(1):60–5.
150. Popat H, Richmond S, Marshall D, Rosin PL. Three-dimensional assessment
of functional change following Class 3 orthognathic correction--a
preliminary report. J cranio-maxillo-facial Surg. 2012;40(1):36–42.
151. Amm C a., Denny AD. Correction of Sagittal Synostosis Using
Foreshortening and Lateral Expansion of the Cranium Activated by Gravity:
Surgical Technique and Postoperative Evolution. Plast Reconstr Surg.
2005;116(3):723–35.
152. Claes P, Walters M, Clement J. Improved facial outcome assessment using
a 3D anthropometric mask. Int J Oral Maxillofac Surg. 2012;41(3):324–30.
153. Bugaighis I, O’Higgins P, Tiddeman B, Mattick C, Ben Ali O, Hobson R.
Three-dimensional geometric morphometrics applied to the study of
children with cleft lip and/or palate from the North East of England. Eur J
Orthod. 2010;32(5):514–21.
154. Bookstein FL. Linear machinery for morphological distortion. Comput
Biomed Res. 1978;11(5):435–58.
155. Paton NI, Yang Y, Sitoh Y-Y, Tha NO. Validation of three-dimensional laser
scanning for the assessment of facial fat changes. HIV Med. 2007;8(8):498–
503.
156. Luximon Y, Ball R, Justice L. The Chinese Face: A 3D Anthropometric
Analysis. In: Proceedings of the TMCE, Anacona, Italy. 2010. pp. 1–11.
157. Toma AM, Zhurov AI, Playle R, Marshall D, Rosin PL, Richmond S. The
assessment of facial variation in 4747 British school children. Eur J Orthod.
2012;34:655–64.
Anas Almukhtar 2016
245
Chapter Six
References
158. Shafi MI, Ayoub a, Ju X, Khambay B. The accuracy of three-dimensional
prediction planning for the surgical correction of facial deformities using
Maxilim. Int J Oral Maxillofac Surg. 2013;42(7):801–6.
159. Maal TJJ, Plooij JM, Verthamme LM, Rangel FA, Berge SJ, Borstlap WA.
One year postoperative hard and soft tissue volumetric changes after a
BSSO mandibular advancement. Int J Oral Maxillofac Surg. 2012;41:1137–
45.
160. Djordjevic J, Lewis BM, Donaghy CE, Zhurov AI, Knox J, Hunter L,
Richmond S. Facial shape and asymmetry in 5-year-old children with
repaired unilateral cleft lip and/or palate: an exploratory study using laser
scanning. Eur J Orthod. 2014;36(5).
161. Baik H-S, Kim S-Y. Facial soft-tissue changes in skeletal Class III
orthognathic surgery patients analyzed with 3-dimensional laser scanning.
Am J Orthod Dentofac Orthop. 2010;138(2):167–78.
162. Kau CH, Zhurov A, Richmond S, Cronin A, Savio C, Mallorie C. Facial
templates: a new perspective in three dimensions. Orthod Craniofac Res.
2006;9(1):10–7.
163. How C, Richmond S, Savio C, Mallorie C. Measuring Adult Facial Morphology
in Three Dimensions. Angle Orthod. 2006;76(5):773–8.
164. Kau CH, Kamel SG, Wilson J, Wong ME. New method for analysis of facial
growth in a pediatric reconstructed mandible. Am J Orthod Dentofac
Orthop. 2011;139(4):e285–90.
165. Maal TJJ, van Loon B, Plooij JM, Rangel F, Ettema AM, Borstlap WA, Bergé
SJ. Registration of 3-dimensional facial photographs for clinical use. J oral
Maxillofac Surg. 2010;68(10):2391–401.
166. Popat H, Richmond S, Playle R, Marshall D, Rosin P, Cosker D. Threedimensional motion analysis – an exploratory study . Part 1: Assessment of
facial movement. Orthod Craniofac Res. 2008;11(4):216–23.
167. Kau CH, Richmond S, Zhurov A, Ovsenik M, Tawfik W, Borbely P, English
JD. Use of 3-dimensional surface acquisition to study facial morphology in
5 populations. Am J Orthod Dentofac Orthop. 2010;137(4suppl):S56.e1–9;
discussion S56–7.
168. Nada R, Loon B Van, Maal T. Three-dimensional evaluation of soft tissue
changes in the orofacial region after tooth-borne and bone-borne surgically
assisted rapid maxillary expansion. Clin oral investig. 2013;17(9):2017–24.
169. Naudi KB, Benramadan R, Brocklebank L, Ju X, Khambay B, Ayoub A. The
virtual human face: superimposing the simultaneously captured 3D
photorealistic skin surface of the face on the untextured skin image of the
CBCT scan. Int J Oral Maxillofac Surg. 2013;42(3):393–400.
Anas Almukhtar 2016
246
Chapter Six
References
170. Marchetti C, Bianchi a, Muyldermans L, Di Martino M, Lancellotti L, Sarti A.
Validation of new soft tissue software in orthognathic surgery planning. Int
J Oral Maxillofac Surg. 2011;40(1):26–32.
171. Verhoeven TJ, Coppen C, Barkhuysen R, Bronkhorst EM, Merkx MA, Bergé
SJ, Maal TJ. Three dimensional evaluation of facial asymmetry after
mandibular reconstruction: validation of a new method using
stereophotogrammetry. Int J Oral Maxillofac Surg. 2013;42(1):19–25.
172. Maal TJJ, Verhamme LM, Loon B Van, Plooij JM, Rangel FA, Kho A.
Variation of the face in rest using 3D stereophotogrammetry. Int J Oral
Maxillofac Surg. 2011;40(11):1252–7.
173. Nada RM, van Loon B, Maal TJ, Bergé SJ, Mostafa YA, Kuijpers-Jagtman
AM, Schols JG. Three-dimensional evaluation of soft tissue changes in the
orofacial region after tooth-borne and bone-borne surgically assisted rapid
maxillary expansion. Clin Oral Investig. 2013;17(9):2017–24.
174. Miller L, Morris DO, Berry E. Visualizing three-dimensional facial soft tissue
changes following orthognathic surgery. Eur J Orthod. 2007;29(1):14–20.
175. Claes P, Walters M, Vandermeulen D, Clement JG. Spatially-dense 3D
facial asymmetry assessment in both typical and disordered growth. J
Anat. 2011;219(4):444–55.
176. Mao Z, Ju X, Siebert JP, Cockshott WP, Ayoub A. Constructing dense
correspondences for the analysis of 3D facial morphology. Pattern Recognit
Lett. 2006;27(6):597–608.
177. Kau CH, Richmond S. Three-dimensional imaging for orthodontics and
orthognathic surgery. London. Wiely-Blackwell; 2010. pp 28-44.
178. Qureshi F. Constructing Anatomically Accurate Face Models using
Computed Tomography and Cyberware data. 2000. p. 11, 47, 63. Available
from: http://www.cs.ucla.edu/~dt/theses/qureshi-ms-thesis.pdf
179. Goodall C. Procrustes Methods in the Statistical Analysis of Shape. J R Stat
Soc Ser B. 1991;53(2):285–339.
180. Al-Hiyali A, Ayoub A, Ju X, Almuzian M, Al-Anezi T. The Impact of
Orthognathic Surgery on Facial Expressions. J oral Maxillofac Surg. 2015;In
Press.
181. Higgins JE. Curve Extraction and Facial Analysis Using Statistical
Techniques. PhD thesis submited to the University of Glasgow 2009. p.
100–3.
182. Chau H, Dasgupta R, Sauret V, Kenyon G. Use of an optical surface scanner
in assessment of outcome following rhinoplasty surgery. J Laryngol Otol.
2008;122(9):972–7.
Anas Almukhtar 2016
247
Chapter Six
References
183. Sforza C, Grandi G, Pisoni L, Di Blasio C, Gandolfini M, Ferrario VF. Soft
tissue facial morphometry in subjects with Moebius syndrome. Eur J Oral
Sci. 2009;117(6):695–703.
184. Sforza C, Grandi G, Binelli M, Dolci C, Menezes M De, Ferrario VF. Age- and
sex-related changes in three-dimensional lip morphology. Forensic Sci Int.
2010;200(1-3):182e1–183e7.
185. Bell A, Lo TW, Brown D, Bowman AW, Siebert JP, Simmons DR, Millett DT,
Ayoub AF. Three-dimensional assessment of facial appearance following
surgical repair of unilateral cleft lip and palate. Cleft palate-craniofacial
J. 2014;51(4):462–71.
186. Joss CU, Joss-Vassalli IM, Bergé SJ, Kuijpers-Jagtman AM. Soft tissue
profile changes after bilateral sagittal split osteotomy for mandibular
setback: a systematic review. J oral Maxillofac Surg. 2010;68(11):2792–
801.
187. Louis PJ, Austin RB, Waite PD, Mathews CS. Soft tissue changes of the
upper lip associated with maxillary advancement in obstructive sleep
apnea patients. J oral Maxillofac Surg. 2001;59(2):151–6.
188. Chew MT. Soft and Hard Tissue Changes after Bimaxillary Surgery in
Chinese Class III Patients. Angle Orthod. 2005;75(6):1–5.
189. Sinthanayothin C. Computerized Cephalometric Line Tracing Technique on
X-ray Images.In: 13th International Conference on Biomedical Engineering.
Springer Berlin Heidelberg, 2009. ICBME proceedings. 2008. pp. 265–9.
190. Benson PE, Richmond S. A critical appraisal of measurement of the soft
tissue outline using photographs and video. Eur J Orthod. 1997;19(4):397409.
191. Sarver DM, Weissman SM. longterm soft tissue responce to maxillary
superior reositioning. Angle Orthod. 1991;61(4):267–76.
192. Mansour S, Burstone C, Legan H. An evaluation of soft-tissue changes
resulting from Le Fort I maxillary surgery. Am J Orthod. 1983;84(1):37–47.
193. Bell WH, Jacobs JD. Surgical-Orthodontic Correction of Maxillary Retrusion
by Le Fort I Osteotorny and Proplast. J Maxillofac Surg. 1980;8(2):84–94.
194. Freihofer H. Changes in nasal profile after maxillary advancement in cleft
and non-cleft patients. J Maxillofac Surg. 1977;5(1):20–7.
195. Wen-Ching Ko E, Figueroa AA, Polley JW. Soft tissue profile changes after
maxillary advancement with distraction osteogenesis by use of a rigid
external distraction device: a 1-year follow-up. J Oral Maxillofac Surg.
2000;58(9):959–69.
196. Stella JP, Streater MR, Epker BN, Sinn DP. Predictability of Upper Lip Soft
Tissue Changes With Maxillary Advancement. J Oral Maxillofac Surg.
1989;47(7):697–703.
Anas Almukhtar 2016
248
Chapter Six
References
197. Schendel SA, Eisenfeld JH, Bell WH, Epker BN. Superior repositioning of
the maxilla: Stability and soft tissue osseous relations. Am J Orthod.
1976;70(6):663–74.
198. Williams R. The diagnostic line. Am J Orthod. 1969;55(5):458–76.
199. Teuscher U, Sailer HF. Stability of Le Fort I Osteotomy in Class Ill Cases
with Retropositioned Maxillae. J cranio-maxillo-facial Surg. 1982;10(2):80–
3.
200. Engel GA, Quan RE, Chaconas SJ. Soft-tissue change as a result of maxillary
surgery. A preliminary study. Am J Orthod. 1979 Mar;75(3):291-300.
201. Carlotti AE, Aschaffenburg PH, Schendel S A. Facial changes associated
with surgical advancement of the lip and maxilla. J Oral Maxillofac Surg.
1986;44(8):593–6.
202. Freinhofer HPM. The Lip Profile after Correction of Retromaxillism in Cleft
and Non-Cleft Patients. J Maxillofac Surg. 1976;4(3):136-41.
203. Hack G, Otterloo J Van, Nanda R. Long-term stability and prediction of
soft tissue changes after Le Fort I surgery. Am J Orthod Dentofac Orthop.
1993;104(6):544–55.
204. Bergman RT. Cephalometric soft tissue facial analysis. Am J Orthod
Dentofac Orthop. 1999;116(4):373–89.
205. Smith J, Thomas P, Proffit W. A comparison of current prediction imaging
programs. Am J Orthod Dentofac Orthop. 2004;125(5):527–36.
206. Gallagher DM, Bell WH, Storum KA. Soft tissue changes associated with
advancement genioplasty performed concomitantly with superior
repositioning of the maxilla. J Oral Maxillofac Surg. 1984;42(4):238–42.
207. Donatsky O, Hillerup S. Computerized cephalometric orthognathic surgical
simulation, prediction and postoperative evaluation of precision. Int J Oral
Maxillofac Surg. 1992;21(4):199–203.
208. Clemente-Panichella D, Suzuki S, Cisneros G. Soft to hard tissue movement
ratios: orthognathic surgery in a Hispanic population. Int J Adult Orthodon
Orthognath Surg. 2000;15(4):255–64.
209. Bell WH, Dann JJ. Correction of dentofacial deformities by surgery in the
anterior part of the Jaws. Am J Orthod. 1973;64(2):162–87.
210. Rosen HM. Lip-Nasal Aesthetics Following Le Fort I Osteotomy. Plast
Reconstr Surg. 1988;81(2):171–9.
211. Dann J, Fonseca R, Bell W. Soft tissue changes associated with total
maxillary advancement a preliminary study. J Oral Surg. 1976;34(1):19-23.
212. Burstone CJ. Lip posture and its significance in treatment planning. Am J
Orthod. 1967;53(4):262-84.
Anas Almukhtar 2016
249
Chapter Six
References
213. Jakobsone G, Stenvik A, Espeland L. Importance of the vertical incisor
relationship in the prediction of the soft tissue profile after Class III
bimaxillary surgery. Angle Orthod. 2012;82(3):441-447.
214. Mobarak K, Krogstad O, Espeland L, Lyberg T. Factors influencing the
predictability of soft tissue profile changes following mandibular setback
surgery. Angle Orthod. 2001;71(3):216–27.
215. Aulsebrook WA, Becker PJ, Işcan MY. Facial soft-tissue thicknesses in the
adult male Zulu. Forensic Sci Int. 1996;79(2):83–102.
216. Schweckendiek W. Nasal abnormalities in facial clefts. J Maxillofac Surg.
1976;4(3):141–9.
217. Hui E, Hägg E, Tideman H. Soft tissue changes following maxillary
osteotomies in cleft lip and palate and non-cleft patients. J CranioMaxillofacial Surg. 1994;22(3):182–6.
218. Knowles C. Changes in the profile following surgical reduction of
mandibular prognathism. Br J Plast Surg. 1965;18(4):432–4.
219. Aaronson S. A cephalometric investigation of the surgical correction of
mandibular prognathism. Angle Orthod. 1967;37(4):251–60.
220. Lin S, Kerr W. Soft and hard tissue changes in Class III patients treated by
bimaxillary surgery. Eur J Orthod. 1998;20(1):25–33.
221. Hu J, Wang D, Luo S, Chen Y. Differences in soft tissue profile changes
following mandibular setback in Chinese men and women. J oral Maxillofac
Surg. 1999;57(10):1182–6.
222. Mobarak KA, Espeland L, Krogstad O, Lyberg T. Soft tissue profile changes
following mandibular advancement surgery: predictability and long-term
outcome. Am J Orthod Dentofac Orthop. 2001;119(4):353–67.
223. Joss C, Joss IM, Kiliarids S, Kuijpers-Jagtman AM. Soft Tissue Profile After
Bilateral Sagittal Split Osteotomy for Mandibular Advancement : A
Systematic Review. J oral Maxillofac Surg. 2010;68(6):1260–9.
224. Jones RM, Khambay BS, McHugh S, Ayoub AF. The validity of a computerassisted 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. 2007;36(10):900–8.
225. Cevidanes L, Tucker S, Styner M, Kim H, Chapuis J, Reyes M, Proffit
W, Turvey T, Jaskolka M. Three-dimensional Surgical Simulation. Am J
Orthod Dentofac Orthop. 2011;138(3):316–71.
226. Schendel S a., Lane C. 3D Orthognathic Surgery Simulation Using Image
Fusion. Semin Orthod. 2009;15(1):48–56.
227. Mollemans W, Schutyser F, Nadjmi N, Maes F, Suetens P. Predicting soft
tissue deformations for a maxillofacial surgery planning system: from
Anas Almukhtar 2016
250
Chapter Six
References
computational strategies to a complete clinical validation. Med Image
Anal. 2007;11(3):282–301.
228. Schutyser F, Cleynenbreugel J Van, Ferrant M, Schoenaers J, Suetens P.
Image-based 3D planning of maxillofacial distraction procedures including
soft tissue implications. In: Medical Image Computing and ComputerAssisted Intervention – MICCAI 2000;104:999–1007.
229. Xia J, Samman N, Yeung RW, Shen SG, Wang D, Ip HH, Tideman H. Threedimensional virtual reality surgical planning and simulation workbench for
orthognathic surgery. Int J Adult Orthodon Orthognath Surg.
2000;15(4):265–82.
230. Keeve E, Girod S, Kikinis R, Girod B. Deformable modeling of facial tissue
for craniofacial surgery simulation. Comput Aided Surg. 1998;3(5):228–38.
231. Chabanas M, Luboz V, Payan Y. Patient specific finite element model of
the face soft tissues for computer-assisted maxillofacial surgery. Med
Image Anal. 2003;7(2):131–51
232. Sarti A, Gori R, Lamberti C. A physically based model to simulate maxillofacial surgery from 3D CT images. Futur Gener Comput Syst.
1999;15(2):217–21.
233. Teschner M, Girod S, Girod B. Direct Computation of Nonlinear Soft-Tissue
Deformation. In: Vision, Modeling, and Visualization VMV’00. 2000.
234. Meehan M, Teschner M, Girod S. Three‐dimensional simulation and
prediction of craniofacial surgery. Orthod Craniofacial Res. 2003;6
(supl.1):102–7.
235. Cotin S, Delingette H, Ayache N. Real-Time Elastic Deformations of Soft
Tissues for Surgery Simulation. IEEE Trans Vis Comput Graph. 1999;5(1):62–
73.
236. Ullah R. The validity of 3dMD Vultus in predicting soft tissue morphology
following orthognathic surgery. M.phil thesis submitted to the University of
Birmingham. 2014. p. 96.
237. Schendel S a, Jacobson R, Khalessi S. Three-Dimensional Facial Simulation
in Orthognathic Surgery: Is It Accurate? J oral Maxillofac Surg.
2013;71(8):1406–14.
238. Park S-B, Kim Y-I, Hwang D-S, Lee J-Y. Midfacial soft-tissue changes after
mandibular setback surgery with or without paranasal augmentation: conebeam computed tomography (CBCT) volume superimposition. J craniomaxillo-facial Surg. 2013;41(2):119–23.
239. Cook J, Chandran V, Sridharan S, Fookes C. Face recognition from 3D data
using Iterative Closest Point algorithm and Gaussian mixture models. in: 3D
Data Processing, Visualization and Transmission(3DPVT) 2004. Proceedings
of the 2nd International Symposium 2004. pp.502-509.
Anas Almukhtar 2016
251
Chapter Six
References
240. Vezzetti E, Marcolin F. 3D human face description: landmarks measures
and geometrical features. Image Vis Comput. 2012;30(10):968–712.
241. De Assis Ribeiro Carvalho F, Cevidanes LHS, da Motta ATS, de Oliveira
Almeida MA, Phillips C. Three-dimensional assessment of mandibular
advancement 1 year after surgery. Am J Orthod Dentofac Orthop.
2010;137(4):S53.e1–S53.e12.
242. Heymann GC, Cevidanes L, Cornelis M, Clerck HJ De, Tulloch JFC. Threedimensional analysis of maxillary protraction with intermaxillary elastics to
miniplates. Am J Orthod Dentofac Orthop. 2010;137(2):274–84.
243. Khambay B, Ullah R. Current methods of assessing the accuracy of threedimensional soft tissue facial predictions: technical and clinical
considerations. Int J Oral Maxillofac Surg. 2015;44(1):132–8.
244. Zhili Mao, Xiangyang Ju, J. Paul Siebert, W. Paul Cockshott AA.
Constructing dense correspondences for the analysis of 3D facial
morphology. Pattern Recognit Lett. 2006;27(6):597–608.
245. Aynechi N, Larson BE, Leon-Salazar V, Beiraghi S. Accuracy and precision
of a 3D anthropometric facial analysis with and without landmark labeling
before image acquisition. Angle Orthod. 2011;81(2):245–52.
246. Almukhtar A, Khambay B, Ayoub A, Ju X, Al-Hiyali A, Macdonald J, Jabar N
and Goto T. “Direct DICOM Slice Landmarking” A Novel Research
Technique to Quantify Skeletal Changes in Orthognathic Surgery. PLoS
One. 2015;10(8):e0131540.
247. Hajeer MY, Ayoub AF, Millett DT. Three-dimensional assessment of facial
soft-tissue asymmetry before and after orthognathic surgery. Br J Oral
Maxillofac Surg. 2004;42(5):396–404.
248. Kau CH, Richmond S, Palomo JM, Hans MG. Three-dimensional cone beam
computerized tomography in orthodontics. J Orthod. 2005;32(4):282–93.
249. Kamburoğlu K, Kolsuz E, Kurt H, Kılıç C. Accuracy of CBCT measurements
of a human skull. J Digit Imaging. 2011;24(5):787–93. Available from:
250. Brown A a, Scarfe WC, Scheetz JP, Silveira AM, Farman AG. Linear
accuracy of cone beam CT derived 3D images. Angle Orthod.
2009;79(1):150–7.
251. Ghoneima A, Kula K. Accuracy and reliability of cone-beam computed
tomography for airway volume analysis. Eur J Orthod. 2013;35(2):256–61.
252. Farman AG, Scarfe WC. Development of imaging selection criteria and
procedures should precede cephalometric assessment with cone-beam
computed tomography. Am J Orthod Dentofac Orthop. 2006;130(2):257–65.
253. Moerenhout B a MML, Gelaude F, Swennen GRJ, Casselman JW, Van Der
Sloten J, Mommaerts MY. Accuracy and repeatability of cone-beam
computed tomography (CBCT) measurements used in the determination of
Anas Almukhtar 2016
252
Chapter Six
References
facial indices in the laboratory setup. J cranio-maxillo-facial Surg.
2009;37(1):18–23.
254. Jokic D, Uglešic V, Macan D, Knezevic P. Soft tissue changes after
mandibular setback and bimaxillary surgery in Class III patients. Angle
Orthod. 2013;83(5):817–23.
255. Marchetti C, Bianchi A, Bassi M, Gori R, Lamberti C, Sarti A. Mathematical
Modeling and Numerical Simulation in Maxillo-Facial Virtual Surgery (VISU).
J Craniofac Surg. 2006;17(4):661–7.
256. Tucker S, Cevidanes LHS, Styner M, Kim H, Reyes M, Proffit W, Turvey T.
Comparison of actual surgical outcomes and 3-dimensional surgical
simulations. J oral Maxillofac Surg. 2010;68(10):2412–21.
257. Beldie L, Walker B, Lu Y, Richmond S, Middleton J. Finite element
modelling of maxillofacial surgery and facial expressions—a preliminary
study. Int J Med Robot Comput Assist Surg. 2010;6(4):422–30.
258. Deuflhard P, Weiser M, Zachow S. Mathematics in Facial Surgery. Notes
AMS. 2006;53(9):1012–6.
259. Soncul M, Bamber MA. Evaluation of facial soft tissue changes with optical
surface scan after surgical correction of Class III deformities. J Oral
Maxillofac Surg. 2004;62(11):1331–40.
260. Luffingham J, Campbell H. The need for orthodontic treatment. A pilot
survey of 14 year old school children in Paisley, Scotland. Trans Eur Orthod
Soc. 1974;259–67.
261. Kim J, Viana MAG, Graber TM. The effectiveness of protraction face mask
therapy: A meta-analysis. Am J Orthod Dentofac Orthop. 1999;115(6):675–
85.
262. Proffit WR and Fields HW. Contemporary Orthodontics. St. Louis: MosbyYear Book, Inc., 2013: 225-264.
263. Kau CH, Cronin a, Durning P, Zhurov a I, Sandham a, Richmond S. A new
method for the 3D measurement of postoperative swelling following
orthognathic surgery. Orthod Craniofac Res. 2006;9(1):31–7.
264. Dolce C, Hatch JP, Sickels JE Van, Rugh JD. Five-year outcome and
predictability of soft tissue profiles when wire or rigid fixation is used in
mandibular advancement surgery. Am J Orthod Dentofac Orthop.
2003;124(3):249–56.
265. Kor HS, Yang HJ, Hwang SJ. Relapse of skeletal class III with anterior open
bite after bimaxillary orthognathic surgery depending on maxillary
posterior impaction and mandibular counterclockwise rotation. J CranioMaxillofacial Surg. 2013;42(5):e230–8.
Anas Almukhtar 2016
253
Chapter Six
References
266. Kim Y-K, Kim Y-J, Yun P-Y, Kim J-W. Evaluation of skeletal and surgical
factors related to relapse of mandibular setback surgery using the
bioabsorbable plate. J cranio-maxillo-facial Surg. 2009;37(2):63–8.
267. Blomqvist JE, Ahlborg G, Isaksson S, Svartz K. A comparison of skeletal
stability after mandibular advancement and use of two rigid internal
fixation techniques. J oral Maxillofac Surg. 1997;55(6):568–74; discussion
574–5.
268. Ko EW-C, Lin SC, Chen YR, Huang CS. Skeletal and dental variables related
to the stability of orthognathic surgery in skeletal Class III malocclusion
with a surgery-first approach. J Oral Maxillofac Surg. 2013;71(5):e215–23.
269. Jung H-D, Jung Y-S, Kim SY, Kim DW, Park H-S. Postoperative stability
following bilateral intraoral vertical ramus osteotomy based on amount of
setback. Br J Oral Maxillofac Surg. 2013;51(8):822–6.
270. Liang X, Lambrichts I, Sun Y, Denis K, Hassan B, Li L. A comparative
evaluation of Cone Beam Computed Tomography (CBCT) and Multi-Slice CT
(MSCT). Part II: On 3D model accuracy. Eur J Radiol. 2010;75(2):270–4.
271. Olmez H, Gorgulu S, Akin E, Bengi AO, Tekdemir I, Ors F. Measurement
accuracy of a computer-assisted three-dimensional analysis and a
conventional two-dimensional method. Angle Orthod. 2011;81(3):375–82.
272. Chang Z-C, Hu F-C, Lai E, Yao C-C, Chen M-H, Chen Y-J. Landmark
identification errors on cone-beam computed tomography-derived
cephalograms and conventional digital cephalograms. Am J Orthod
Dentofac Orthop. 2011;140(6):e289–97.
273. Gribel BF, Gribel MN, Frazäo DC, McNamara J a, Manzi FR. Accuracy and
reliability of craniometric measurements on lateral cephalometry and 3D
measurements on CBCT scans. Angle Orthod. 2011;81(1):26–35.
274. Aydil B, Özer N, Marşan G. Facial soft tissue changes after maxillary
impaction and mandibular advancement in high angle class II cases. Int J
Med Sci. 2012;9(4):316–21.
275. Ghang M-H, Kim H-M, You J-Y, Kim B-H, Choi J-P, Kim S-H, Choung PH..
Three-dimensional mandibular change after sagittal split ramus osteotomy
with a semirigid sliding plate system for fixation of a mandibular setback
surgery. Oral Surg Oral Med Oral Pathol Oral Radiol. 2013;115(2):157–66.
276. Kim B-R, Oh K-M, Cevidanes LHS, Park J-E, Sim H-S, Seo S-K, Reyes M, Kim
YJ, Park YH. Analysis of 3D soft tissue changes after 1- and 2-jaw
orthognathic surgery in mandibular prognathism patients. J oral Maxillofac
Surg. 2013;71(1):151–61.
277. Walters M, Claes P, Kakulas E, Clement J. Robust and regional 3D facial
asymmetry assessment in hemimandibular hyperplasia and hemimandibular
elongation anomalies. Int J Oral Maxillofac Surg. International Association
of Oral and Maxillofacial Surgery; 2013;42(1):36–42.
Anas Almukhtar 2016
254
Chapter Six
References
278. Mao Z, Siebert J, Cockshott WP, Ayoub A. Constructing dense
correspondences to analyze 3d facial change. In :The 17th International
Conference on Pattern Recognition (ICPR’04). Proceedings. 2004;3:1–5.
279. Oh K-M, Seo S-K, Park J-E, Sim H-S, Cevidanes LHS, Kim Y-JR, Park YH..
Post-operative soft tissue changes in patients with mandibular prognathism
after bimaxillary surgery. J cranio-maxillo-facial Surg. 2013;41(3):204–11.
280. Nkenke E, Vairaktaris E, Kramer M, Schlegel A, Holst A, Hirschfelder
U, Wiltfang J, Neukam FW, Stamminger M. Three-dimensional analysis of
changes of the malar-midfacial region after Le Fort I osteotomy and
maxillary advancement. Oral Maxillofac Surg. 2008;12(1):5–12.
281. Vasudavan S, Jayaratne Y, Padwa B. Nasolabial soft tissue changes after Le
Fort I advancement. J Oral Maxillofac Surg. 2012;70(4):e270–7.
282. Metzler P, Geiger EJ, Chang CC, Sirisoontorn I, Steinbacher DM.
Assessment of three-dimensional nasolabial response to Le Fort I
advancement. J Plast Reconstr Aesthet Surg. 2014;67(6):756–63.
283. Ko EW-C, Figueroa A, Polley JW. Soft tissue profile changes after maxillary
advancement with distraction osteogenesis by use of a rigid external
distraction device: A 1-year follow-up. J Oral Maxillofac Surg.
2000;58(9):959–69.
284. Mansour S, Burstone C, Legan H. An evaluation of soft-tissue changes
resulting from Le Fort I maxillary surgery. Am J Orthod. 1983;84(1):37–47.
285. Dantas WRM, da Silveira MMF, do Egito Vasconcelos BC, Porto GG.
Evaluation of the nasal shape after orthognathic surgery. Braz J
Otorhinolaryngol. 2015;81(1):19–23.
286. van Loon B, van Heerbeek N, Bierenbroodspot F, Verhamme L, Xi T, de
Koning MJ, Ingels KJ, Bergé SJ, Maal TJ. Three-dimensional changes in
nose and upper lip volume after orthognathic surgery. Int J Oral Maxillofac
Surg. 2015;44(1):83–9.
287. Park S-B, Kim Y-I, Hwang D-S, Lee J-Y. Midfacial soft-tissue changes after
mandibular setback surgery with or without paranasal augmentation: conebeam computed tomography (CBCT) volume superimposition. J craniomaxillo-facial Surg. 2013;41(2):119–23.
288. Almeida RC, Cevidanes LHS, Carvalho F a R, Motta a T, Almeida M a O,
Styner M, Turvey T, Proffit WR, Phillips C. Soft tissue response to
mandibular advancement using 3D CBCT scanning. Int J Oral Maxillofac
Surg. 2011;40(4):353–9.
289. McCollum AGH, Gardener GJM, Evans WG, Becker PJ. Soft-Tissue Changes
Related to Mandibular Advancement Surgery. Semin Orthod.
2009;15(3):161–71.
Anas Almukhtar 2016
255
Chapter Six
References
290. Iizuka T, Eggensperger N, Smolka W, Thüer U. Analysis of soft tissue profile
changes after mandibular advancement surgery. Oral Surgery, Oral Med
Oral Pathol Oral Radiol Endodontology. 2004;98(1):16–22.
291. Raschke GF, Rieger UM, Bader R. Soft tissue outcome after mandibular
advancement — an anthropometric evaluation of 171 consecutive patients.
Clin Oral Invest. 2013;17:1415–23.
292. Dicker GJ, Koolstra JH, Castelijns J a, Van Schijndel R a, Tuinzing DB.
Positional changes of the masseter and medial pterygoid muscles after
surgical mandibular advancement procedures: an MRI study. Int J Oral
Maxillofac Surg. 2012;41(8):922–9.
293. Conley RS, Boyd SB. Facial soft tissue changes following maxillomandibular
advancement for treatment of obstructive sleep apnea. J oral Maxillofac
Surg. 2007;65(7):1332–40.
294. Claes P, Walters M, Vandermeulen D, Clement JG. Spatially-dense 3D
facial asymmetry assessment in both typical and disordered growth. J
Anat. 2011;219(4):444–55.
295. Gerbino G, Bianchi FA, Verzé L, Ramieri G. Soft tissue changes after
maxillo-mandibular advancement in OSAS patients: a three-dimensional
study. J cranio-maxillo-facial Surg. 2014;42(1):66–72.
296. Bianchi A, Muyldermans L, Di Martino M, Lancellotti L, Amadori S, Sarti A.
Facial soft tissue esthetic predictions: validation in craniomaxillofacial
surgery with cone beam computed tomography data. J oral Maxillofac
Surg. 2010;68(7):1471–9.
297. Chabanas M, Payan Y. A 3D Finite Element model of the face for simulation
in plastic and maxillo-facial surgery. In: The Medical Image Computing and
Computer-Assisted Intervention – MICCAI. 2000. Proceedings of the Third
International Conference Pittsburgh, PA, USA, October 11-14, 2000. pp.
1068–75.
298. Koch RM, Gross MH, Carls FR, Von BUren DF, Fankhauser G, Parish YIH.
Simulating Facial Surgery Using Finite Element Models. In: SIGGRAPH '96.
Proceedings of the 23rd annual conference on Computer graphics and
interactive techniques, 1996. pp. 421-428.
299. Keeve E, Girod S, Girod B. Craniofacial Surgery Simulation. in: 4th
International Conference, VBC '96, Hamburg, Germany, September 22 - 25,
1996. Proceedings of the Visualization in Biomedical Computing. 1996. pp.
541–6.
300. Zachow S, Gladiline E, Hege H, Deuflhard P. Finite-Element Simulation of
Soft Tissue Deformation. In: Computer Assissted Radiology and Surgery
(CARS). 2000. Proceedings of the 14th International Congress & Exhibition,
San Francisco, USA. 2000. pp. 23–8.
Anas Almukhtar 2016
256
Chapter Six
References
301. Keeve E, Girod S, Pfeifle P, Girod B. Anatomy-Based Facial Tissue Modeling
Using the Finite Element Method. At: The 7th Visualization Conference
proceedings (VIS’96). 1996; pp. 21–8.
302. Gladilin, E.; Zachow, S.; Deuflhard, P.; Hege, H.-C. A biomechanical model
for soft tissue simulation in craniofacial surgery. In: Medical Imaging and
Augmented Reality, 2001. Proceedings. International Workshop. 2001.
pp.137-141,
303. Ulusoy I, Akagunduz E, Sabuncuoglu F, Gorgulu S, Ucok O. Use of the
dynamic volume spline method to predict facial soft tissue changes
associated with orthognathic surgery. Oral Surg Oral Med Oral Pathol Oral
Radiol Endod. 2010;110(5):e17–23.
304. Delingette H. Toward realistic soft-tissue modeling in medical simulation.
in: Proceedings of the IEEE. 1998;86(3):512-523.
305. Platt JC, Barr AC. Constraint Methods for Flexible Models. Comput Graph
(ACM). 1988;22(4):279–88.
306. Black J, Hasting G. Handbook of Biomedical properties. London: Chapman
and Hall; 1998.PP 490-500
307. Jabar N, Robinson W, Goto TK, Khambay BS. The validity of using surface
meshes for evaluation of three-dimensional maxillary and mandibular
surgical changes. Int J Oral Maxillofac Surg. 2015;44(7):914–20.
308. Terzic A, Combescure C, Scolozzi P. Accuracy of computational soft tissue
predictions in orthognathic surgery from three-dimensional photographs 6
months after completion of surgery: a preliminary study of 13 patients.
Aesthetic Plast Surg. 2014;38(1):184–91.
309. Ullah R, Turner J, Khambay B. The Validity of 3dMD Vultus in Predicting
Soft Tissue Morphology Following Orthognathic Surgery. Br J Oral
Maxillofac Surg. 2014;52(8):e58.
310. Van Hemelen G, Van Genechten M, Renier L, Desmedt M, Verbruggen E,
Nadjmi N. Three-dimensional virtual planning in orthognathic surgery
enhances the accuracy of soft tissue prediction. J cranio-maxillo-facial
Surg. 2015;43(6):918–25.
311. Liebregts J, Xi T, Timmermans M, De Koning M, Bergé S, Hoppenreijs T.
Accuracy of three-dimensional soft tissue simulation in bimaxillary
osteotomies. J cranio-maxillo-facial Surg. 2015;43(3):329–35.
312. Liebregts J, Timmermans M, De Koning M, Bergé S, Maal T. Threedimensional facial simulation in bilateral sagittal split osteotomy: a
validation study of 100 patients. J Oral Maxillofac Surg. 2015;73(5):961-70.
Anas Almukhtar 2016
257
7
C
Appendices
ontents
7.1
APPENDIX 1 PRESENTATIONS AND AWARDS............................................................................ 259
7.1.1
VERBAL PRESENTATIONS ......................................................................................................... 259
7.1.2
POSTER PRESENTATIONS ......................................................................................................... 260
7.1.3
AWARDS AND RECOGNITIONS .................................................................................................. 261
7.2
APPENDIX 2 PUBLICATIONS ................................................................................................ 262
7.2.1
PUBLISHED JOURNAL ARTICLES ................................................................................................. 263
7.2.2
ACCEPTED FOR PUBLICATION ................................................................................................... 263
Chapter Seven
7.1 Appendix 1 Presentations and awards
Appendices
7.1.1 Verbal presentations
1. “Internal DICOM slice landmarking” a novel method to quantify skeletal
movement following orthognathic surgery. 3D Bologna international conference in
Italy 2014 (Bologna, Italy).
2. Computer assisted surgical planning, current practice at the Glasgow dental
hospital and school. Collaboration visit to the University of (Lille, France), funded
by the EuroCleft scientific Foundation 2014 (ESF).
3. Comparison of the Accuracy of Voxel Based Registration and Surface Based
Registration for 3D Assessment of Surgical Change following Orthognathic
Surgery. 3D User meeting 2013 (London, UK).
4. Comparison between different 3D image registration methods used in the analysis
orthognathic of facial changes following orthognathic surgery. Graduate Research
competition 2013 (Glasgow, UK).
5. Computer assisted 3D planning in orthognathic surgery. A four Hours of theory
and Hands-On training course for Post Graduate students and staff members on 3D
Planning in orthognathic surgery 2013-2015 (Glasgow, UK).
6. Regular seminars presentation. Ppost-Graduate seminars and journal club
meetings 2011-2015 (Glasgow, UK).
7. Effect of Le Fort I osteotomy on Alar base width. Research audit Verbal
presentation 2012 (Glasgow, UK).
Anas Almukhtar 2016
259
Chapter Seven
7.1.2 Poster presentations
Appendices
1. State-of-the art analysis of soft tissue changes in response to Le Fort I maxillary
advancement. The First HCED Iraq initiative meeting 2015; London, UK.
2. “Internal DICOM slice landmarking” a novel method to quantify skeletal
movement following orthognathic surgery. The annual International British
Craniofacial society meeting 2014; Oxford, UK.
3. Comparison between different 3D image registration methods used in the analysis
orthognathic of facial changes following orthognathic surgery The international
Brand-Spasel Symposium 2014; Basel, Switzerland.
4. “Internal DICOM slice landmarking” a novel method to quantify skeletal
movement following orthognathic surgery. The international Brand-Spasel
Symposium 2014; Basel, Switzerland.
5. Effects of Le Fort I Osteotomy on the Nasopharyngeal Airway—6-Month FollowUp. The annual BOS conference 2014; Edinburgh, UK.
Anas Almukhtar 2016
260
Chapter Seven
7.1.3 Awards and recognitions
Anas Almukhtar 2016
Appendices
261
Chapter Seven
Anas Almukhtar 2016
Appendices
262
Chapter Seven
Appendices
7.2 Appendix 2 Publications
7.2.1 Published Journal articles
Almukhtar A, Ju X, Khambay B, McDonald J, Ayoub A. Comparison of the
Accuracy of Voxel Based Registration and Surface Based Registration for 3D
Assessment of Surgical Change following Orthognathic Surgery. PLoS ONE. 2014;
9(4): e93402.
Anas Almukhtar, Balvinder Khambay , Ashraf Ayoub, Xiangyang Ju , Ali AlHiyali, James McDonald, Norhayati Jabar, and Tazuko Goto. "Direct DICOM slice
landmarking” A novel research technique to quantify skeletal changes in
orthognathic surgery. PLoS One. 2015;10(8):e0131540.
Mohammed Almuzian, Anas Almukhtar, Xiangyang Ju, Ali Al-Hiyali, Philip
Benington, Ashraf Ayoub. Effects of Le Fort I Osteotomy on the Nasopharyngeal
Airway—6-Month Follow-Up.J Oral Maxillofac Surg. 2016;74(2):380-91.
Mohammed Almuzian, Anas Almukhtar, Michael O’Neil, Philip Benington,
Thamer Al Anezi and Ashraf Ayoub. Innovation in prediction planning for anterior
open bite correction. Aust Orthod J. 2015; 31: 78–86.
7.2.2 Accepted for publication
Almukhtar A, Ju X, Khambay B, McDonald J, Ayoub A. State-of-the art analysis
of soft tissue changes in response to Le Fort I maxillary advancement. Br J Oral
Maxillofac Surg. 2016. XXXX.
Almuzian M, Ju X, Almukhtar A, Ayoub A, Al-Muzian L, McDonald JP. Does
rapid maxillary expansion affect nasopharyngeal airway? A prospective Cone
Beam Computerised Tomography (CBCT) based study. Surgeon. 2016. XXXX.
Khambay B, Cheungmy, Almukhtar A, Keeling A.J, tchsung, Ju X , McDonald
JP, Ayoub A. The accuracy of conformation of a generic surface mesh for the
analysis of facial soft tissue changes. PLoS ONE. 2016 XXXX.
Anas Almukhtar 2016
263