108 105 PB

ISSN 2219-6749
The Journal of
Forensic
Odonto-Stomatology
42, n. 21 -- Apr
Volume 36, 2024 2018
December
OFFICIAL PUBLICATION OF THE

INTERNATIONAL ORGANIZATION FOR
FORENSIC ODONTO-STOMATOLOGY
JFOS
ISSN: 2219-6749
THE JOURNAL OF FORENSIC ODONTO-STOMATOLOGY

Editor
Vilma Pinchi, DDS, PhD, Msc
Section of Forensic Medical Sciences
University of Florence (Italy)
Associate Editor
Douglas Sheasby (United Kingdom)
Francesco Pradella (Italy)
Ademir Franco (Brazil)
Editorial Board
Ashith B. Acharya (India)
Salem Seed Altalie Aldhuhoori (United Arab Emirates)
Ricardo Henrique Alver da Silva (Brazil)
Herman Bernitz (South Africa)
Hrvoje Brkic (Croatia)
Jelena Dumančić (Croatia)
Sigrid Kvaal (Norway)
Maria Gabriela Haye Biazevic (Brazil)
Ruediger Lessig (Germany)
Helen Liversidge (United Kingdom)
Anastasia Mitsea (Greece)
Phrabhakaran Nambiar (Malaysia)
Cristiana Palmela Pereira (Portugal)
Webmaster and Publishing Director

Stefano Garatti (Italy)
Editorial Staff
Luka Banjšak (Croatia)
Ilenia Bianchi (Italy)
Ana Margarida Peleias Rodrigues (Portugal)
The Journal of Forensic Odonto-Stomatology (JFOS)is the official journal of the
International Organization for Forensic Odonto-Stomatology (IOFOS) since 1983.
JFOS is an international journal dedicated to forensic odontology, legal odonto-
stomatology, dental anthropology, translational research in forensic sciences, litigation in
dentistry, dental ethics and law in the perspective of justice administration and legal
rights protection.
JFOS is open-access free of charge for readers and authors, published three times per
year (in April, August, and December). The quality of the publications is guaranteed by
the blind review procedure by internationally accredited peer reviewers.
The Journal of Forensic Odonto-Stomatology (JFOS) is referenced and fully indexed in
the largest databases such as PubMed-MEDLINE, PubMed Central, Scopus, and
EBSCO.
The Journal welcomes international material for publishing. We would also like to
encourage submissions from new authors, recognising the importance of first
publications. All papers should comply with the “Guidance to Authors” document
printed on the IOFOS.eu website.
Disclaimer: The views/opinions expressed in the publication are those of the authors
and not of the editorial board or the editor. The JFOS is not responsible for any
consequences arising from the use of the information contained in the published articles.
The Journal of Forensic Odonto-Stomatology
Vol 42 n. 1 Apr 2024
CONTENTS
CONTENT PAGE I
Authors Page
Original Articles
The role of orthodontics in children identification: a Aliki Rontogianni , 2

case report of two victims of mass disaster Anastasia Mitsea,
Kety Karayianni
Canine sexual dimorphism in crown and root Mohammad Tajik, 12

dimensions: a cone-beam computed tomographic Najmeh Movahhedian
study
Machine learning assisted 5-part tooth segmentation Rizky Merdietio Boedi, 22
method for CBCT-based dental age estimation in Simon Shepherd,
adults Fahmi Oscandar,
Ademir Franco,
Scheila M
Review
Artificial intelligence in age and sex determination Sraddha Singh, 30

using maxillofacial radiographs: a systematic Bhoopendra Singh,
review Shanu Kumar
Are computed tomography images of the mandible Luciana Munhoz, 38

useful in age and sex determination? A forensic Shunsuke Okada,
science meta-analysis Miki Hisatomi,
Yoshinobu Yanagi,
Emiko Saito Arita1,
Junich Asaumi
I
JFOS - Journal of Forensic Odonto-Stomatology Vol 42 n. 1 - Apr - 2024
The role of orthodontics in children identification:

a case report of two victims of mass disaster
Copyright © 2024 International Organization ABSTRACT

for Forensic Odonto-Stomatology - IOFOS
Introduction: Human Identification based on dental evidence
Aliki Rontogianni 1, Anastasia cannot be accomplished if antemortem dental records are
Mitsea2, Kety Karayianni2 unavailable or of poor quality. The involvement of the
orthodontist in mass disaster victim identification processes
1 Department of Orthodontics, School may be crucial in relation to the amount and quality of the
of Dentistry, National and Kapodistrian records which can be obtained before, during, and following
University of Athens, Greece. the treatment.
2Department of Oral Diagnosis and
Aim: The aim of the study is the description of the
Radiology, School of Dentistry,
contribution of the findings drawn from orthodontic records
National and Kapodistrian University
to the identification of victims of mass disasters who had
of Athens, Greece.
received an orthodontic treatment, through the presentation
of two cases. The first case involves the identification of a child
victim of a plane crash and the second case involves the
identification of two identical twin girls who died in a fire. In
Corresponding author:
both cases, the identification was based on the findings
[email protected]
obtained from the ante-mortem records provided by the
orthodontist.
Conclusions: The orthodontists apply customized orthodontic
appliances and keep a comprehensive file of images, casts,
The authors declare that they radiographs, and other records in their practice. As a result,
they can make a substantial contribution to the identification
have no conflict of interest.
of young people or even adult victims of mass disasters in any
case in which the authorities make a request.
INTRODUCTION
In recent decades mass disasters became a common global
issue of concern. A mass disaster or a mass fatality incident, is
KEYWORDS an event in which there is a critical relationship between the
number of casualties and resources, both human and material,
Mass disaster, available at the occurrence site, usually unpredictable and
Human identification, abrupt. 1 Environmental (such as tornadoes, hurricanes,
Orthodontic records earthquakes), transportation, industrial events, or terrorist acts,
might all result in mass fatalities .2
Depending on the location of the occurrence and the country/
countries of origin of the victims, the disaster might be
characterized as local, national, or international.2-4 Mass
catastrophes can also be classified as “closed” (an aircraft-related
accident) or “open” (an earthquake), based on whether the
J Forensic Odontostomatol number of victims and those involved are known details.2, 5- 7
2024. Apr;(42): 1-2:11 Human identification is the process to recognize an individual
ISSN :2219-6749
as a unique being. Forensic odontology is the branch of
dentistr y applied in criminal and civil law cases. The
DOI doi.org/10.5281/zenodo.11061609
participation of forensic odontology in the human identification
2
process is fundamental when human remains are open and national. The cases show the valuable
found, since the dental arches have particularities contribution of orthodontics in the identification
that make such a process possible.8 According to of children when casualties of mass fatality
the International Criminal Police Organization events.
(Interpol), given the individuality of the teeth and
other anatomical str uctures of the CASE I
dentomaxillofacial region, forensic odontology is On August 14, 2005, a Boeing 737-300 plane
one of the three primary human identifiers. crashed in a mountainous area of Attica, Greece.
These are classified as primary or secondary The plane took-off from Larnaca airport with
according to their ability to achieve an accurate final destination Prague, Czech Republic, and
identification. Primary methods include dental stop over in Athens airport, Greece. While
analysis, fingerprinting, and DNA analysis. approaching Athens airport the plane lost
Considering the accuracy and reliability provided communication with the air traffic control center
to the process, these are the methods of choice in and soon crashed. All the people on board, 115
cases of human identification (‘https:// passengers (adults and children) and 6 crew
www.interpol.int/en/How-we-work/Forensics/ members, lost their lives.
Disaster-Victim-Identification-DVI'), equally Depending on each victim's individual situation,
reliable, trustworthy.9,10 The dental evidence is all primary human identification techniques
reliable thanks to the uniqueness and stability of (fingerprinting, DNA, and dental evidence) were
the dental tissues and treatments and includes used since it was impossible to identify the
dental records, study models, photographs, X- victims visually. Whenever possible, multiple
rays.11-13 identification techniques were used in each case.
However, for a successful comparison process of
ante-mortem and post-mortem dental evidence, Post-mortem data:
it is imperative to overcome several difficulties.14 Both jaws were available for examination. Post-
The most frequent challenge is when incomplete, mortem dental examination and charting were
ambiguous, or incorrectly documented ante- completed. The child was undergoing
mortem dental records are avalilable.14 Other orthodontic treatment, as evidenced by the post-
obstacles that may influence the identification mortem data of his corpse, which demonstrated
procedures are the administrative and that fixed orthodontic appliances had been fitted
collaborative issues with foreign authorities about in his maxilla.
the collection of the ante-mortem data.15 In particular, the first upper premolars were
Although the DNA evidence is a powerful tool in missing bilaterally, the first upper permanent
identification cases and especially in mass molars had orthodontic stainless-steel rings
disasters victims identification, there are various (bands), in which a trans-palatal arch was fixed.
limitations to be considered.16 Both the upper permanent canines, lateral and
Orthodontics is the dental speciality in charge of central incisors had been erupted and had
establishing a functional and aesthetic harmony metallic brackets, with hooks on the canines.
to the face. Since the treatment lasts a long time Also, the second permanent molars were erupted
and its planning is complicated, the orthodontist bilaterally (Fig.1).
takes several records in the pre-treatment, In the mandible, the first premolars were missing
treatment and post treatment phese. Mostly the bilaterally, the first permanent molars had
records consist of patient’s history, dental models, orthodontic rings and all the rest of the
extra oral and intraoral photographs, and X-rays permanent teeth had brackets. Moreover, the
(at least an Orthopantomogram and a lateral s e c o n d p e r m a n e n t m o l a r s w e r e e r u p te d
Cephalogram). All these data taken from the bilaterally. Supplementar y to the clinical
orthodontic records can help in the identification examination, post-mortem photographs were
of a patient.17 Dental evidence is a reliable tool taken (Fig.1). Age estimation based on tooth
thank to the uniqueness and stability of the eruption was between 12.5 and 16.5.
dental tissues and treatments.11
Two mass disaster cases are presented in this Ante-mortem data:
study; the first case can be characterized as Among the victims' postmortem dental records
closed and international and the second case as obtained by the police there was the orthodontic
3
record of a male child. The record included notes, Comparison of post-mortem and ante-mortem dental
a panoramic radiograph taken before the evidence.
orthodontic treatment,1 extraoral photograph, The little youngster was positively recognized based
3 intraoral photographs, and a set of on the similarities found from the analysis of the
orthodontic models (Fig. 2,3,4,5). The ante- dental records' ante-mortem and post-mortem data.
mortem data were collected and charted. The Specifically, there was no difference between the
study of the ante-mortem records revealed extracted teeth found in the ante-mortem and post-
that the boy had orthodontic stainless-steel mortem dental records. In both the antemortem
rings (bands) in his maxillary first molars in and postmortem reports, the fixed trans-palatal
which a trans-palatal arch was fixed, all teeth arch was precisely the same. And both the dental
(from # 15 to # 25) had metallic brackets with records, antemortem and postmortem, showed the
hooks on the canines. metal brackets with hooks on the canines.
Figure 1. Post-mortem photographs from both jaws which were available for examination. a) A stainless
steel trans-palatal arch was fixed on the palate. b) Fixed orthodontic appliances were placed on the
maxilla, first premolars bilaterally were not present, first permanent molars had orthodontic stainless-
steel rings (bands). Both permanent canines, lateral and central incisors had erupted and had metallic
brackets, with hooks on the canines. c) Right lateral view of the maxilla and mandible, d) Left lateral
view of the maxilla and mandible, second permanent molars had erupted bilaterally.
4
Figure 2. A panoramic radiograph taken by the orthodontist before the orthodontic treatment started
Figure 3. Extraoral photograph taken by the orthodontist. Face frontal lips relaxed
5
Figure 4. Intra-oral photographs taken by the orthodontist before treatment. a) Frontal in occlusion.
b) Upper occlusal a stainless steel trans-palatal arch fixed on the palate. c) Lower occlusal
Figure 5. Orthodontic plaster models
CASE II
Two homozygous twin girls were among the 102 establish the identity of the victims was the
people that died in a fire which broke out in a evaluation of dental evidence.
village in Attica in July 2018. Visual identification
was impossible. Moreover, the identification of Post-mortem data:
the children’s corpses could not be based on the The maxilla and the mandible were available for
DNA analysis process since the individuals were post-mortem examination in both cases. The
homozygous twins and shared exactly the same children were both in the phase of mixed
DNA.Therefore, the only suitable process to dentition (Fig. 6).
6
In the maxilla of both children, the first kind of dental intervention in their teeth.
permanent molars were erupted bilaterally, first Ho w e v e r, w e f o u n d d i f f e r e n c e s i n t h e
premolars were erupted bilaterally, both upper arrangement of the anterior maxillary teeth.
incisors were erupted, as well as the second According to the first plaster model, the left
deciduous molars bilaterally. central and lateral incisors had erupted labially,
The different finding among the post-mortem while the right lateral incisor had not yet been
dental findings of the children was that in the erupted since there was not enough space for its
maxilla of one of them, the left central and lateral eruption (Fig 7 and Fig. 8). On the contrary, the
incisors had erupted labially, while the right maxillary teeth of the second child’s orthodontic
lateral incisor had not been erupted yet since it plaster model seemed to have been erupted
seemed that there was not enough space for its normally (Fig. 9).
eruption. On the contrary, the maxillary teeth of
the second child had been erupted normally (Fig. Comparison of post-mortem and ante-mortem dental
6). Τhe dentition of both children did not evidence
present any kind of dental intervention. It was possible to identify the homozygous twins
based on the maxillary tooth eruption pattern,
Ante-mortem data. which was documented post-mortem and with the
The information the family members gave the comparison of the ante-mortem dental plaster
police authorities indicated that the youngsters models (Fig.8 and Fig. 10). One orthodontic plaster
had attended a private orthodontist office model's eruption pattern matched the eruption
recently. The orthodontist had the initial models. pattern of the first girl’s left central and lateral
We obtained the orthodontic models by the incisors in the maxilla. One other consistent finding
orthodontist once the police found him. between the postmortem and ante mortem dental
The study of the two orthodontic plaster models records was the right lateral incisor not yet erupted
that belonged to the homozygous twin girls due to limited space. The second homozygous twin
revealed that both children did not present any girl's maxillary incisors had all erupted.
Figure 6. Post-mortem dental records included photographs of the maxillae and mandibles of both
children’s corpses, that were taken during the post-mortem examination. Both children were in the phase
of mixed dentition. Photographs a, and b belonged to one child. Photographs c and d belonged to the other
7
Figure 7. Ante-mortem dental records included plaster models (a and b) taken by the orthodontist for one
of the homozygous twin girls. The left central and lateral incisors in the maxilla had erupted labially, while the
right lateral incisor had not yet erupted, since there was insufficient space available for its eruption
Figure 8. a) Plaster model of the maxilla of one child taken by the orthodontist, ante-mortem records,
in the maxilla the left central and lateral incisors had erupted labially, while the right lateral incisor had
not been erupted yet since it seems that there was not enough space for its eruption. b) Post-mortem
dental records of the same child
8
Figure 9. Ante-mortem dental records for the second of the homozygous twin girls from the
orthodontist included plaster models (a and b). As can be observed, the maxillary incisors had erupted
Figure 10. a) Post-mortem dental records of the second homozygous twin girl. b) Plaster model of the
maxilla of the same child taken by the orthodontist, in ante-mortem records, the maxillary teeth of the
second homozygous twin girl had erupted
DISCUSSION
The orthodontist, from the very first days of his impressions, and any other document included in
post graduate training, is used to take routinely a a patient's dental record are the evidence the
set of diagnostic records pre-treatment, during it orthodontist relies on to make initial diagnoses,
and at its end. These records are essential for assess prognoses and progress of the treatment
planning and carr ying out the appropriate outcome.18 These include case history recording,
t h e r a p i e s . Wr i t t e n i n f o r m a t i o n , d e n t a l a full set of intra and extra oral photographs, with
radiographs, photographs, plaster casts, digital complete clinical and a thorough medical
9
examination. Moreover, the orthodontist has missing people such as radiographic images out of
extensive knowledge about the various aspects of the dentomaxillofacial area could be considered
te e t h p o s i t i o n , c r a n i o f a c i a l g r o w t h , a g e an useful identification tool.31 Furthermore, when
assessment and can deal with the skeletal and there is an extreme lack of antemortem medical
dental remains. A well-detailed, accurate, and records, the analysis of images from other sources
updated dental record is an indicator of the (like social media) might be suitable for human
excellence of the orthodontic services provided. identification.32
Excellent documentation is directly related to the In order to use dental records for mass disaster
o r t h o d o n t i s t 's c l i n i c a l c o m p e t e n c e a n d human identification, a number of requirements
contributes to the increase of his prestige and should be satisfied: trained personnel must
trustworthiness. General dentists are more likely collect and evaluate pre- and post-mortem dental
to keep incomplete, insufficient, and out-of-date records; a uniform dental notation and coding
records.19 system should be adopted; the necessar y
Furthermore, orthodontic appliances are equipment must be available; and the personnel
generally manufactured of commercially pure must submit to media pressure. Human remains
titanium, titanium-aluminum-vanadium alloys, or must be adequately stored and recorded in order
nickel, molybdenum, copper, and stainless-steel to safeguard the dentomaxillofacial evidence. The
elements. The metals and alloys mentioned above process for identifying teeth should adhere to
have unique physical and mechanical properties, quality control guidelines. Since various DNA
as well as corrosion resistance.20-23 Fixed as well as tests that require specific expertise are available,
removable orthodontic appliances are valuable the DNA analysis should be used when dental
dental identification tools due to the evidence or fingerprint analysis are
aforementioned properties of the orthodontic inconclusive.15,16
materials and their resistance to extremely Given the complexities of orthodontic cases and
unfriendly circumstances.24-26 Therefore, all the long-term treatment plans, orthodontists should
data an orthodontist collects can be a significant always update the patient records on a regular
source of information in cases where basis. As a consequence, orthodontic records are
identification with soft tissue is impossible due to valuable in forensic cases.27-29
the putrefaction of the bodies. The success of an identification process based on
Creating and maintaining detailed, accurate forensic dental evidence depends on the
updated and well organized dental records is availability, accuracy and good quality of the ante-
essential for a variety of reasons.27-29 Concerning mortem and post-mortem dental records.20 Due
forensic odontology, detailed and accurate dental to their excellent quality, orthodontic dental
records, may help not only with the identification records are extremely valuable tools and when
of unidentified corpses or human remains by they are available the human identification
comparing antemortem and post-mortem dental procedures are greatly facilitated.33,34
evidence, but also to the reconstruction of an The present study highlights the contribution of
individual's biological profile and additionally orthodontics to establish the identity of mass
with the resolve of other criminal and civil law catastrophes victims. In the two cases presented
cases.22,30 the identification was based exclusively on the
The unavailability of dental records or the excellent quality and accuracy of the orthodontic
availability of insufficient dental records are records. The contribution of orthodontics
major concerns which complicates the records was even more prominent for the case of
identification procedures in isolated cases and in the identical twins where all the other primary
mass disasters. In this case scenario the collection forensic methods (DNA and finger prints) failed
of all available ante mortem medical data of to reach a definite result.
REFERENCES
1. Marrone M, Tarantino F, Stellacci A, Baldassarra SL, Cazzato 2. Brough A.L., Morgan B. RGN. The basics of disaster
G, Vinci F D. A. Forensic Analysis and Identification victim identification. Journal of Forensic Radiology and
Processes in Mass Disasters: Explosion of Gun Powder in the Imaging. 2015;3(1):29–37.
Fireworks Factory. Molecules. 2022;27(1):244. 3. Nathan MD SD. Dentistry and mass disaster - a review. J
10
Clin Diagn Res. 2014;8(7):ZE01-3. identification. Acta odontol Venez. 2005;43:67–74.

4. Pereira CP SJ. How to do identify single cases according 18. Silva RF, Chaves P, Paranhos LR, Lenza MA DJ. Use of
to the quality assurance from IOFOS. The positive orthodontic records in human identification. Dental.
identification of an unidentified body by dental Press J Orthod. 2011;16:52–7.
parameters: a case of homicide. J Forensic Leg Med. 19. Abdelkarim A, J. L. ‘Orthodontic chart documentation.’,
2013;20(3):169–73. Am J Orthod Dentofac Orthop., 2017; 152(1), pp. 126–130.
5. Spencer DE. Forensic odontology: an overview. J Calif 20.Nakajima H OT. Titanium in dentistry: development and
Dent Assoc. 2014;42(6):397–405. research in the U.S.A. Dent Mater J. 1996;15(2):77–90.
6. Matsuda S YH. Possibilities and challenges in digital 21. Brantley WA ET. Orthodontic materials: scientific and
personal identification using teledentistry based on clinical aspects. 1st ed New York: Thieme, Stuttgart;
integration of telecommunication and dental 2000;
information: a narrative review. J Int Med Res. 22. Park J KY. Metallic biomaterials. In: Wong JY, Bronzino
2022;50(4). JD, editors. Biomaterials. 1st ed Boca Raton: CRC Press;
7. Pretty IA SD. A look at forensic dentistry-Part 1: The 2007;
role of teeth in the determination of human identity. Br 23. Sfondrini MF, Cacciafesta V, Maffia E, Massironi S S, A,
Dent J 14;190(7):359-66. 2001;190(7):359–66. Alberti G, Biesuz R KC. Chromium release from new
8. Junior, E. F., Legal, D.O, Claudio, L. & Moura L. A stainless steel, recycled and nickel-free orthodontic
importancia dos arcos dentarios na identificacio humana. brackets. Angle Orthod. 2009;79(2):361–7.
Rev Bras Odontol. 2014;7(5):262. 24. González-Colmenares G, Calvo-Díaz L, N.-E. M. and
9. Vidal-Parra M FG. Registros ortodóncicos para la Bertel-Ruíz MM, Garzón-Ramírez I, R.-S. M. ‘Effect of
identificación forense: una revisión exploratoria. Revista high temperatures on teeth fixed with an orthodontic
Española de Medicina Legal. 2022;48:78–83. bracket. An in vitro study.’, Forensic Sci Int., 2020; 308, pp.
10. Dyke AEC, Cunningham S, Hunt N RC. A comparative 1–6.
study to investigate the effect of orthodontic treatment 25. Seals Jr RR, S. D. ‘The importance of denture
on the uniqueness of the human anterior dentition. identification.’, 1985;Spec Care Dentist., 5(4), pp. 164–8.
Forensic Sci Int 2018 Aug;289:368-373. 289:368–73. 26. Gosavi S GS. Forensic odontology: a prosthodontic view.
11. Puri P, Shukla SK ,Haque I. Developmental dental J Forensic Dent Sci. 2012;4(1):38–41.
anomalies and their potential role in establishing identity 27. SK. J. Role of dentist in forensic investigations. J
in post-mortem cases: a review. Med Leg J. 2019; Mar; Forensic Res. 2012;3:148.
87(1):13–8. 28. Krishan K, Kanchan T GA. Dental evidence in forensic
12. Martin-de-las-Heras, S.; Valenzuela, A.; Luna. J. D. & identification–an overview, methodology and present
Bravo M. The utility of dental patterns in forensic status.
dentistry. Forensic Sci Int,. 2010;195:166e1-e5. 29. Open Dent J. 2015;9:250–6.
13. Moura, R. N. V.; Ferreira. E. F.; Coste, S. C.; Corradi 30. Re d d y G, Re d d y V P, S h a r m a M A M . Ro l e o f
LM&, Travassos D V. Identifying missing people: the orthodontics in forensic odontology–a social
contribution of forensic dentistry and DNA. Rev responsibility. J Clin Diagn Res. 2016;10(4):ZE01-3.
Odontol UNESP,. 2017;46(6):313–8. 31. Correia AM, Alencar PNB, Sobrinho MLS, Silva RF CF,
14. Terada, A. S. S. D.; Araujo, L. G.; Paranhos, L.R.; TP. B. Orthodontic records helping to identify a victim
Guimaraes MA., Silveira, T. C. P. & Silva RHA. of execution. Int J Odontostomat. 2021;15:403–8.
Orthodontic use of documentation in identification of a 32. Bianchi IA, Focardi MB, Grifoni R, Raddi S, Rizzo A,
skeletonized body in legal dental practice. Int J Defraia B, Pinchi V. Dental identification of unknown
Odontostomat, 2014;8(1):41–6. bodies through antemortem data taken by non-dental X-
15. Samarika Dahal, Gopal Kumar Chaudhary, Mani Raj rays. Case reports. J Forensic Odontostomatol.
Maharjan, Eugen Dolma Walung, A Dental Perspective 2021;39(3):49-57.
on the Successes and Limitations of the Disaster Victim 33. Franco RPAV, Franco A, da Silva RF, Pinto PHV, Alves
Identification Response to the Nepal Earthquake, da Silva RH. Use of non-clinical smile images for human
Forensic Sciences Research, Volume 7, Issue 3, September i d e n t i f i c a t i o n : a s y s te m a t i c r e v i e w. J Fo r e n s i c
2022, Pa g e s 3 6 6 – 3 7 0, https://doi.org/ Odontostomatol. 2022;40(1):65-73.
10.1080/20961790.2022.2034716 34. Fonseca GM, Cantín M LJ. Odontología forense II: la
16. Jordan D and Mills D (2021) Past, Present, and Future of identificación inequívoca. Int J Odontostomat.
DNA Typing for Analyzing Human and Non-Human 2013;7(2):327–34.
Forensic Samples. Front. Ecol. Evol. 9:646130. doi: 35. Al-Amad S, McCullough M, Graham J, Clement J, Hill A.
10.3389/fevo.2021.646130 Craniofacial identification by computer-mediated
17. Silva RF, Cruz BVM, Daruge Júnior E, Daruge E FJ(. s u p e r i m p o s i t i o n . J Fo r e n s i c O d o n t o s t o m a t o l .
The Importance of dental documentation in human 2006;24(2):47-52.
11
Canine sexual dimorphism in crown and root

dimensions: a cone-beam computed tomographic
study
Copyright © 2024 International Organization

ABSTRACT
Mohammad Tajik 1, Najmeh The primary step in forensic odontological analysis is sex
Movahhedian1 determination. The present study is one of the few studies that
evaluated the accuracy of the combination of canine tooth root
1 Department of Oral and length and crown measurements for sex determination. The
Maxillofacial Radiology, School of s t u d y s a m p l e c o m p r i s e d 1 9 6 c o n e - b e a m c o m p u te d
Dentistry, Shiraz University of Medical tomographic scans of individuals aged 20-80 years distributed
Sciences, Shiraz, Iran. in five age categories: 20-29, 30-39, 40-49, 50-59, and 60+ years
old. Different parameters, such as width, length, and ratio
measurements for the crown and root of each maxillary and
mandibular canine tooth, were examined and recorded. The
Corresponding author: findings indicated that maxillary canines had greater sex
[email protected] dimorphism ability (87.3%) than mandibular canines (80.6%).
Total tooth length and root length of maxillary canine were the
most pronounced variables in the differentiation of sex groups.
When the combination of the mandibular and maxillary
mea surements wa s considered, the accuracy for sex
The authors declare that they dimorphism was 85.7%. By using ratio variables, the accuracy
have no conflict of interest. was reduced to 68.9%. According to the findings of this study,
total tooth length and root length are the most discriminant
variables of canine teeth. These variables are more reliable sex
indicators than crown measurements.
INTRODUCTION
KEYWORDS Forensic odontology is a subspecialty of forensic medicine that
assists in establishing a post-mortem biologic profile,1 which
Sexual Dimorphism, primarily includes verifying age, sex, stature, and ancestry.2-4
Forensic Dentistry, The primary step in this context is sex determination.
Canine tooth, Although sex can be most accurately assessed by DNA analysis5
or examination of skeletal remains such as the pelvis, 6
Tooth length, craniofacial bones, 7 and mandible, 8 their applicability is
Cone-beam Computed restricted due to a variety of factors, including the need for
Tomography expensive equipment and time-consuming procedures or
suboptimal condition of the bones, especially those severely
mutilated.9 In such situations, teeth are considered a practical
adjacent in sex determination because of their structural
durability to pre- and post-mortem insults 10, 11 and their ability
to predict sex with an accuracy of 51.13% to 100%. 12-15
J Forensic Odontostomatol Moreover, since the complete development of dentition
2024. Apr;(42): 1-12:21 precedes skeletal maturation, teeth play an essential role in sex
ISSN :2219-6749 determination for younger individuals.1
The application of dentition in sex determination is primarily
based on the distinctions between the dimensions or
12
morphology of the teeth in males and females.3, 16, m ore practical for forensic applications,
17 Among different types of teeth, the canine particularly in skeletal imaging and odontology.20
tooth has been consistently used for forensic Therefore, the present study aimed to assess the
purposes and is regarded as a "key tooth" in sex accuracy of sex determination in adult individuals
determination1, 18 primarily because they show using CBCT images based on the different
the most significant sex dimorphism in their length, width, and ratio measurements of the
dimensions. Moreover, they are the least crown and root of canine teeth in mesiodistal and
frequently extracted teeth,3 highly resistant to buccolingual dimensions in both jaws.
dental/periodontal diseases, and are more likely Additionally, it attempted to present a specific
to survive post-mortem trauma. formula for sex determination based on the
Many studies on the effectiveness of canine teeth canine teeth using the discriminant function
in sex determination used the canine index, analysis. The present study is the first to be
which considers the mesiodistal width of the conducted among the Iranian population. It is
canine teeth and inter-canine arch width.1-3, 15 one of the few studies evaluating the accuracy of
However, some studies showed that the canine canine sex dimorphism considering the tooth
index has poor sex dimorphism ability, and its root length and crown measurements.
application in forensic works should be confined,
and absolute measurements of the canine tooth MATERIALS AND METHODS
are better sex indicators.1, 3, 18 Another limitation In this cross-sectional study, the CBCT scans
of such studies is that they have not considered of 196 individuals (96 men and 100 females)
root measurements. This factor may impair the referred to the Dental School of Shiraz
accuracy of sex determination since it has been University of Medical Sciences (Shiraz, Iran)
demonstrated that the Y chromosome has a more for purposes other than the present study were
decisive influence over root length growth than evaluated. The study complied with all relevant
the X chromosome.12 principles of the Declaration of Helsinki at the
Radiographic examinations are a non-destructive time of imaging. All subjects signed a written
and ethical approach for evaluating the whole informed consent form authorizing the use of
tooth if radiographs are taken due to clinical their anonymous radiographic data in research
indications. Capitaneanu et al.19 used panoramic and publications. The Ethics Committee of
radiographs to compare the length and width Shiraz University of Medical Sciences, Shiraz,
variables and ratios to determine the applicability Ir a n , approved the study
of various maxillary and mandibular teeth in (IR.Sums.Dental.REC. 1399.206). The study
sexual determination. They showed that the sample age ranged from 20 to 80 years and was
tooth length of the mandibular canine was the divided into five age groups: 20-29, 30-39,
most sexually dimorphic measurement. However, 40-49, 50-59, and 60+ years old. Except for the
panoramic radiographs have disadvantages, such final group, which comprised 16 males due to a
as unequal magnification and unpredictable lack of individuals matching the inclusion
distortion due to the patient positioning and criteria, all groups included 40 CBCT scans
location within the focal trough.20 These factors distributed evenly between the two sexes. The
preclude accurate measurements, and the results CBCT images were digitally captured using a
cannot accurately reproduce or represent direct New Tom VGi Evo CBCT unit (QR SRL
measurements. In recent years, computed Company, Verona, Italy) with 3 mA, 110 kVp,
tomography (CT) or cone-beam CT (CBCT) has and 0.3 mm voxel size, in accordance with the
been increasingly used in forensic investigations, manufacturer's instructions for positioning and
particularly for sex and age determination.21 CT exposure.
and CBCT provide images with a sub-millimeter The inclusion criteria of the study were CBCT
resolution free from distortion, magnification, scans with good image quality, the presence of
and superimposition of the adjacent structures, fully developed and erupted maxillary and
[22] which are not available in projection imaging mandibular permanent canine teeth, the
and panoramic radiography and can positively absence of restorations, significant occlusal
affect the findings of the studies. 21 The wear on the crown of the canine teeth, root
portability, lower cost, ease of image acquisition, resorption, dilacerations, significant
and user-friendliness of CBCT over CT make it buccolingual inclination in the canines, any
13
pathology or skeletal disorders, and history of the most incisal tooth point and the root apex)
trauma or orthodontic treatment. and the maximum root length (RL) (defined as
For each scan, the measurements were taken on the distance between the cemento-enamel
the canine teeth of both jaws, i.e., maxilla and junction (CEJ) and root apex) (Figure 1).
mandible. Since previous studies showed no
significant difference in canine measurements • The width measurements: The maximum
between the left and right sides of the jaw,3, 23, 24 buccolingual width of the crown (BL), the
all the measurements were made on the left buccolingual width of the tooth at the CEJ level
canine teeth. Yet, if the left canine tooth did not (CEJBL), the maximum mesiodistal width of
meet the inclusion criteria, the procedure was the crown (MD), the mesiodistal width of the
made on its right counterpart. Each canine tooth tooth at the CEJ level (CEJMD) (Figure 2).
underwent the following measurements: • The ratio and proportions: TL/RL, TL/MD, TL/
BL, TL/ CEJMD, TL/ CEJBL, RL/MD, RL/BL,
• The length measurements: The maximum tooth
RL/ CEJMD, RL/ CEJBL, MD/ CEJMD, BL/
length (TL) (defined as the distance between CEJBL, CEJBL/ CEJMD, BL/MD
Figure 1. Shows the tooth and root length measurements (TL & RL, respectively) used in the study:
Axial images used for reconstruction of mesiodistal cross-section; b the maximum tooth length; c the
maximum root length
Based on the axial section at the mid-root level, recorded all the measurements separately using
cross-sections perpendicular to the canine tooth NNT Viewer software (NNT V2.21, Image
were prepared for the measurements. The length works, Verona, Italy). The examiners were
measurements (TL and RL) were done on the blinded to the subject's age and sex. To assess
mesiodistal cross-sections. The thickness of these the intra- and inter-observer reliability, the
cross-sections was considered around 7 mm to examiners randomly selected one-third of the
account for the buccolingual inclination of the images and re-evaluated them after a two-week
canine tooth and to include both the incisal edge interval.
and the apex of the tooth in one section. MD and
CEJMD measurements were done on 1 mm thick Statistical analysis
mesiodistal cross-sections representing the Data were analyzed using SPSS software, version
maximum crown width. Buccolingual cross- 22.0 (IBM Corp., Armonk, N.Y., USA). For each
sections with 1 mm thickness were prepared to variable, the two sexes were compared using an
measure BL and CEJBL variables. For each independent t-test. P values less than 0.05 were
subject, additional information such as the considered statistically significant. To find the
subjects' birth dates, date of acquisition of the best variables for sex determination, the
CBCT scans, and sex were also recorded. discriminant analysis and stepwise selection
Two examiners, an experienced oral and method were used. The accuracy of inter- and
maxillofacial radiologist and a well-trained post- the intera-examiner agreement was determined
graduate oral and maxillofacial radiology student, by Intra-class correlation coefficient test (ICC).
14
Figure 2. Shows the width measurements used in the study: an Axial image used for reconstructing mesiodistal
cross-sections; b The maximum mesiodistal width of the crown (MD) and the mesiodistal width at the CEJ level
(CEJMD); c Axial image used for reconstructing buccolingual cross-sections; d The maximum buccolingual
width of the crown (BL) and the buccolingual width of the tooth at the CEJ level (CEJBL)
RESULTS
Based on the ICC values, there was a high inter were the TL and RL of maxillary canines. Other
and intra-obser ver agreement for all variables based on the magnitude of Standardized
odontometric measurements (r>0.90 and >0.95, Discriminant Function Coefficients (SDFC)
respectively). For analyzing the data, an average scores were CEJMD of maxillary canine, RL,
for each value was used. MD, and CEJMD of mandibular canines,
According to the independent t-test results, all respectively.
the maxillary and mandibular canine teeth length The discriminant function for all maxillary and
and width measurements significantly differed mandibular measurements was formulated using
between the sexes (all P values <0.05). There were the Canonical Discriminant Function Coefficient
also significant differences between males and (CDFC) as follows:
females in eight ratios of the upper jaw and two Formula 1 (for both jaws): D = [-20.254+ 0.571 (MD
of the lower jaw (Table 1). of mandibular canine) + 0.463 (TL of maxillary
The results of the stepwise discriminant analysis canine) - 0.270 (RL of maxillary canine) + 0.168
are shown in Table 2. When all measurements (RL of mandibular canine) + 0.976 (CEJMD of
from both jaws were considered, the most maxillary canine) + 0.434 (CEJMD of mandibular
pronounced variables in sex group differentiation canine)].
15
Based on this formula, the discriminant D-score As presented in Table 2, the maxillary canine had
was 0.023 (values greater than 0.023 indicate an accuracy rate of 87.3% (87.5% for males and
males, while lower values indicate females). 87.0% for females), and the mandibular canine
had an accuracy rate of 80.6% (79.2% for males
Based on the corresponding CDFCs, the and 82.0% for females). When both jaws were
following formulae were presented for obtaining considered, the accuracy rate was 85.7% (86.0%
the discriminant function separately for each jaw for males and 85.4% for females).
measurements:
Formula 2 (for the maxi#a): D = -18.568+ 0.502 (TL) Based on the magnitude of SDFC scores, the best
- 0.247 (RL) + 1.286 (CEJMD) + 0.529 (CEJBL) ratio variables for sex discrimination were MD/
The discriminant D-score was 0.022 (values CEJMD and TL/MD of the maxillary canine.
greater than 0.022 indicate males, while lower Based on the corresponding CDFCs, the
values indicate females). discriminant function for ratio variables was
Formula 3 (for the mandible): D = -18.829+ 0.962 formulated as follows:
(MD) + 0.261 (RL) + 0.952 (CEJMD) + 0.542 Formula 4: D = -6.330 – 1.171(TL/MD of maxillary
(CEJBL) canine) + 7.995(MD/ CEJMD of maxillary canine)
The discriminant D-score was 0.019 (values With the discriminant D-score of 0.017 (values
greater than 0.019 indicate males, while lower greater than 0.017 indicate males, while lower
values indicate females). values indicate females). The overall accuracy of
Based on the stepwise anal ysis, the best these ratio variables was 68.9%, which was the
differentiating variables for each jaw were TL for lowest among all the reported accuracies in the
the maxilla and CEJMD for the mandible. present study (Table3).
Table 1. Comparisons of all the variables of the permanent canine teeth between the two sexes)
Variables Mean±SD P value
Male Female
MD 7.12±0.47 6.63±0.51 <0.001*
BL 8.27±0.56 7.57±0.56 <0.001*
Maxilla TL 26.30±1.82 23.43±1.86 <0.001*
RL 19.16±1.84 17.01±1.67 <0.001*

Width
CEJMD 5.58±0.49 4.89±0.44 <0.001*
and
CEJBL 7.53±0.51 6.90±0.52 <0.001*
length
MD 6.38±0.43 5.83±0.44 <0.001*
measurements
BL 7.75±0.51 7.14±0.54 <0.001*
TL 24.70±1.81 22.47±1.79 <0.001*

Mandible
RL 17.43±1.75 15.85±1.67 <0.001*
CEJMD 5.47±0.53 4.84±0.56 <0.001*
CEJBL 7.10±0.51 6.52±0.54 <0.001*
16
TL/RL 1.37±0.07 1.38±0.06 0.601
BL/MD 1.16±0.08 1.14±0.08 0.150
RL/BL 2.32±0.23 2.26±0.24 0.038*
RL/MD 2.70±0.28 2.58±0.28 0.002*
TL/BL 3.19±0.25 3.10±0.28 0.026*
TL/MD 3.70±0.31 3.55±0.35 0.001*
TL/CEJBL 3.50±0.28 3.41±0.30 0.026*

Maxilla
BL/CEJBL 1.09±0.03 1.09±0.03 0.349
RL/CEJBL 2.56±0.26 2.48±0.27 0.040*
MD/CEJMD 1.28±0.09 1.36±0.12 <0.001*
CEJBL/ 1.36±1.35 1.42±0.15 0.003*

CEJMD
TL/CEJMD 4.74±0.47 4.82±0.53 0.0727

Ratios
RL/CEJMD 3.45±0.38 3.49±0.41 0.0572
and
TL/RL 1.42±0.10 1.42±0.10 0.940
proportions
BL/MD 1.21±0.09 1.22±0.98 0.384
RL/BL 2.25±0.26 2.22±0.27 0.440
RL/MD 2.74±0.34 2.73±0.33 0.788
TL/BL 3.19±0.26 3.15±0.29 0.331
TL/MD 3.70±0.31 3.54±0.34 0.839
TL/CEJBL 3.50±0.28 3.40±0.30 0.577

Mandible
BL/CEJBL 1.09±0.03 1.09.0.04 0.903
RL/CEJBL 2.46±0.29 2.44±0.30 0.042
MD/CEJMD 1.18±0.12 1.21±0.13 0.041*
CEJBL/ 1.30±0.14 1.36±0.15 0.024*

CEJMD
TL/CEJMD 4.54±0.49 4.67±0.55 0.0751
RL/CEJMD 3.21±0.41 3.29±0.46 0.0660
The results of the independent T-test are shown in the table

* P value less than 0.05 was considered statistically significant
MD: the maximum mesiodistal width of the crown; BL: the maximum buccolingual width of the crown; TL: the maximum
tooth length; RL: the maximum root length; CEJMD: the mesiodistal width of the tooth at the CEJ level; CEJBL: the
buccolingual width of the tooth at the CEJ level
17
Table 2. The length and width measurements of canine teeth with significant differentiating function
based on discriminant analysis
Overall
Jaw Variable C.D.F.C * S.D.F.C**
Accuracy (%)
TL2 0.463 0.857
RL2 -0.270 -0.475
CEJMD2 0.976 0.458

Both Jaws 85.7
RL1 0.168 0.288
MD1 0.571 0.253
CEJMD1 0.434 0.217
TL 0.502 0.928
Maxillary CEJMD 1.286 0.603

87.3
canine RL -0.247 -0.435
CEJBL 0.529 0.277
CEJMD 0.952 0.477
Mandibular MD 0.962 0.428

80.6
canine RL 0.261 0.448
CEJBL 0.542 0.286

* Canonical Discriminant Function Coefficient
** Standardized Discriminant Function Coefficient
TL: the maximum tooth length; RL: the maximum root length; CEJBL: the buccolingual width of
the tooth at the CEJ level; MD: the maximum mesiodistal width of the crown; CEJMD: the
mesiodistal width of the tooth at the CEJ level
1 and 2 indicate mandibular and maxillary measurements, respectively.
Table 3. The ratio variables of canine teeth with significant differentiating function based on
discriminant analysis
Overall
Jaw Variables C.D.F.C * S.D.F.C**
accuracy (%)
Maxillary MD/CEJMD 7.995 0.841

68.9
canine TL/MD -1.171 -0.386
* Canonical Discriminant Function Coefficient

** Standardized Discriminant Function Coefficient
TL: the maximum tooth length; MD: the maximum mesiodistal width of the crown; CEJMD:
the mesiodistal width of the tooth at the CEJ level
DISCUSSION
Exact sex estimation is the foremost step in the teeth, the canine tooth has consistently been
identification process. Teeth and skeleton-based considered the most sexually dimorphic and
methods have a prominent application in sex critical tooth across various populations.12, 18, 25, 26
determination as they are usually the best- Most studies on the applicability and accuracy of
preserved remains.9 Among different types of the canine teeth in sex dimorphism are based on
18
the crown measurements and canine index, and based measurements using NewTom scanning
root measures have rarely been utilized for sex and direct measurements by caliper was just 0.07
dimorphism analysis. This is mainly due to the ± 0.41mm. In addition, Kim et al.36 found no
inaccessibility of the tooth root.27 The root, statistically significant differences in the crown
preserved in the bony socket, is considered a and root length in CBCT scans and direct
more resilient structure than the tooth crown measurements of the premolar teeth using a
and, unlike the crown, is not affected by wear.28 d i g i t a l c a l i p e r. Ho w e v e r, C B C T- b a s e d
According to Lähdesmäki and Alvesalo,29 the Y measurements demonstrated that the total tooth
chromosome may play a more significant role in length of the premolars was 0.18 ± 0.44 mm
root length development than the X shorter than the direct measurements. This study
chromosome, which may be responsible for the also found a weak positive correlation between
sexual dimorphic characteristics of the root. the crown and root length in the canine and
Z o r b a e t a l .28 s h o w e d t h a t r o o t l e n g t h premolar teeth, indicating that the crown length
measurements of single-rooted teeth were a cannot accurately estimate root length in most
reliable indicator of sex dimorphism. Moreover, tooth types.
other studies reported that the root length of To lessen the impact of dental wear, the studies
permanent teeth had a higher degree of sexual using tooth crown measurements for sex
dimorphism than the crown measurements.9, 30 estimation12, 19, 37, 38 had to restrict the age range of
According to Garn et al.,9 root length alone their study samples to young adults. According to
shows comparable or greater sexual dimorphism Lambrechts et al.,39 the annual enamel vertical
as the crown measurements. They also showed loss in vivo ranged between 20 and 38
that combining root and crown measurements micrometers. Consequently, the teeth with
enhances the sexual discriminant power of each significant attrition were excluded from the
set of measurements on its own. However, few present investigation. On the other hand, since
studies9, 13, 19, 28, 31 considered canine tooth root the present study included tooth root
measurements for sex determination. measurements, which were not affected by wear,
Moreover, since different populations exhibited the age range of the study sample was set at
distinct patterns of sexual dimorphism, the 20-80 years old. This age range was comparable
findings of such studies were considered to studies by Zorba et al.28 and Kazzazi and
population-specific 13, 32 and should not be Kranioti,13 which similarly used root length and
generalized. Only one study has examined tooth root volume measurements for sex
root dimensions for sex determination in the determination.
archaeological Iranian population.13 This study The current study’s findings indicated that all the
was a CT volumetric examination of 52 length and width measurements of the maxillary
archaeological skeletal remains dating from and mandibular canine teeth were significantly
around 1400 to 800 BCE, which may not reliably higher in men than in women. However, only 8
represent the present population. ratio variables of the maxillary canines and 2 ratio
CBCT has been widely used for forensic purposes variables of the mandibular canines demonstrated
in recent years. 21, 33 Besides sub-millimeter a significant difference between the sexes.
resolution, CBCT offers other advantages over Similarly, Capitaneanu et al.19 assessed all the
conventional two-dimensional radiographs, maxillary and mandibular teeth in the panoramic
including distortion-free images, magnification, image and reported that males had higher mean
superimposition of the adjacent structures, and tooth length and mean width measures. Other
the capacity to adjust the structure orientation.21 studies 1, 12, 24, 37 that evaluated the crown or
These features enhance the ability to locate the cervical tooth measurements also found that men
anatomic structures better, which leads to more had higher values than women.
accurate results. Sherrard et al.34 investigated the It is estimated that for the dental measurements
reliability of the tooth measurements on CBCT to be used as the sole sex predictor, the accuracy
scans. They found that the total tooth and root should be at least 80%.19 In the current study,
length measurements were not significantly discriminant analysis results indicated that
different from the direct measurements on the maxillary canines had a higher sexual dimorphic
extracted teeth. Similarly, Stratemann et al.35 ability with an accuracy of 87.3%, whereas
reported that the difference between CBCT- mandibular canines had an accuracy of 80.6%.
19
When the measurements of both jaws were that combining tooth variables using ratios did
considered, the discriminant ability was found not increase the discriminant ability. They
to be 85.7%. Dumančić et al.,40 who explored introduced ratio variables to reduce the
sex dimorphism in canine teeth, reported an undesirable effect of unequal magnification of
accuracy of 73.5% for mandibular canine crown panoramic radiography. However, employing ratio
dimension and morphology. Zorba et al. 28 variables, which likewise have less accuracy in sex
evaluated the root length of single-rooted teeth dimorphism, seems unnecessary for this purpose,
and discovered that the maxillar y lateral as using CBCT makes it possible to have real-
incisors and canine were the most dimorphic sized images.
teeth. Similar to the present study, they
reported that maxillar y canine had higher CONCLUSION
sexual discriminant accuracy than mandibular Based on the discriminant analysis results, the
canine. They reported 80% and 76.9% maxillar y canine tooth was more sexually
discriminant accuracy for right and left dimorphic than the mandibular canine tooth
maxillary canine and 74.4% and 77.6% for right (87.3% vs. 80.6%). However, based on the
and left mandibular canine, which was lower accuracy, both had reliable sex dimorphism
than the present study. While Zorba et al.28 abilities. The use of ratio variables reduced the
only measured the root length, the current accuracy of sex dimorphism to 68.9%. The TL
study considered the length measurements of and RL of maxillary canine were the most
the tooth and root and the mesio-distal and discriminant variables of canine teeth. Root and
buccolingual width measurements of the canine total tooth measurements were more reliable sex
tooth crown. indicators than crown measurements. Thus, it is
In the present study, the tooth length and root r e co m m e n d e d t h a t r o o t a n d to t a l to o t h
length of maxillary canines showed the most mea surements be considered in forensic
prominent sex dimorphism among all the investigations for sex determination.
variables. This finding indicated that root
dimension and tooth length measurements are ACKNOWLEDGMENT
more sex dimorphic than crown measurement The authors thank the Vice-Chancellery of Shiraz
and should be considered in forensic University of Medical Sciences for supporting
investigations. These findings were in this research (Grant number: 22084). This
agreement with the findings of Capitaneanu et manuscript is based on the thesis by Dr.
a l . ,19 w h o r e p o r t e d t h a t t h e l e n g t h Mohammad Tajik, a post-graduate student of
measurements of mandibular and maxillary Oral and Maxillofacial Radiology under the
canines had a higher discriminant ability than supervision of Dr. Najmeh Movahhedian. The
the width measures. They also reported that authors also thank Dr. Mehrdad Vossoughi of the
mandibular canine tooth length was the most Center for Research Improvement of the School
discriminative variable among al l teeth. of Dentistry for the statistical analysis.
According to the present study’s findings, when
the ratio variables were considered, the Statements and Declarations:
accuracy for sex dimorphism decreased to This work was supported by Shiraz University of
68.9%. Similarly, Capitaneanu et al.19 showed Medical Sciences (Grant number 22084)
REFERENCES
1. Acharya AB, Mainali S. Limitations of the mandibular dental age in radiographic images. J Dent Child (Chic)
canine index in sex assessment. J Forensic Leg Med 2018;85(2):45-50.
2009;16(2):67-9. 5. Rötzscher K. Forensic and legal dentistry. 4 ed. Switzerland.
2. Azevedo Á, Pereira ML, Gouveia S, Tavares JN, Caldas Springer; 2014.
IM. Sex estimation using the mandibular canine index 6. Steyn M, İşcan MY. Metric sex determination from the
components. Forensic Sci Med Pathol 2019;15(2):191-7. pelvis in modern Greeks. Forensic Sci Int 2008;179(1):86.e 1-6.
3. Gupta S, Chandra A, Gupta OP, Verma Y, Srivastava S. 7. Williams BA, Rogers T. Evaluating the accuracy and
Establishment of sexual dimorphism in North Indian precision of cranial morphological traits for sex
population by odontometric study of permanent determination. J Forensic Sci 2006;51(4):729-35.
maxillary canine. J Forensic Res 2014;5(2):1000224. 8. Hu KS, Koh KS, Han SH, Shin KJ, Kim HJ. Sex
4. Movahedian N, Abedi S, Memarpour M. Comparison of determination using nonmetric characteristics of the
the Demirjian and root resorption methods to estimate mandible in Koreans. J Forensic Sci 2006;51(6):1376-82.
20
9. Garn SM, Cole PE, Van Alstine WL. Sex discriminatory 26. da Silva PR, Lopes MC, Martins-Filho IE, Haye Biazevic
effectiveness using combinations of root lengths and MG, Michel-Crosato E. Tooth crown mesiodistal
crown diameters. Am J Phys Anthropol 1978;50(1):115-8. mea surements for the determination of sexual
10. Pretty IA, Sweet D. A look at forensic dentistry–Part 1: dimorphism across a range of populations: A systematic
The role of teeth in the determination of human identity. review and meta-analysis. J Forensic odontostomatol
Br Dent J 2001;190(7):359-66. 2019;37(1):2-19.
11. Capitaneanu C, Willems G, Thevissen P. A systematic 27. Bakkannavar SM, Manjunath S, Nayak VC, Kumar GP.
review of odontological sex estimation methods. J Canine index–A tool for sex determination. Egypt J
Forensic Odontostomatol 2017;35(2):1-19. Forensic Sci 2015;5(4):157-61.
12. Angadi PV, Hemani S, Prabhu S, Acharya AB. Analyses of 28. Zorba E, Vanna V, Moraitis K. Sexual dimorphism of root
odontometric sexual dimorphism and sex assessment l e n g t h o n a G r e e k p o p u l a t i o n s a m p l e . Ho m o
accuracy on a large sample. J Forensic L eg Med 2014;65(2):143-54.
2013;20(6):673-7. 29. Lähdesmäki R, Alvesalo L. Root lengths in 47, XYY
13. Kazzazi SM, Kranioti EF. A novel method for sex males’ permanent teeth. J Dent Res. 2004;83:771-75.
estimation using 3D computed tomography models of 30. Ha r r i s E F, C o u c h J r W M . T h e r e l a t i v e s e x u a l
tooth roots: A volumetric analysis. Arch Oral Biol dimorphism of human incisor crown and root
2017;83:202-208. dimensions. Dent Anthropol 2006;19(3):87-95.
14. Martins Filho IE, Lopez-Capp TT, Biazevic MG, Michel- 31. G o v i n d a r a m D, B h a r a n i d h a r a n R , R a m y a R ,
Crosato E. Sexual dimorphism using odontometric Rameshkumar A, Priyadharsini N, Rajkumar K. Root
indexes: Analysis of three statistical techniques. J Forensic Length: As a determinant tool of sexual dimorphism in an
Leg Med 2016;44:37-42. e t h n i c Ta m i l p o p u l a t i o n . J Fo r e n s i c D e n t S c i
15. Nadendla LK, Paramkusam G, Pokala A, Devulapalli RV. 2018;10(2):96-100.
Identification of gender using radiomorphometric 32. Vossoughi M, Movahhedian N, Ghafoori A. The impact
measurements of canine by discriminant function of age mimicry bias on the accuracy of methods for age
analysis. Indian J Dent Res 2016;27(1):27-31. estimation based on Kvaal’s pulp/tooth ratios: a
16. Davoudmanesh Z, Shariati M, Azizi N, Yekaninejad S, bootstrap study. Int J Legal Med 2022;136(1):269-78.
Ho z h a b r H , Ka d k h o d a e i - O l i a d a r a n i F. S e x u a l 33. Baglivo M, Winklhofer S, Hatch GM, Ampanozi G, Thali
dimorphism in permanent canine teeth and formulas for MJ, Ruder TD. The rise of forensic and post-mortem
sex determination. Biomed Res India 2017;28:2773-7. radiology—analysis of the literature between the year
17. Macaluso PJ Jr. Sex discrimination potential of 2000 and 2011. J Forensic Radiol Imaging 2013;1(1):3-9.
permanent maxillary molar cusp diameters. J Forensic 34. Sherrard JF, Rossouw PE, Benson BW, Carrillo R,
OdontoStomatol 2010;28(1):22: 31-22: 31. Buschang PH. Accuracy and reliability of tooth and root
18. Acharya AB, Mainali S. Univariate sex dimorphism in the lengths measured on cone-beam computed tomographs.
Nepalese dentition and the use of discriminant functions Am J Orthod Dentofacial Orthop 2010;137(4 Suppl):S100-08.
in gender assessment. Forensic Sci Int 2007;173(1):47-56. 35. Stratemann SA, Huang JC, Maki K, Miller AJ, Hatcher
19. Capitaneanu C, Willems G, Jacobs R, Fieuws S, Thevissen DC. Comparison of cone beam computed tomography
P. Sex estimation based on tooth measurements using imaging with physical measures. Dentomaxi#ofac Radiol
panoramic radiographs. Int J Legal Med 2017;131(3):813-21. 2008;37(2):80-93.
20.Mallya SM, Lam E. White and Pharoah's Oral Radiology: 36. Kim SY, Lim SH, Gang SN, Kim HJ. Crown and root
Principles and Interpretation. 8th ed. St. Louis: Elsevier lengths of incisors, canines, and premolars measured by
Health Sciences; 2018. cone-beam computed tomography in patients with
21. Sarment DP, Christensen AM. The use of cone beam malocclusions. Korean J Orthod 2013;43(6):271-8.
computed tomography in forensic radiology. J Forensic 37. Acharya AB, Angadi PV, Prabhu S, Nagnur S. Validity of
Radiol Imaging 2014;2(4):173-81. the mandibular canine index (MCI) in sex prediction:
22. Paknahad M, Dokohaki S, Khojastepour L, Shahidi S, Reassessment in an Indian sample. Forensic Sci Int
Haghnegahdar A. A Radio-Odontometric analysis of 2011;204(1-3):207.e1-4.
sexual dimorphism in first molars using cone-beam 38. Khangura RK, Sircar K, Singh S, Rastogi V. Sex
computed tomography. Am J Forensic Med Pathol determination using mesiodistal dimension of permanent
2022;43(1):46-51. maxillary incisors and canines. J Forensic Dent Sci
23. Castillo L, Castro AM, Lerma C, Lozada D, Moreno F. 2011;3(2):81-5.
Mesiodistal and bucolingual dental diameters in a group 39. Lambrechts P, Braem M, Vuylsteke-Wauters M, Vanherle
of mixed ethnicity population in Cali, Colombia. Rev G. Quantitative in vivo wear of human enamel. J Dent Res
Estomat 2012;20(1):16-22. 1989;68(12):1752-4.
24. Kazzazi SM, Kranioti EF. Sex estimation using cervical 40.Dumančić J, Scott GR, Savić Pavičin I, Anić-Milošević S,
dental measurements in an archaeological population Me d a n č i ć N , B r k i ć H . C a n i n e C r o w n S e x u a l
from Iran. Archaeol Anthrop Sci 2018;10:439-48. Dimorphism in a Sample of the Modern Croatian
25. Prabhu S, Acharya AB. Odontometric sex assessment in Population. Dent J (Basel) 2023;11(7):175.
Indians. Forensic Sci Int 2009;192(1-3):129.e 1-5.
21
Machine learning assisted 5-part tooth segmentation

method for CBCT-based dental age estimation in
adults
Copyright © 2024 International Organization ABSTRACT

Background: The utilization of segmentation method using
Rizky Merdietio Boedi 1,2, volumetric data in adults dental age estimation (DAE) from
Simon Shepherd3, Fahmi cone-beam computed tomography (CBCT) was further
Oscandar 4, Ademir Franco 5, expanded by using current 5-Part Tooth Segmentation ( SG5t )
method. Additionally, supervised machine learning modelling
Scheila Mânica1
—namely support vector regression (SVR) with linear and
polynomial kernel, and regression tree — was tested and
1 Centre of Forensic and Legal
Medicine and Dentistry, University of
compared with the multiple linear regression model.
Dundee, Dundee, United Kingdom. Material and Methods: CBCT scans from 99 patients aged
2Department of Dentistry, Faculty of
between 20 to 59.99 was collected. Eighty eligible teeth
Medicine, Universitas Diponegoro,
Semarang, Indonesia. 3 Department including maxillary canine, lateral incisor, and central incisor
of Oral Surgery, School of Dentistry, were used in this study. Enamel to dentine volume ratio, pulp
University of Dundee, Dundee, United to dentine volume ratio, lower tooth volume ratio, and sex was
Kingdom. 4Department of Oral and
Maxillofacial Radiology - Forensic utilized as independent variable to predict chronological age.
Odontology, Faculty of Dentistry, Results: No multicollinearity was detected in the models. The
Universitas Padjadjaran, Bandung, best performing model comes from maxillary lateral incisor
Indonesia. 5 Division of Forensic 2
Dentistry, Faculdade São Leopoldo using SVR with polynomial kernel ( Radj = 0.73). The lowest
Mandic, Campinas, Brazil error rate achieved by the model was given also by maxillary
lateral incisor, with 4.86 years of mean average error and 6.05
years of root means squared error. However, SG5t demands a
Corresponding author: complex approach to segment the enamel volume in the crown
[email protected] section and a lengthier labour time of 45 minutes per tooth.
INTRODUCTION
The authors declare that they The emergence of cone beam computed tomography (CBCT)
have no conflict of interest. systems in dentistry has opened new possibilities in diagnostics
and image analysis. Utilizing CBCT for patient diagnosis
provides dentists with a clear and detailed visualization of the
dentomaxillofacial feature. 1 The improved anatomical
KEYWORDS visualization provided by CBCT leads to more predictable
post-operative outcomes and safer clinical practices.2 On the
Forensic Dentistry, image analysis front, researchers have used volumetric
Age Determination by Teeth, information to deeply analyze anatomical structures,3 human
Cone Beam Computed growth,4 and other changes that happens in the craniofacial
Tomography, region.5 This volumetric information has also proven to be
highly valuable in various dental analyses, including dental age
Supervised Machine Learning
estimation performed by forensic odontologists.6
The importance of volumetric information provided by the
J Forensic Odontostomatol CBCT allows forensic odontologists to conduct a thorough
2024. Apr;(42): 1-22:29 investigation of regressive changes in teeth, which is essential
ISSN :2219-6749 in dental age estimation of adults when dental growth has
DOI: doi.org/10.5281/zenodo.11061543 ceased.7 Recent research has expanded the capabilities of
volumetric measurement for dental a ge estimation by
22
segmenting it into various anatomical regions to exposure time between 1.7 to 8.7 seconds. The
enhance the reliability of predictive models, voxel size used in this study were 0.125mm3 (n =
namely: Lower Tooth Volume Ratio (LTVR), 14), 0.2mm3 (n = 36), 0.3mm3 (n = 43), and 0.4 mm3
derived from the lower root chamber volume (n =6). In relation to the respective fields of view,
(LRCV) and lower hard tissue volume (LHTV);8 the smaller voxel size was taken for the small field
Pulp to Dentine Volume Ratio (PDVR), derived of view (250 x 250 x 250), whilst the 0.3 and 0.4
from the pulp chamber volume (PCV) of the mm3 size was used for the 400 x 400 x 400 scans.
crown and dentine volume (DV); and Enamel to Importantly, no patient was subjected to
Dentine Volume Ratio (EDVR), derived from radiation specifically for the study, as the sample
enamel volume (EV) and DV.9 All these ratios collection was retrospective and taken from an
can be included in a single predictive model. existing image database of CBCT scans acquired
However, the introduction of more independent for clinical purposes.
variables into a predictive linear model carries The inclusion sample for the current study
the risk of multicollinearity, 10 lower model consists of maxillary teeth with fully erupted,
reliability,11 and inflated R2.12 closed apex, and a visible cemento-enamel
These problems were argued that it can be junction (CEJ). Teeth with restorations, caries,
solved by calculating the variance inflation impaction, resorption, associated tumors, cysts,
2
factor (VIF) and adjusted R2 ( Radj ) of the model. pulp calcification, visible accessory root canals, or
Moder n solutions — such a s Super vised any inter ventions that may affect tooth
Machine Learning (SML) — have also been formation or structure were excluded. The total
repor ted to reduce the multicol linearity sample size for this study includes 240 teeth
problem 13 while minimizing the error-rate of a (Table 1) consisting of maxillary canines (C, n =
dental age estimation model when compared to 80), lateral incisors (Li, n = 80), and central
the traditional multiple linear regression incisors (Ci, n = 80).
model.14
Considering these aspects, the aim of this Table 1. Total sample size of each examined
research is twofold: (1) to examine the predictive tooth for each sex and age range.
performance of the 5-Part Tooth Segmentation M F
( SG5t ) method by combining the variables Age (Years)
introduced by Merdietio Boedi et al. 8-9 and (2) C C Li Ci Li Ci
assessing the effectiveness of SML — namely, 20-24.99 5 5 5 5 5 5
support vector regression (SVR) and regression
tree (RT) — to improve the reliability of the 25-24.99 5 5 5 5 5 5
variables in predicting chronological age (CA). 30-34.99 5 5 5 5 5 5
MATERIAL AND METHODS 35-35.99 5 5 5 5 5 5

40-44.99 5 5 5 5 5 5
Data acquisition
This observational cross-sectional study received 45-49.99 5 5 5 5 5 5
ethical approval from the Research Ethics 50-54.99 5 5 5 5 5 5
Committee of Universitas Padjajaran (Approval
No. 899/UN6.KEP/EC/2021). The sample used in 55-59.99 5 5 5 5 5 5
this study was acquired from Merdietio Boedi et M = Male, F = Female, C = Maxillary Canine, Li =
al.'s (2023) and was adapted and reanalyzed for Maxillary Lateral Incisor, Ci = Maxillary Central Incisor
the current study.9 The sample consisted of 45
males and 54 females of Bandung, Indonesian Sample processing
origin, aged between 20 and 59.99 years (Mean The volumetric information of each tooth was
CA 40.69 ± 11.23). CBCT scans were acquired at segmented into 5 distinct components: PCV, DV,
Universitas Padjajaran Dental Hospital using the E V, L RC V, a n d L H T V ( Fi g . 1 ) . T h e s e
In s t r u m e n t a r i u m Dental OP300 measurements were carried out using ITK-SNAP
(Instrumentarium Dental, Tuusula, Finland) with ver. 3.8 (ITK-SNAP, UPenn & UNC, USA).15 The
patient-specific exposure settings, including 85 primary settings in ITK-SNAP were consistently
kV, tube current ranging from 3 to 8 mA, and an configured as follows throughout the
measurement process: 3D brush settings ON, all
23
label opacities were set to 35, "initialize with the specific tooth's volumetric ratio calculation
c u r r e n t s e g m e n t a t i o n" o p t i o n O N , a n d (i.e., E DV R Li represents the EDVR calculation
"thresholding" option was chosen in the for tooth Li). Generally, the meantime taken for
segmentation mode. All volumetric measurements the SG5t method was 45 minutes for each tooth.
are conducted through the built-in region of
interest (ROI) function within ITK-SNAP. Figure 1. Illustration of the 5-Part Tooth
The SG5t method star ts with the volume Segmentation method. The demarcation between
calculation for the crown region (i.e., PCV, DV, the crown area and root area is defined using the
and EV) followed the protocol described by highest or most apical point of the cemento-
Merdietio Boedi et al. in their 2023 work,9 while enamel junction in the sagittal section of the
for the root region (i.e., LRCV and LHTV), the CBCT.
procedure follows to their 2022 work. 8 The
volumetric measurement started with the
calculation of PCV, followed by the entire crown
volume, demarcated by the highest or most apical
C E J i n t h e s a g i t t a l v i e w o f t h e C B C T.
Subsequently, the volumetric information was
partitioned into DV and EV. Lastly, the calculation
extended to LRCV and LHTV (Fig. 2). In the
separation of the crown and root regions, the
regions of interest (ROI) were ensured to overlap,
ensuring the continuity of segmentation. All the
calculated volume will then be converted to 3
volumetric ratios:
EV PC V L RC V
EDVR ( ), PDVR ( ), and LTVR ( ).
DV DV LHTV
The superscripted letter notation corresponds to
Figure 2. 5-Part Tooth Segmentation sequence in ITK-SNAP three-dimensional rendering. A: Pulp Chamber
Volume (red), B: Whole Crown Volume (green), C: Separation between Dentine (green) and Enamel (light blue)
Volume, D: Lower root chamber volume (blue), E: Lower hard tissue volume (yellow), F: Visualization of the
cross-sectioned volumetric data depicting the separation between 5 volumetric information.
Data analysis odontologist with 7 years of experience, with

Inter- and intra-observer variabilities were calculations repeated after a two-week interval
calculated using the Intraclass Correlation between observations. Furthermore, the inter-
Coefficient (ICC). Inter-observer agreement was observer analysis was accomplished by comparing
determined by the first author— a forensic the observations of the first author with a second
24
observer — an oral and maxillofacial radiologist = 0.57), and PDV R Li to LT V R Li (r = 0.38),

with 3 years of experience — recruited for the respectively.
study. These analyses were based on a randomly
selected set of 20 CBCT scans from unpublished Table 2. Intra and Inter-Class Correlation
data. Coefficient between and within observers.
Initial analysis was performed by calculating ICC
Pearson’s correlation coefficient ( r ) cross- Variables
tabulation for each variable. Then, three Intra Inter
predictive models, namely MLR, SVR, and RT, PCV 0.78 0.78
were developed using CA as the dependent
DV 0.82 0.81
variable, volumetric ratios as independent
variables, and sex as a covariate. All data analysis EV 0.7 0.76
wa s conducted using R (version 3.4.1, R LRCV 0.92 0.92
Foundation for Statistical Computing, Vienna,
Austria) with the caret extension. Model training LHTV 0.9 0.73
parameters in caret were set using the "repeated
cross-validation" training method with 5-fold
cross -validation and 2 repetitions. The Table 3. Pearson's correlation coefficient values
hyperparameters for both SVR and RT were set for each variable.
to center and scale for the preProcess function, Tooth EDVR PDVR LTVR CA
and the tuneLength parameter was set to 5. Two
SVR kernel will be tested: linear (SVR-L) and EDVR - 0.57 0.51 -0.67
polynomial (SVR-Poly) kernels. PDVR 0.57 - 0.44 -0.68
2
Model reliability was assessed using Radj , root C
mean squared error (RMSE), and mean absolute LTVR 0.51 0.44 - -0.66
e r r o r ( M A E ) . Ad d i t i o n a l l y, p e r f o r m a n ce CA -0.67 -0.68 -0.66 -
evaluation included the use of the variance
inflation factor (VIF) to detect multicollinearity, EDVR - 0.47 0.60 -0.61
with VIF > 5 indicating multicollinearity among
PDVR 0.47 - 0.38 -0.65
independent variables. To ensure reproducibility Li
in the randomization, the pseudo-random LTVR 0.60 0.38 - -0.73
number generator in R was controlled using
set.seed = 30. CA -0.61 -0.65 -0.73 -
EDVR - 0.55 0.47 -0.60

RESULTS
Volumetric measurements demonstrated good PDVR 0.55 - 0.51 -0.65
consistency, with intra- and inter-obser ver Ci
LTVR 0.47 0.51 - -0.61
agreement reported in Table 2.
All r-values in the cross-tabulation (Table 3) were CA -0.60 -0.65 -0.61 -
statistically significant. Negative correlations
C = Maxillary Canine, = Li = Maxillary Lateral Incisor, Ci =
were observed between CA and independent
Maxillary Central Incisor, EDVR = Enamel to Dentine
variables, indicating that as CA increases, the Volume Ratio, PDVR = Pulp to Dentine Volume Ratio,
ratio measurements reflecting regressive changes LTVR = Lower Tooth Volume Ratio
in the observed tooth decrease. The highest
negative correlation between CA and an MLR models were calculated as follows:
independent variable was observed in PDV RC (r = y = 69.83 − 17.69E DV RC − 168.59PDV RC − 132.77LT V RC
-0.68). Positive r-values between independent y = 62 − 164.06PDV R Li − 196L T V R Li
variables indicate that both variables change in y = 70.83 − 20.57E DV RCi − 164.47PDV RCi − 132.28LT V RCi
the same direction; in this case, all independent where y is the estimated dental age. The sex
variables decrease over time. The highest and covariate did not exhibit a significant contribution
lowest positive correlations between independent to the MLR models, and similarly, PDV R Li did not
variables were observed for PDV RC to E DV RC (r show significance in predicting CA. All VIF values
25
were below 5, indicating no multicollinearity among the results evade the age mimicry phenomenon,
the independent variables. where the final error rate may be skewed toward a
A comparison of the models revealed that SVR- certain age group or population reference. 18
Poly consistently achieved superior performance in Secondly, the proposed methodology must
predicting CA compared to the other modelling incorporate inter- and intra-observer error,
approaches, as evidenced in Table 4. Notably, the certainly with a proper quantification analysis to
2
highest Radj values also translated to a lower MAE ensure methodological reproducibility. For
and RMSE metrics. The third-degree SVR-Poly instance, when dealing with ordinal data (e.g.,
model in C utilizing all independent variables and staging, atlases), Cohen's Kappa should be
sex as predictor yielded the lowest MAE (4.97 employed for analysis, while continuous data
2
years) and RMSE (6.37 years), with Radj value of (e.g., measurements) necessitate the use of ICC.19
0.71. Conversely, the Ci SVR-L model employing Thirdly, the reliability of the model is signified by
all independent variables produced the highest the error-rate that should be shown along with
MAE (6.6 years) and RMSE (7.97 years), with the the conversion of the age-related dental data, and
2
lowest Radj value of 0.55. this can be achieved through scoring systems,
diagrams, or models. Importantly, the error-rate
should be calculated from a different dataset than
Table 4. Model performance comparison for
the one used for model creation to prevent
each tooth. All error values were reported from
overfitting — a scenario that occurs when a
the cross-validated data using 5-fold cross-
model is too closely tailored to the training data
validation with 2 repetitions.
and may not perform well on new, unseen data.
Models The most straightforward approach to prevent
Tooth Metrics
MLR SVR-L SVR-Poly RT overfitting is to separate the dataset into training
and testing subsets. For instance, considering a
2
Radj 0.66 0.65 0.71 0.62 dataset denoted as x , the training data could use
the x − i portion of the data, while the model
C MAE 5.42 5.23 4.97 5.77
derived from the training dataset is evaluated
RMSE 6.86 6.8 6.37 7.18 against the i data. Another method — as
2
employed in this research — involves k-fold
Radj 0.69 0.7 0.73 0.69
cross-validation. This technique divides the data
Li MAE 5.14 5.15 4.86 5.28 into k parts or folds, utilizing k − 1 segments for
training the dataset and reserving the untrained
RMSE 6.46 6.42 6.05 6.67 fold for testing. This process is repeated k times,
2
Radj 0.56 0.55 0.68 0.56 with each fold serving as the test data exactly
once.
Ci MAE 6.4 6.6 5.27 6.26 CBCT can be used for dental staging,20 alveolar
bone loss measurement,21 metric assessment,22
RMSE 7.7 7.97 6.67 7.7
and specifically for volumetric measurement
C = Maxillary Canine, = Li = Maxillary Lateral Incisor, Ci = application. Volumetric data has shown the
Maxillary Central Incisor, highest reliability within the adult population.6
2 = Adjusted R2, MAE = Mean Average Error, RMSE = Root
Radj
This is primarily attributed to the absence of
Mean Squared Error, MLR = Multiple Linear Regression, SVR-L =
dental maturation process, a method commonly
Support Vector Regression with Linear Kernel, SVR-Poly =
Support Vector Regression with Polynomial Kernel, RT = e m p l o y e d f o r c h i l d r e n a n d a d o l e s ce n t s .
Regression Tree Regressive dental changes are subtle when
compared to dental maturation, necessitating
DISCUSSION data granularity to ensure even the smallest shifts
Although studies on dental age estimation might in adult tooth structure are captured accurately.
use different approaches, several key objectives Unlike conventional radiographs — which
need to be addressed by researchers to establish a provide only two-dimensional measurements,
robust methodology.16 Firstly, an adequate sample volumetric data captures these details through
size is essential, not only in the quantity of CBCT voxels to approximate radiographic
analyzed data but also the distribution across age structure volume, acquired from the height,
group and sex.17 This key objective ensures that width, and depth of the images.24 Therefore, the
26
accuracy of the volumetric method has much to by using geometric approximations of upper
recommend it for the assessment of the dental central incisors to reduce operating time.32 A
age of adults. direct comparison of R2 values with the current
This study utilizes volumetric approximations of methodology reveals a higher R2 value of 0.71 for
a tooth obtained through CBCT scans employing the Li SVR-Poly model when compared to Zhang
various voxel sizes. As noted in previous studies, et al. (R2 = 0.42), Pinchi et al. (R2 = 0.58), or similar
inconsistencies in measuring or observing age- population study conducted by Anjani et al. (R2 =
related variables can introduce errors in the final 0.58). However, it is important to note that the
estimated age.25 Lee et al. recommend a 0.2mm3 SG5t a p p r o a c h i n v o l v e s a m o r e co m p l e x
voxel for optimal dental structure evaluation.27 segmentation process, and hence a greater time
Even so, Adisen et al. concluded that different commitment, for each part of the tooth to
voxel sizes have not significantly improved overall achieve enhanced model performance.
result.26 While smaller voxel sizes offer increased The error rate in adult dental age estimation was
clarity, methods utilizing CBCT volumetric found to be higher and possible causes include
information can still function appropriately with external factors, regressive dental changes, and
larger voxel sizes with good quality scans. population-specific influences.33 For instance,
Whenever feasible, uniform sampling if images individuals with bruxism may experience a higher
based on the same voxel size should be preferred rate of attrition, resulting in reduced enamel
to standardize the methodological settings. volume. Additionally, error rates may var y
The potential variability in CBCT-based significantly depending on the specific population,
measurements was seen by Yang et al. whom as seen in the Indonesian populations. Marroquin
conducted a comparison of pulp size obtained et al. (2018) conducted pulp/tooth volume
from an Archimedes' principle experiment to calculations on two distinct samples —Colombian
CBCT-derived volumes, yielding an acceptable and Malaysian— concluding the necessity of
error rate of ±7.6%.28 Further work by Star et al. population-specific methods for age estimation
revealed a more substantial discrepancy between due to discriminatory results.34 Du et al. (2021)
software measurements and the gold standard, observed significant differences in error rates
with differences of up to 21% and 16% for pulp when a model derived from Chinese samples was
and tooth volume, respectively.29 Adding to this applied to Black Americans, because the original
complexity, previous research has identified a MAE of 7.9 years Increased to 14.04 years for the
significant correlation between individual stature later population.35 Hence, the model performance
and tooth volume.30 This implies that larger reported in this study may be population-specific,
individuals may tend to have larger teeth, underscoring the importance of methodological
potentially introducing bias into the volumetric validation when a model is applied to a new or
measurements. This study used a ratio as its different populations.
independent variable, mitigating the influence of The limitations of this study are two-fold and
both intrinsic variability in CBCT measurements linked primarily to sample size and methodological
and the correlation between stature and tooth constraints. First, the sample size was 80 per tooth
size. position, above the required minimum sample size
This study employed a segmentation method to of 67 from G*Power analysis for a-priori sample
achieve higher model performance by discerning size determination with effect size of 0.3, Power
multiple volumetric regressive changes in anterior 0.95, and 4 estimated predictors. However, a
teeth using CBCT images. Previous studies have bigger sample size to achieve a lower effect size
primaril y utilized tooth volumetric data, may result in better age estimation performance
conducting this approach across various that more accurately representing dental regressive
populations,7 each with their own modifications changes in adults.36 Moreover, a non-uniform
aimed at either improving model performance or distribution of voxel sizes was obtained during
adopting a more user-friendly approach. For sampling process — a situation justified by the
instance, (1) Zhang et al. utilized only the enamel institutional protocol that require patient-specific
and pulp chamber of an impacted third molar to modeling of energy parameters for ima ge
eliminate the influence of external factors acquisition. Second, the inclusion of EV
affecting the rate of enamel attrition,31 and (2) calculation affected the operating time of the
Pinchi et al. simplified volumetric measurements segmentation, approximately 45 minutes, and with
27
a steeper learning curve. In contrast, similar the maxillary canine affording the best overall
segmentation study that divided the tooth performance in comparison to other anterior
volumetric information into four segments maxillary teeth examined. Nonetheless, this
without EV calculation took only 10 minutes.8 improvement in model performance comes with
This drawback certainly comes with a better greater labor time costs and a steep learning
overall performance. As reported in the previous curve for methodology application. Future
study, the current SG5t approach shows an research might consider the application of this
increase of R2 between 0.1 to 0.23. Hence, the segmentation methodology to additional teeth,
current SG5t is better suited for evaluating particularly molars, given their relatively greater
individual dental identification cases, where p r o t e c t i o n w i t h i n t h e o r a l c a v i t y, t h u s
time constraints are presumably more lenient potentially enhancing their usefulness in disaster
compared to mass disaster cases. scenarios.
CONCLUSION ACKNOWLEDGMENTS
The SG5t approach combined with SVR-Poly This work was supported by the Ministry of
model gives an overall better performance when Education and Culture, Universitas Diponegoro
compared to other modelling approaches, with No.497/UN7.P/HK/2021.
REFERENCES
1. Ludlow JB, Timothy R, Walker C, Hunter R, Benavides 11.Ioannidis JP. Why most published research findings are false.
E, Samuelson DB, Scheske MJ. Effective dose of dental PLoS Med 2005;2:e124.
CBCT-a meta analysis of published data and additional data for 12.Helland IS. On the Interpretation and Use of R2 in Regression
nine CBCT units. Dentomaxi#ofac Radiol 2015;44:20140197. Analysis. Biometrics 1987;43:61-9.
2. Wang X, Shujaat S, Shaheen E, Ferraris E, Jacobs R. 13.Chan JYL, Leow SMH, Bea KT, Cheng WK, Phoong
Trueness of cone-beam computed tomography-derived sku# models SW, Hong ZW, Chen YL. Mitigating the Multico#inearity
fabricated by different technology-based three-dimensional Problem and Its Machine Learning Approach: A Review.
printers. BMC Oral Health 2023;23:397. Mathematics 2022;10:1283.
3. Lenza MG, Lenza MM, Dalstra M, Melsen B, Cattaneo 14.Rocha LT, Ingold MS, Panzarella FK, Santiago BM, de
PM. An analysis of different approaches to the assessment of upper Oliveira RN, Bernardino ID, Makeeva I, Junqueira JLC,
airway morphology: a CBCT study. Orthod Craniofac Res Manica S, Franco A. Applicability of Wi#ems method for age
2010;13:96-105. estimation in Brazilian children: performance of multiple linear
4. Anderson PJ, Yong R, Surman TL, Rajion ZA, Ranjitkar S. regression and artificial neural network. Egypt J Forensic Sci
Application of three-dimensional computed tomography in 2022;12:1-9.
craniofacial clinical practice and research. Aust Dent J 15.Yushkevich PA, Gerig G. ITK-SNAP: An Intractive
2014;59:174-85. Medical Image Segmentation Tool to Meet the Need for Expert-
5. Van Dessel J, Nicolielo LF, Huang Y, Coudyzer W, Salmon Guided Segmentation of Complex Medical Images. IEEE Pulse
B, Lambrichts I, Jacobs R. Accuracy and reliability of different 2017;8:54-7.
cone beam computed tomography (CBCT) devices for structural 16.Bjork MB, Kvaal SI. CT and MR imaging used in age
analysis of alveolar bone in comparison with multislice CT and estimation: a systematic review. J Forensic Odontostomatol
micro-CT. Eur J Oral Implantol 2017;10:95-105. 2018;36:14-25.
6. Merdietio Boedi R, Shepherd S, Manica S, Franco A. 17.Schmeling A, Geserick G, Reisinger W, Olze A. Age
CBCT in dental age estimation: A systematic review and meta estimation. Forensic Sci Int 2007;165:178-81.
analysis. Dentomaxi#ofac Radiol 2022;51:20210335. 18.Vossoughi M, Movahhedian N, Ghafoori A. The impact
7. Barbosa MG, Franco A, de Oliveira RDB, Mamani MP, of age mimicry bias on the accuracy of methods for age estimation
Junqueira JLC, Soares MQS. Pulp volume quantification based on Kvaal's pulp/tooth ratios: a bootstrap study. Int J Legal
methods in cone-beam computed tomography for age estimation: A Med 2022;136:269-78.
critical review and meta-analysis. J Forensic Sci 2023;68:743-56. 19.Aldridge VK, Dovey TM, Wade A. Assessing Test-Retest
8. Merdietio Boedi R, Shepherd S, Oscandar F, Manica S, Reliability of Psychological Measures Persistent Methodological
Franco A. Regressive changes of crown-root morphology and Problems. Eur Psychol 2017;22:207-18.
their volumetric segmentation for adult dental age estimation. J 20.Cantekin K, Sekerci AE, Buyuk SK. Dental computed
Forensic Sci 2022;67:1890-8. tomographic imaging as age estimation: morphological analysis of
9. Merdietio Boedi R, Shepherd S, Oscandar F, Manica S, the third molar of a group of Turkish population. Am J Forensic
Franco A. 3D segmentation of dental crown for volumetric age Med Pathol 2013;34:357-62.
e s t i m a t i o n w i t h C B C T i m a g i n g. In t J L e ga l Me d 21.Daluz A, Saliba-Serre B, Foti B, Lan R. Age estimation
2023;137:123-30. &om alveolar bone loss, re-evaluation of Ruquet's method.
10.Alin A. Multico#inearity. Wiley Interdiscip Rev: Comput Stat Forensic Sci Med Pathol 2023:1-10. Available from: https://
2010;2:370-4. link.springer.com/article/10.1007/s12024-023-00617-2
[cited 10 November 2023].
28
22.Salemi F, Farhadian M, Sabzkouhi BA, Saati S, Nafisi N. 30.Hatipoğlu FP, Arıcıoğlu B, Hatipoğlu Ö, Köse TE,
Age estimation by pulp to tooth area ratio in canine teeth using cone- Günaçar DN. Prediction of root canal lengths and pulp volume of
beam computed tomography. Egypt J Forensic Sci 2020;10:1-8. the maxillary permanent first molar based on stature, crown
23.Helmy MA, Osama M, Elhindawy MM, Mowafey B. diameters, and facial morphometry. Anat Sci Int 2023;98:454-62.
Volume analysis of second molar pulp chamber using cone beam 31.Zhang ZY, Yan CX, Min QM, Li SQ, Yang JS, Guo YC,
computed tomography for age estimation in Egyptian adults. J Jin WF, Li LJ, Xing PF, Li J. Age estimation using pulp/enamel
Forensic Odontostomatol 2020;3:25-34. volume ratio of impacted mandibular third molars measured on
24.Spin-Neto R, Gotfredsen E, Wenzel A. Impact of voxel CBCT images in a northern Chinese population. Int J Legal Med
size variation on CBCT-based diagnostic outcome in dentistry: a 2019;133:1925-33.
systematic review. J Digit Imaging 2013;26:813-20. 32.Pinchi V, Pradella F, Buti J, Baldinotti C, Focardi M,
25.25. Abdul Rahim AH, Davies JA, Liversidge HM. Norelli GA. A new age estimation procedure based on the 3D
Reliability and limitations of permanent tooth staging techniques. CBCT study of the pulp cavity and hard tissues of the teeth for
Forensic Sci Int 2023;346:111654. forensic purposes: A pilot study. J Forensic L eg Med
26.Adisen MZ, Keles A, Yorubulut S, Nalcaci R. Age 2015;36:150-7.
estimation by measuring maxi#ary canine pulp/tooth volume 33.Phulari RGS, Dave EJ. Evolution of dental age estimation
ratio on cone beam CT images with two different voxel sizes. Aust methods in adults over the years &om occlusal wear to more
J Forensic Sci 2018;52:71-82. sophisticated recent techniques. Egypt J Forensic Sci 2021;11:1-14/
27.Lee S-M, Oh S, Kim J, Kim Y-M, Choi Y-K, Kwak HH, 34.Marroquin TY, Karkhanis S, Kvaal SI, Kruger E,
Kim Y-I. Age estimation using the maxi#ary canine pulp/tooth Tennant M. Overcoming population differences for dental age
ratio in Korean adults: A CBCT buccolingual and horizontal estimation in adults through pulp/tooth volume calculations: a
section image analysis. J Forensic Radiol Imaging 2017;9:1-5. pilot study. Aust J Forensic Sci 2018;52:500-7.
28.Yang F, Jacobs R, Willems G. Dental age estimation 35.Du H, Li G, Zheng Q, Yang J. Population-specific age
through volume matching of teeth imaged by cone-beam CT. estimation in Black Americans and Chinese people based on pulp
Forensic Sci Int 2006;159:S78-S83. chamber volume of first molars &om cone beam computed
29.Star H, Thevissen P, Jacobs R, Fieuws S, Solheim T, tomography. Int J Legal Med 2022;136:811-9.
Willems G. Human dental age estimation by calculation of pulp- 36.Sgheiza V, Liversidge H. Reference and target sample age
tooth volume ratios yielded on clinica#y acquired cone beam distribution impacts between model types in dental
computed tomography images of monoradicular teeth. J Forensic d e v e l o p m e n t a l a g e e s t i m a t i o n . In t J L e g a l Me d
Sci 2011;56:S77-82. 2022;137:383-93.
29
Artificial intelligence in age and sex determination

using maxillofacial radiographs: A systematic review

ABSTRACT
Sraddha Singh1, Bhoopendra In the past few years, there has been an enormous increase in
Singha1, Shanu Kumar1 the application of artificial intelligence and its adoption in
multiple fields, including healthcare. Forensic medicine and
1 Department of Forensic medicine & forensic odontology have tremendous scope for development
Toxicology, Rajendra Institute of using AI. In cases of severe burns, complete loss of tissue,
Medical Science, Ranchi, India complete or partial loss of bony structure, decayed bodies,
mass disaster victim identification, etc., there is a need for
prompt identification of the bony remains. The mandible, is
the strongest bone of the facial region, is highly resistant to
Corresponding author: undue mechanical, chemical or physical impacts and has been
[email protected] widely used in many studies to determine age and sexual
dimorphism. Radiographic estimation of the jaw bone for age
and sex is more workable since it is simple and can be applied
equally to both dead and living cases to aid in the identification
The authors declare that they process. Hence, this systematic review is focused on various AI
have no conflict of interest. tools for a ge and sex determination in maxillofacial
radiographs. The data was obtained through searching for the
articles across various search engines, published from January
2013 to March 2023. QUADAS 2 was used for qualitative
synthesis, followed by a Cochrane diagnostic test accuracy
review for the risk of bias analysis of the included studies. The
KEYWORDS results of the studies are highly optimistic. The accuracy and
precision obtained are comparable to those of a human
Artificial Intelligence, examiner. These models, when designed with the right kind of
Convolutional Neural data, can be of tremendous use in medico legal scenarios and
Networks, disaster victim identification.
Artificial Neural Networks,
INTRODUCTION
Machine Learning, Forensic odontology deals with the proper administration and
Deep learning, evaluation of dental evidence in criminal or civil legal
Mandible, proceedings in the interest of justice.1 It involves identification
Forensic Odontology, by assessing and analysing the unique structures in the oral
cavity.2 Identification is the most important aspect of forensic
OPG, and the application of forensic odontology involves teeth and
Lateral Cephalogram j a w b o n e s to a i d i n t h e p r o c e s s o f i d e n t i f i c a t i o n .
Identification through age and sex of bony remains is viable
due to its ability to last longer under physical, mechanical
J Forensic Odontostomatol stress.3 Identification is attained with 100% accuracy if a
2024. Apr;(42): 1-30:37 complete skeleton is available, although this is rarely the case.
ISSN :2219-6749 In cases of mass disasters, with complete or partial loss of
DOI doi.org/10.5281zenodo.11088513 human structure or burns, it is impossible to find an intact
human skeleton.4 For age and sex determination after pelvis,
30
skull is the most dimorphic structure. The forensic odontology. Studies have shown that AI
Mandible being the strongest bone of the skull, is has the accuracy and precision equivalent to
the bone of choice for a ge and sex trained examiner in age and sex determination of
determination. 5,6 Radiographs are simple, the individuals.12,13 Hence this systematic review
accessible, and cost effective and demonstrate is aimed at reporting on Artificial intelligence in
high accuracy for age and sex determination. age and sex determination in forensic odontology.
They can be easily advised in both living and
dead.7 MATERIALS AND METHODS
Artificial intelligence (AI) can be described as an
intelligent computer system with the unique Search strategy and Review !amework
ability to mimic human potential in decision To conduct this systematic review, articles that
making, problem solving, understanding language were based on the application of artificial
and learning. AI models developed using machine intelligence in the mandible for age and sex
learning, can predict events with given sets of determination were selected. Preferred reporting
observations in the form of images or data. AI items for systematic reviews and meta-analyses
not only overcomes the subjectivity of any PRISMA guideline for diagnostic test accuracy
diagnosis or individual examination but also were used Fig.1.14 The data search was conducted
reduces the overall cost. 8 Deep learning is through various search engines, namely Pubmed,
another type of neural networks where the model Cochrane, Google Scholar, and Scopus. The
on its own learns about the data and how to search duration spans 10 years, from January 2013
process the given data. Deep learning neural to March 2023. The following key words were
network have neurons in hidden layers in used in various combinations to improve the
thousands and millions.9,10 AI and its advent in search strategy like Artificial Intelligence (AI)
the field of medical science have made the Convolutional Neural Networks (CNN),
possibility of sound diagnosis and prompt Artificial Neural Networks (ANN), Mandible,
prediction with decision making achievable.11 The Fo r e n s i c O d o n to l o g y, O P G, a n d L a te r a l
major feature of decision making with already Cephalogram. The PICO guideline was used for
known observation is the main feature of AI searching for the right kind of study to be
which will be of great significance in the field of included in the systematic review (Table 1).
Table 1. Table showing PICO (P = Population, I = Intervention, C = Comparison, O = Outcome) and

Research Question.
RESEARCH QUESTION
How will Artificial intelligence tool be used in mandibular age and sex determination in
forensic cases?
Population Patients maxillofacial radiographs (Orthopantomogram, lateral Cephalogram,

CBCT images)
Intervention AI algorithms for age estimation and sex determination
Comparison Comparison across different AI models, reference standards, expert opinions
Outcome Measurable or predictive outcomes such as accuracy, Correlation Coefficient,

sensitivity, specificity, ROC = receiver operating characteristic curve, AUC =
Area Under the Curve.
31
Figure 1. PRISMA Flow chart showing the selection of articles for the systematic review.
Total records obtained

(n = 595)
Records after removal of duplicate articles (n=278)

Screening
Articles assessed for Records excluded on applying

eligibility inclusion exclusion criteria (n
(n =278) =270)
Full text articles included in Full text article excluded

the qualitative synthesis after qualitative synthesis
(n =8) (n=1)
Studies included in
Included
systematic review
(n =7)
Methodology There should be a measurable outcome to assess

For the preliminary search, 585 articles were the role of AI in the study.
identified using database searches, and through
other sources, an additional 11 articles were Exclusion criteria
identified. The net article available after the first Articles on tools apart from AI technology.
stage was 595. The retrieved articles were full Articles with only abstracts and not full text.
texts and abstracts. In the first stage articles were Articles in languages other than English.
chosen, based on their titles and the abstracts
that addressed our research question. In the Data management
second stage, 278 articles were eligible for review Post application of the above eligibility criteria,
after the removal of 278 articles because of only 8 articles were obtained. Qualitative
duplication. The following inclusion and assessment was done for each article based on
exclusion criteria were considered for the QUADAS 2 (Quality Assessment and Diagnostic
eligibility of the studies: Accuracy Tool) by the author.15 One article was
found to be a misfit amongst the 8 articles (Fig 1).
Inclusion criteria 7 articles were available for final analysis, and
The article must be about the application of AI these were read in full length so that AI
to the mandibular bone. There should be specific application in the mandibular age and sex
mention of the AI based model used in the study. determination could be analysed. Each article was
There should be a specified data set for each discussed among 4 domain namely patient
model to be applied for age or sex determination. selection, index test, reference standard, and flow
32
and timing. For each domain risk of bias was low risk. Patient selection, index test, reference
assessed in terms of high risk, unclear risk and standard also assessed the applicability.
RESULTS
While conducting the systematic review, 7 Risk of bias
articles were analysed for quantitative data. All Risk of bias assessment was conducted under 4
studies that were finally analysed were conducted domains, in patient selection, one study showed
in the past five years, although inclusion criteria high risk17 and one showed unclear risk.20 Overall,
stated a duration of 10 years. All the studies were 72% of studies showed low risk in patient
selection. All studies focused on age and sex
based on the AI tool and its application in
determination through the AI models therefore,
mandibular age and sex determination. The there was an overall low risk for reference
models used in the study were ANN16, ML17, standards and index test. For the flow and timing
KCNN-KNN18, BCNN19, and CNN.20-22 These of the study, one study showed unclear risk17. This
neural network models were used to evaluate risk of bias in assessment and applicability
p a n o r a m i c r a d i o g r a p h s 1 6 - 2 0, 2 2 a n d l a te r a l concern is shown in Fig. 2. The applicability
cephalograms.21 showed similar results.
Figure 2. Figure showing the number of studies and their proportion under high risk, unclear and low
risk categories for both Risk of Bias and Applicability Concerns.
DISCUSSION
Age and sex determination play a crucial role in determination in radiographs fails to provide the
the process of identification and in establishing proper identification. The ability of artificial
an anthropological profile.23 The resilience of jaw intelligence to minimise error and precision in
bones provides immense information for f e a t u r e a s s e s s m e n t , f o l l o we d b y p r o m p t
establishing the identity of any unknown body. prediction, makes it a tool of choice.15
Identification through bony and dental remains is In the present systematic review, an effort has
critical in cases of chemical burns, burns, and been made to analyse various AI tools for prompt
complete or partial tissue loss.24 The traditional identification involving age and sex in
morphometric measurement used for age and sex maxillofacial radiographs ( Table 2).
33
Table 2. Details of studies in the systematic review involving AI based models in age and sex
determination using maxillofacial radiographs.
Sl. No. 1. 2. 3. 4. 5. 6. 7.
Authors Banjsak et al 22 Baydogan et al 20 Patil et al 16 Khazaei et al 21 Lee et al 17 Back et al 19 Sharifonnasabl

et al 18
Year of 2020 2022 2020 2022 2022 2020 2022
Publication
Algorithm CNN CNN ANN CNN ML BCNN HCNN-KNN
Age detection Artificial neural Hybrid HCNN-

Convolutional CNN model Machine learning
for Deep network for Age estimation KNN Model for
neural network for sex algorithm for age
learning from gender using Bayesian Age estimation
Objective to estimate age determination group prediction
Dental determination convolutional accuracy in
of archaeological using lateral using panaromic
panaromic using mandibular neural network orthopantomogr
skull remains cephalogram metric parameters
radiographs parameters aphy
Training data 3228 627 509 1180 471 2400 1537

set
Validation 403 Not clear 63 296 Not clear Not clear 384
data set
Testing data 89 Not clear 63 296 Not clear Not clear none
set
Comparison 3 models none 3 models 3 models none none none
Mandibular Mandible and

Mandible and Mandibular Upper and Upper and
Study factor architecture Mandible dental metric
teeth parameters lower jaws lower jaws
and teeth parameters
Orthopantomo Orthopantomo Orthopantomo Lateral Orthopantomo Orthopantomo orthopantomo

Modality
grams grams grams cephalogram grams grams grams
Discriminant
DenseNet Concordance
Model1:53% analysis:69.1%
121:90% correlation
Result Model2:42% 84% Logistic 87% 99.80%
ResNet:62% coefficient
Model3:73% regression:69.9%
VGG 75% ccc=0.91
ANN:75%
Effective(+)
Non- Effective(+) Effective(+) Effective(+) Effective(+) Effective(+) Effective(+) Effective(+)
effective(-)
Neutral(-)
The model
correctly
CNN based
Outcome of In machine classified age with
The study DenseNet121 Bayesian CNN
the study learning 99.8% accuracy.
demonstrated The study has high quantifies the
depicts higher models age Evaluating the
noteworthy depicted high predictive predictive
accuracy for group proposed model
Outcome accuracy in accuracy of accuracy in sex uncertainty
gender prediction on a new dataset
placing images dental age determination which is
prediction accuracy was with different
in correct age estimation using lateral important in
using ANN more than races also proved
group cephalometric legal context
model acceptable the superior
images
performance of
the model
The results also

needs to be Additional
In future studies, Further studies Results are
extrapolated classification with
Further different feature with larger encouraging
using larger larger sample size
development extraction and sample size are although the
population is needed to
Recommend of neural net increasing desirable to more accuracy is not
considering other determine none
ation and training number of data accurately yet at the level
confounding whether the
would offer to be included for determine sex that warrants
factors like origin, prediction
better results state- of- the art correctly on routine
race,age, performance will
classifier larger scale application
masticatory increase
muscles
34
In a study conducted by Back et al16 for age determination from lateral cephalometric
estimation( 15-25 years age), Bayesian radiographs using CNN. The three models,
Convolutional neural network was used to attain DenseNet, ResNet, and VGG, were evaluated for
t h e co n co r d a n ce co r r e l a t i o n co e f f i c i e n t their ability to correctly determine the sex of the
ccc=0.910. Age estimation in orthopantomograms lateral cephalogram. DenseNet 121 architecture
was formulated using a regression task and CNN depicted the highest accuracy of 90%. The study
was designed in which Inception V3 architecture also showed the importance of transfer learning
was used. There was a considerable agreement to attain high accuracy in sex determination. The
between true age and predicted age; however, the study gives desirable results and can be
mean absolute error was almost 2 years, which is implemented with a larger sample size.
highly unacceptable for legal purposes. The In a study conducted by Lee et al 17 for age
results were encouraging for the application of AI determination (11 to 69 years), machine learning
in the form of BCNN, but not to the level that was used. Five machine learning models were
confirms routine application. formulated. Each model demonstrated high
Banjsak et al22 used CNN for archaeological age accuracy for age group prediction. Overall
estimation (19 to 85 years of age) through transfer accuracy of 87% was achieved, although a larger
learning. A VGG architecture pre trained on the sample size is needed to determine the true
ImageNet feature extractor was used. Three prediction accuracy.
models were used for age prediction by just The above mentioned studies have used various
changing the hyperparameters. Model 1 showed models in order to provide a ge or sex
53% accuracy, Model 2 showed 42% accuracy, and determination with accuracy higher than the
Model 3 showed 73% accuracy. This study already existing methodology. All these studies
demonstrated a noteworthy accuracy that should have certainly high accuracy, but translating the
be considered for further analysis using an results into routine applications will require a
orthopantomogram. deeper and more detailed study. Although the
In another study conducted by Baydogan et al20 present study is based on AI application on
for age determination (2 to 13 and 13 to 31 years) mandibular bone, expanding the application to
using orthopantomograms, CNN was used with dental age and sex determination can be of great
the Alexnet architecture for feature extraction, help if interpreted together. Teeth based age and
followed by four classification algorithms. The K sex determination using a hybrid transformer
nearest neighbour algorithm showed the highest model can add significant value to overall
accuracy (84% accuracy). This study depicts high results.26 The ability of AI to provide objective
accuracy with the intervention of AI in age and accurate decision making can be used
determination. explicitly in other domains like bitemark, facial
In a study conducted by Patil et al16 for sex identification, and dental comparison.27 AI is a
determination using ANN, orthopantomogram game changer when the correct data is available.
was used. Logistic regression and discriminant Data in the form of maxillofacial radiographs is a
analysis were used for sex determination, along prerequisite for such studies. In several studies
with an artificial neural network. ANN had an deep convolutional network have proven to be of
accuracy of 75%, much higher than the other two high accuracy and practical significance with
models. This study stated that ANN could be a accuracy as high as 94.7%.28 The advent of AI in
promising prediction tool for sex determination the field of forensic medicine and odontology will
in forensics. certainly aid in quality decision making. This
Sharifonnasabi et al 18 conducted a study on overarching domain requires skilled manpower
orthopantomograms for age determination (15 to and high quality data to provide highly objective
23 years) using the HCNN-KNN model. This and accurate results. Artificial intelligence
model had a very high accuracy of 99.8% for age provides an opportunity to revolutionise the
determination. The precision was as high as +_ 6 disciple, but comes with certain challenges. The
months. Principal component analysis was used issue of data privacy being one of the major
to compensate for overfitting. This study proves hindrances, followed by the availability of a huge
that a substantial model design could facilitate amount of good quality data is a challenge. This
prompt age classification. in turn could perpetuate bias and amplify
Khazaei et al 21 conducted a study on sex incorrect discrimination ability.29 In order to
35
Al
go
rit
h
m
mitigate the possible disadvantages offered by AI, taking over various sectors. Forensic medicine,
extensive research, an expanding knowledge base, forensic odontology, and anthropology are no
and exploring the true potential of deep learning exceptions. The advent of AI, with the help of
and machine learning algorithms are to be done. good-quality data and vast information, can be an
Good quality data is the cornerstone of all AI excellent tool. For age and sex determination,
applications. This data can be helpful for quality and quantity of data are imperative to
obtaining prompt results in a ge and sex facilitate the development of proper models or
determination in real-life scenarios, and could be tools that the authorities can rely upon. The
of great help in cases of mass victim development of such a model can be utilised by
identification, decomposed bodies, and criminal medical, legal, police, and forensic authorities.
cases where it is important to establish age and Therefore, future studies are recommended by
sex along with other parameters. utilising the application of AI in age and sex
determination in a wider maxillofacial
CONCLUSION radiographic data set.
We are at the doorstep of another technological
advancement, which will be directed by artificial Statements and Declarations:
intelligence. AI and its ability to mimic human No competing interest have appeared to affect
capabilities and decision-making are already the work.
REFERENCES
1. Luntz LL. History of forensic dentistry. Dent Clin North 11. Park WJ, Park JB. History and application of artificial
Am. 1977 Jan;21(1):7–17. neural networks in dentistr y. Eur J Dent. 2018
2. Johnson A. A Morphological Study of Tongue and its Role in Oct;12(4):594–601.
Forensics Odontology. J Forensic Sci Crim Investig [Internet]. 12. An automated technique to stage lower third molar
2018 Feb 19 [cited 2023 Apr 21];7(5). Available from: https:// development on panoramic radiographs for a ge
juniperpublishers.com/jfsci/JFSCI.MS.ID.555723.php estimation: a pilot study. - PMC [Internet]. [cited 2023
3. Berketa JW, James H, Lake AW. Forensic odontology Apr 22]. Available from: https://www.ncbi.nlm.nih.gov/
involvement in disaster victim identification. Forensic Sci pmc/articles/PMC6100230/
Med Pathol. 2012 Jun;8(2):148–56. 13. Roh BY, Lee WJ, Seo JU, Lee UY, Lee SS. Analysis of
4. Scheuer: Application of osteology to forensic medicine - forensic odontological examinations at the National
G o o gle Sc h ol ar. [ c ite d 2 0 2 3 A p r2 1 ] .h ttps:// Forensic Service of Korea from 2011 to 2015. Leg Med
s c h o l a r. g o o g l e . co m / s c h o l a r _ l o o k u p ? Tokyo Jpn. 2018 May;32:37–42.
journal=Clin+Anat&title=Application+of+osteology+to+foren 14. McGrath TA, Alabousi M, Skidmore B, Korevaar DA,
sic+medicine&author=L+Scheuer&volume=15&publication_ Bossuyt PMM, Moher D, et al. Recommendations for
year=2002&pages=297-312&pmid=12112359& reporting of systematic reviews and meta-analyses of
5. Saini V, Srivastava R, Rai RK, Shamal SN, Singh TB, diagnostic test accuracy: a systematic review. Syst Rev.
Tripathi SK. Mandibular ramus: an indicator for sex in 2017 Oct 10;6(1):194.
fragmentary mandible. J Forensic Sci. 2011 Jan;56 Suppl 15. Whiting PF, Rutjes AWS, Westwood ME, Mallett S,
1:S13-16. Deeks JJ, Reitsma JB, et al. QUADAS-2: a revised tool for
6. More CB, Vijayvargiya R, Saha N. Morphometric analysis the quality assessment of diagnostic accuracy studies.
of mandibular ramus for sex determination on digital Ann Intern Med. 2011 Oct 18;155(8):529–36.
orthopantomogram. J Forensic Dent Sci. 2017;9(1):1–5. 16. Patil V, Vineetha R, Vatsa S, Shetty DK, Raju A, Naik N,
7. Johnson A, Singh S, Thomas A, Chauhan N. Geometric et al. Artificial neural network for gender determination
morphometric analysis for sex determination using lateral using mandibular morphometric parameters: A
cephalograms in Indian population: A preliminary study. J comparative retrospective study. Jin Z, editor. Cogent
Oral Maxillofac Pathol JOMFP. 2021;25(2):364–7. Eng. 2020 Jan 1;7(1):1723783.
8. Schwendicke F, Samek W, Krois J. Artificial Intelligence 17. Lee YH, Won JH, Auh QS, Noh YK. Age group
in Dentistry: Chances and Challenges. J Dent Res. 2020 prediction with panoramic radiomorphometric
Jul;99(7):769–74. parameters using machine learning algorithms. Sci Rep.
9. Javed S, Zakirulla M, Baig RU, Asif SM, Meer AB. 2022 Jul 9;12:11703.
Development of artificial neural network model for 18. Sharifonnasabi F, Jhanjhi NZ, John J, Obeidy P, Band SS,
prediction of post-streptococcus mutans in dental caries. Alinejad-Rokny H, et al. Hybrid HCNN-KNN Model
C o m p u t Me t h o d s Pr o g r a m s B i o m e d . 2 0 2 0 Enhances Age Estimation Ac c u r a c y in
Apr;186:105198. O r t h o p a n t o m o g r a p h y. Fr o n t P u b l i c He a l t h .
10. K h a n a g a r S B , A l - E h a i d e b A , Ma g a n u r P C , 2022;10:879418.
Vishwanathaiah S, Patil S, Baeshen HA, et al. 19. Back W de, Seurig S, Wagner S, Marré B, Roeder I,
Developments, application, and performance of artificial Scherf N. Forensic age estimation with Bayesian
intelligence in dentistry - A systematic review. J Dent Sci. convolutional neural networks based on panoramic
2021 Jan;16(1):508–22. dental X-ray imaging. 2022 Jul 13 [cited 2023 Apr 23];
36
Av a i l a b l e f r o m : h t t p s : / / o p e n r e v i e w. n e t / f o r u m ? 25. Sikka A, Jain A. Sex Determination of Mandible: A

id=SkesoBY49E Morphological and Morphometric Analysis. Int J
20.Baydoğan MP, Baybars SC, Tuncer SA. Age Detection by Contemp Med Res, 2016;3: 1869-72.
Deep Learning from Dental Panoramic Radiographs. 26. Fan F, Ke W, Dai X, Shi L, Liu Y, Lin Y, Cheng Z,
Artif Intell Theory Appl. 2022 Oct 1;2(2):51–8. Zhang Y, Chen H, Deng Z. Semi-supervised automatic
21. Khazaei M, Mollabashi V, Khotanlou H, Farhadian M. dental age and sex estimation using a hybrid
S e x d e te r m i n a t i o n f r o m l a te r a l ce p h a l o m e t r i c t r a n s f o r m e r m o d e l . In t J L e g a l M e d . 2 0 2 3
May;137(3):721-731. doi: 10.1007/s00414-023-02956-9.
r a d i o g r a p h s u s i n g a n a u to m a te d d e e p l e a r n i n g
Epub 2023 Jan 31.
convolutional neural network. Imaging Sci Dent. 2022 27. Vodanović M, Subašić M, Milošević DP, Galić I, Brkić
Sep;52(3):239–44. H. Artificial intelligence in forensic medicine and
22. Banjšak L, Milošević D, Subašić M. Implementation of forensic dentistry. J Forensic Odontostomatol. 2023
artificial intelligence in chronological age estimation Aug 27;41(2):30-41.
f r o m o r t h o p a n t o m o g r a p h i c X- r a y i m a g e s o f 28. Ilić I, Vodanović M, Subašić m. Gender Estimation
archaeological skull remains. Bull Int Assoc Paleodont. from Panoramic Dental X-ray Images using Deep
2020 Dec 22;14(2):122–9. Convolutional Networks. IEEE EUROCON 2019
23. Wang X, Liu Y, Miao X, Chen Y, Cao X, Zhang Y, et al. -18th International Conference on Smart Technologies,
DENSEN: a convolutional neural network for estimating 2019, pp. 1-5, Novi Sad, Serbia, 1-4 July 2019. doi:
10.1109/EUROCON.2019.8861726
chronological ages from panoramic radiographs. BMC
29. Vodanović M, Subašić M, Milošević DP, Galić I, Brkić
Bioinformatics. 2022 Oct 14;23(3):426. H. Artificial intelligence in forensic medicine and
24. Pretty IA. Forensic dentistry: 1. Identification of human forensic dentistry. J Forensic Odontostomatol. 2023
remains. Dent Update. 2007 Dec;34(10):621–2, 624–6, Aug 27;41(2):30-41.
629-630 passim.
37
Are computed tomography images of the mandible

useful in age and sex determination? A forensic
science meta-analysis

ABSTRACT
Luciana Munhoz 1, Shunsuke Objectives. This meta-analysis addresses the use of mandibular
Okada2, Miki Hisatomi2, computed tomography (CT) scans for age and/or sex
Yoshinobu Yanagi3, Emiko Saito determination in forensic science.
Methods. Six databases were searched until June 2023, using
Arita1, Junich Asaumi2
the keyword “mandible” combined with keywords related to
“multislice computed tomography” (MCT) or “cone-beam
computed tomography” (CBCT) and keywords related to
1Department of Stomatology, São “skeletal age determination” or “sex determination analysis.”
Paulo University Dentistry School,
Brazil 2Department of Oral and
Main Results. Among the 23 studies included, 11 used MCT
Maxillofacial Radiology, Okayama and 12 used CBCT to perform forensic assessments. Age
University Hospital, Japan determination was the aim of a single study, sex and age
3Department of Dental Informatics,
determinations were the objective of five studies, and the other
Okayama University, Japan studies investigated the determination of sex only. Meta-
analysis could be performed only for sex determination.
Conclusions. Mandible measurements are useful in sex
determination, as the bicondylar and bigonial breadth are
Corresponding author:
[email protected]
larger in males than in females. For the mandible angle, the
meta-analysis results confirm sex dimorphism in CBCT scans
but not in MCT scans. For age estimation, further studies are
needed to prove that the mandible hole is a reliable parameter
The authors declare that they for age estimation. PROSPERO registration number:
CRD42021260967.
have no conflict of interest.
INTRODUCTION
In forensic science, sex and age determination are fundamental
aspects of personal identification. As the mandible is the
KEYWORDS largest and longest-lasting facial bone, it has a critical role in
human identification, particularly in the absence of a complete
Mandible, skull or pelvis.1 Determining individual features using the
Jaw, mandible comprises the use of measurements and macroscopic
Computed tomography, morphological form assessments.1
The integration of forensic science and medical imaging
Forensic Anthropology, technology has contributed to considerable advances in
Human Identification forensic science; it is now possible to evaluate body structures
in highly decomposed or contaminated bodies or in cultures
with low autopsy acceptance.2 Although plain radiographs have
their value in forensic practice, technologies with 3D outputs,
J Forensic Odontostomatol such as computed tomography (CT),2 that provide accurate
2024. Apr;(42): 1-38:57 and reliable imaging of maxillofacial structures have been
ISSN :2219-6749
largely studied and applied in forensic science.
Multislice computed tomography (MCT) and cone-beam
computed tomography (CBCT) are CT imaging techniques
38
that use the same imaging reconstruction Medicine), Embase® (Excerpta Medica Database),
principle, although they differ in radiation dose Scopus® (Elsevier), Cochrane Central Register of
and in spatial, contrast, and temporal resolution.3 Controlled Trials, Web of Science® (Institute of
CBCT is typically applied in cases of dental and Scientific Information – Clariative Analytics), and
jaw disorders; however, it is not applied in cases Google Scholar® (Google). The aforementioned
of neoplastic lesions where the administration of databases were searched without language and time
contrast agents and evaluation of soft tissues is restrictions (until June, 2023). The Boolean
required.3 operators “AND” or “OR” were used to combine
For postmortem human identification, several and optimize the searches.
combinations of distinct mandible landmarks and Itemized search strategies were established for each
linear or angular measurements have been database based on keywords determined by
proposed and extensively studied. However, this “Medical subjects headings” (MESH): “Mandible”
process is complicated by the fact that it is combined with keywords defining MCT as:
difficult to access and assess complex anatomical “Mu l ti sl i ce C om pu ted Tom ogra phy O R
structures or sites in mandibles covered by soft Multidetector Computed Tomography OR
tissue. Thus, MCT and CBCT have been utilized Multidetector-Row Computed Tomography OR
as imaging tools in forensic investigations. Multisection Computed Tomography” or keywords
Frequently, the objective of such investigations defining CBCT as: “ Cone-Beam Computed
was to verify the correlation between the linear Tomography OR CT Scan, Cone-Beam OR Cone-
or angular measurements of mandibular Beam CT OR Cone-Beam Computer-Assisted
anatomical sites and age or sexual dimorphism in Tomography OR Cone-Beam Computerized
different populations. Some of the studies also Tomography OR Volume CT OR Volume
investigated the influence of age and sex on Computed Tomography OR Volumetric CT OR
mandible shape. However, comparisons that Volumetric Computed TomographyCT Scan, Cone-
include distinct populations have not yet been Beam OR Cone-Beam CT OR Cone-Beam
performed. Computer-Assisted Tomography OR Cone-Beam
Hence, the objectives of the present systematic Computerized Tomography OR Volume CT OR
review were to determine 1) the mandible sites Volume Computed Tomography OR Volumetric
that have been studied for skeletal age and sex CT OR Volumetric Computed Tomography”.
determination, 2) the main results and For searches regarding age determination, the
conclusions of the reviewed studies, and 3) f ol l o wi ng keywo rd s wer e ad d ed to t he
whether mandible images are useful in the aforementioned combination: “Age Determination
determination of a ge and sex for human by Skeleton OR Bone Age Measurement OR
identification. It should be noted that only Skeletal Age Measurement OR Skeletal Maturation
studies that used MCT or CBCT for mandible- Index”. For searches regarding sex determination,
based assessment of age and/or sex were included the following keywords were added: “Sex
in the review. Determination Analysis OR Sex Determination
Technics OR Sex Determination Techniques”.
MATERIALS AND METHODS Manual searches were also performed.
Protocol and registration Eligibility criteria: Types of studies and Participant groups
This systematic review and metanalysis is registered Published research articles or technical notes were
at the National Institute for Health Research, considered for inclusion. Abstracts, oral
International Prospective Register of Systematic presentations, case reports and literature reviews
Reviews (PROSPERO). The registration number is were excluded.
CRD42021260967. The Preferred Reporting Items Investigations with mandible measurements or
for Systematic Reviews and Meta-Analyses morphologic classifications for age and sex
(PRISMA) checklist was followed.4 determinations in forensic science, using MCT or
CBCT were considered for inclusion. Investigations
Data selection about development of software or equations for age
The selection of studies potentially eligible for or sex determination without measurements or
inclusion in this was performed using the following morphologic classifications were excluded.
databases: PubMed Central® (United States The articles considering the following assessments
National Institutes of Health’s National Library of were excluded: dental status, canal mandibular,
39
mandibular foramen, mental foramen, alveolar bone selected study. The screening and selection of
evaluations. potentially included studies will be performed using
MCT or CBCT scans performed in human beings Rayyan QRI (https://www.rayyan.ai/).5
were included. Studies performed in dried The search results were summarized in one flow
mandibles or that did not used MCT or CBCT chart, according to PRISMA statement6 (Figure 1 –
were excluded. Investigations that not included data selection) and tables (Tables 1 to Table 4).
mandible bone in the assessment, were not
considered for inclusion. Data analysis – risk of bias
The quality of each original research were be
Data extraction assessed using the Cochrane risk of bias tool for
Data extraction was executed by two independent non-randomized studies, 7 and demonstrated in a
reviewers, who initially screened the titles and figure (Figure 2) using Robvis tool 8 (https://
abstracts, and then evaluated the full text of each mcguinlu.shinyapps.io/robvis/).
Figure 1. PRISMA flow diagram illustrating the literature search
40
Table 1. Year of publication, main subject (skeletal age or sex determination), type of computed
tomography used (Multislice computed tomography or cone beam computed tomography), sample
features and ethnicity of the group studies
Origin of the
Determination Bones Type
Author Year Country Sample features population
sex or age? included of CT
studied
Libyan
200 CT scans: minimum age
Libya and Population in
Atef et al.9 2020 Sex Mandible MCT 18 years; maximum age 60
Egypt Tripoli
years; 87 females, 113 males
Imaizumi 100 CT scans: minimum age

Mandible
et al.10 2020 Japan Sex MCT 23 years; maximum age 65 Japanese
and Skull
years; 114 females, 99 males
120 CT scans, minimum age

23; maximum age 84 years, 57
Gillet et Mandible females and 63 males, divided
2020 France Sex MCT French
al.11 and Skull in three groups: whole
sample, over 40 years, under
40 years.
Egypt, 213 CT scans: minimum age 7

Motawei et Saudi Mandible years; maximum age 63 years;
2020 Age and Sex CBCT Egyptians
al.12 Arabia, 114 females, 99 males
Taiwan
Okkesim 70 CT scans: minimum age 18 Central

and 2020 Turkey Sex Mandible CBCT years; maximum age 29 years; Anatolian
Erhamza13 35 females, 35 males Turkish
Australia 654 CT scans: minimum age Australian?

Fan et al.14 2019 and Sex Mandible CBCT 8.5 years; maximum age 19.5 (Email sent to
Belgium years; 386 females, 268 males the author)

Albalawi et Saudi
2019 Sex Mandible CBCT 18 years; maximum age 60 Saudi Arabia
al.15 Arabia
years; 104 females, 96 males.
300 CT scans: minimum age White (country

Bulut et Germany Sex (according to
2019 Mandible MCT 20 years; maximum age 80 of origin not
al.16 and Turkey age ranges)
years; 150 females, 150 males. specified)
121 CT scans: minimum age Turkish (from

Tassoker et
2019 Turkey Sex and age Mandible CBCT 10 years; maximum age 69 Middle
al.17
years; 71 females, 50 males Anatolia)
79 CT scans: minimum age 18

Alias et al.18 2018 Malaysia Sex Mandible MCT years; maximum age 74 years; Malaysian
31 females, 48 males.
433 CT scans: minimum age 8

Barak et
2018 Turkey Age and Sex Mandible CBCT years; maximum age 31 years; Turkish
al.19
260 females, 173 males.
60 CT scans: 30 females, 30
Barbieri et males. The scans were
2018 Brazil Age and Sex Mandible CBCT divided in groups according Brazilian
al.20
to age (5 examinations for
each decade of life).
Sex (age as Mandible 420 CT scans: minimum age Han adults in

Zheng et
2018 China secondary and CBCT 18 years; maximum age 70 Northeast
al.21
objective) maxilla years; 210 females, 210 males. China
41
Deng et 219 CT scans: minimum age 7 Central

Mandible
2017 China Sex CBCT years; maximum age 20 years; Chinese
al.22
108 females, 111 males
438 CT scans: 214 females,

Tunis et 224 males; male mean age
2017 Israel Sex Mandible MCT Israeli
al.23 53.3 ± 19.9; female mean age
56.2 ± 20.6 years.

Inci et al.24 2016 Turkey Sex Mandible MCT 18 years; maximum age 60 Turkish

Gamba et
2016 Brazil Sex Mandible CBCT 18 years; maximum age 60 Brazilian
al.25
Dong et 203 CT scans: minimum age

2015 China Sex Mandible CBCT 20 years; maximum age 65 Chinese Han
al.26
232 CT scans from cadavers:

Kano et minimum age 16 years;
2015 Japan Sex Mandible MCT Japanese
al.27 maximum age 100 years; 106
females, 116 males
İlgüy et al.28 161 CT scans: minimum age

European
2014 Turkey Sex Mandible CBCT 18 years; maximum age 85
descendants
China and 240 CT scans: minimum age

Lin et al.29 2014 Republic Sex Mandible MCT 21 years; maximum age 70 Korean
of Korea years; 120 females, 120 males.
167 CT scans of fetuses (74

Femur females and 93 males), aged
Minier et
2014 France Age and MCT from 20 to 40 weeks. The Not specified
al.30
mandible mandible was missing in 16
fetuses
Egypt and 500 CT scans: minimum age

Karoshah
2010 Saudi Sex Mandible MCT 6 years; maximum age 60 Egyptian
et al.31
Arabia years; 250 females, 250 males.
Abbreviations: MCT: multislice computed tomography; CT: computed tomography; CBCT: cone beam computed tomography
Statistical Assessment standardized mean difference as the outcome

Variables were assessed only if data provided was measure. A random-effects model was fitted to
available as “mean values” and “standard the data. The amount of heterogeneity was
deviations” from 3 or more investigations using estimated using the maximum-likelihood
exactly the same measurement and the same type estimator. In addition to the estimate of tau², the
of CT. Data was not considered for the meta- I² statistic are reported.
analysis if incomplete or missing the population Meta-analysis assessments were performed using
origin. The analysis was carried out using the Jamovi version 1.6 (The Jamovi Project).
42
Table 2. Summary of the methodology, results and conclusions of the studies included
Methodology
Author Results Conclusions
applied
Mandible can be used to

Quantitative There are differences between male and female in
Atef et differentiate sex as evidenced by
assessment (measures all mandibular parameters except minimal ramus
al.9 that female were higher than
in mandible) breadth and gonial angle.
male except gonial angle.
Morphologic studies The validation results on actual casework skulls The sex estimation method developed
Imaizumi
and machine learning were less acceptable than expected; a larger sample enables to perform objective
et al.10
for shapes creation. is needed to achieve better results. identification of skeletal remains.
Metric and geometric Although the mandible does not

Cranium was the most dimorphic structure,
morphometric appear to be the most dimorphic
Gillet et regardless of which analysis method and individual’s
methods to evaluate structure of the cephalic extremity
al.11 age. The assessment of mandible did not increase
size and shape-related sample, it remains a useful tool in
sex estimation accuracy for the whole skull.
sexual dimorphism. the absence of an intact skull.
The length of the ramus

Motawei There are no sexual dimorphism of The mandibular ramus length is
of the mandible was
et al.12 the mandible ramus length until the age of 17 valuable in age estimation and
measured in lateral CT
years. less valuable in sex determination.
scans.
Linear parameters The development of standards to

Okkesim It was found that all variable of mandibular ramus
were measured using each population for accurate
and on CBCT models showed a statistically significant
the mouse-driven gender identification from
Erhamza13 difference between males and females.
method. skeletal remains is needed.
Growth trajectories of
Growth direction in both males
the mandible in males Mandibular sexual dimorphism already exists at
and females is similar but is faster,
Fan et and females were 9 years of age, but this is mostly in size, but not in
peaks later and occurs over a
al.14 modelled using a non- shape. Significant dimorphism was evident by
longer period in males than in
linear kernel 11 years and increased through adolescence.
females.
regression framework.
The angle formed by the

Mandibular angular intersection of lines from the left
Albalawi Measurements presented differences between
measurements using and right gonion to mention
et al.15 males and females.
3D images. helps in providing
anthropological data.
Measurements and
comparisons of gonial The authors showed that the gonial angle is The results revealed that the
angle, using 3D CT sexually dimorphic in senior adult ages (60 – 80 gonial angle is not a particularly
Bulut et
imaging. Sample years). Females have larger gonial angles in all 3 age good indicator to identify the sex
al.16
divided according to sex groups (no statistic test showed for the from the cranium and should not
and grouped according aforementioned information provided by authors). be used as a sole criterion.
to age ranges.
Authors compared
Panoramic radiography
panoramic radiographs
Tassoker According to CBCT examinations, right and left measurements showed significant
with CBCT using
et al.17 gonial angle are higher in females than males. differences from CBCT in the
linear and angular
mandible.
measurements.
In this study, all parameters were found to be

Morphometric and The mandible could be
greater in male mandibles than in female. By
Alias et morphological distinguished according to the sex
stepwise discriminant function analysis, from the
al.18 parameters analysis in the Malaysian population.
bigonial breath and condylar height were the best
using 3D imaging
parameters selected in the analysis.
43
Condyle cortication
For Males: Chronologic age increased as the
assessment using visual
*Type I mean age observed: 14.14 years stages of the cortication progress
classification: type 1
*Type II mean age observed: 16.11 years from Type I to Type III in male
(no cortication); type
Barak et *Type III mean age observed: 19.39 years and female
2; type 3 (surface with
al.19 For Females: individuals, and all the stages of
similar or higher
*Type I mean age observed: 13.01 years the cortication in the mandibular
density than the
*Type II mean age observed: 15.52 years condyle of male occur compared
surrounding cortical
*Type III mean age observed: 17.95 years to female.
areas).
3D models generated No differences were found between mandibular The structure evaluated cannot
Barbieri
from 3D angular incision measurements in both sexes, or age ranges contribute to forensic
et al.20
measuring tools. groups. anthropology evaluations.
Sex determination of
The maxillofacial
Zheng et Evaluation of variables using CBCT reconstruction maxillofacial region using CBCT
bones were measured
al.21 technology provided a new theoretical basis and has a high accuracy rate and is
in the median sagittal
practical means for sex determination. also applicable to different
position.
countries.
Virtual measurements obtained

Four linear variables It was demonstrated that the breadth size of the
Deng et from 3D images by CBCT may
were selected in the mandible is a useful in sex determination in the
al.22 serve as a substitute for direct
mandible 3D images. studied population.
anatomic measurements.
Linear measurements
Tunis et from 3D Except for mandibular angle, males have a greater The method applied is not age
al.23 reconstructions of the mean value than females. dependent.
mandible.
There was no statistical difference in the

Morphometric measures of the
mandibular flexure angle between males and
upper part of the ramus, can
Inci et Linear distances and females. Mandibular angle values were higher in
provide valuable data to
al.24 angle measurements. females. Comparing the accuracy rates of sex
determine sex in a Turkish
determination, the upper ramus vertical height
population.
showed the highest dimorphism.
Measurements Bicondylar breadth, Ramus

Gamba et
obtained from 3D Authors validated a formula that provided an length, Bicondylar breadth, and
al.25
sagittal views and axial imaging metric that can assist the dental examiner. Gonial angle showed better
views reliability for sex estimation.
Linear or angular
All of the measurements studied were sexually Mandible expresses sexual
Dong et measurements using
dimorphic, with the maximum mandibular length dimorphism in the contemporary
al.26 3D images from CBCT
and bi-condylar breadth being the most dimorphic. adult Han Chinese population.
scans.
Quantitative These findings suggest the

measurements and Although these parameters weakly depended on efficacy of CT morphometry of
Kano et
correlation with body the body height, the correlations were insufficient the mandible for sex
al.27
height as a secondary for stature estimation. discrimination with quantitative
data. assessment.
The sagittal diameter of

foramen magnum seems to be
İlgüy et Measurements were The mean values of mandibular measurements
useful according to the
al.28 performed using 3D were greater for males than females except for
discriminant
imaging. gonial angle.
analysis test for sex
determination.
Males are larger than females in all variables, The upper ramus above flexure
Lin et Measurements using except for mandibular flexure angle, mandibular has the larger potentials than the
al.29 mandible 3D models. flexure depth and mandibular flexure lower border mandibular ramus flexure itself to
and mandible angle. discriminate sexes.
44
Linear and angular Femoral length and mandible measurements presented

Minier et Mandible is a reliable indicator
measurements of the correlation with age; however femoral length
al.30 for estimating fetal age at death.
mandible and femur. correlation was stronger than mandible correlation.
Significant differences and included: bicondylar The overall predictive accuracy of

Karoshah Measurements in 3D
breadth, gonial angle and minimum ramus breadth. the prediction model constructed
et al.31 models
was 83.9%
Table 3. Measurements performed, and their main values, angles or morphometric parameters reported
(values or classification, if applicable) and statistical analysis results from publications that used MCT in
the assessments
Values reported
Mean values Angles measured for the mandible
Measurements provided in mandible or angles measured Statistical analysis
Authors (mm) or morphologic
in mandible morphometric results*
parameters features
Male Female Male Female
*Ramus lengtha
58.2 49.1
*Minimal ramus Except for minimal
23.3 23.4
breadth ramus breadth,
*Coronoid height variables showed
53.7 47.1
Atef et al.9 *Gonion- *Gonial angle 121.51 125.0 statistically significant
60.1 50.3
gnathion length differences. Mandible
*Bicondylar angle was higher in
96.3 87.9
breadth females than males.
77.7 76.7
*Bigonial length
Little/no
Projected
*Mental eminence projecti
Virtual shapes created
Imaizumi et *Gonion on
Everted showed clear sexual
al.10 Little/no
dimorphism.
*Chin eversion
Squared
Oval
Male
Female
*Mandibular 32.25 Male presented higher
29.43
symphyses height mean values than
Gillet et al.11
*Ramus heighta 58.96 *Gonial angle females, except for
54.34 Not reported
*Bigonial breadth 94.94 gonial angle.
87.52
*Bicondylar 104.16
96.77
breadth
No statistically
difference was
observed among the
*Gonial angle age groups in both
(20 – 39 years) 123.73 124.03 sexes; except for the
Bulut et al.16
(40 – 59 years) 123.38 124.16 age range 60 – 80
(60 – 80 years) 122.99 124.69 years, with higher
values for the gonial
angle for females than
males (p = 0.04).
45
*Maximum
breadth of ramus
*Minimum breath
of ramus
*Shapes chin:
*Condylar height
Squared 92% 84%
*Maximum height
Pointed The independent t-
of ramus
Mean values not *Gonial flare test showed significant
Alias et al.18 *Coronoid height
provided Everted 85% 80% difference between
*Mandibular
Inverted males and females.
Body Height
*Muscle markings
*Symphyseal
More prominent
height
Less prominent 90%
*Bicondylar
90%
Breadth
*Bigonial Breadth
*Ramus lengthb
*Ramus width 66.9 58.9
*Body length 31.8 30.2
*Mandibular 79.9 75.0
angle 34.5 31.5
Width Significant differences
*Coronoid width 23.7 22.4 between males and
*Coronoid height 19.4 17.5 females were found for
*Condyle width 20.3 18.4 all mandibular external
Tunis et al.23 *Chin width 28.3 23.2 *Mandibular measurements and for
123.5 125.6
*Bicondylar 122.4 115.7 Angle most of the internal
Breadth measurements. Except
*Bigonial breadth 94.0 87.1 for mandibular angle,
*Chin heighta 21.6 21.0 males have a greater
*Chin thickness 4.0 3.9 mean value than
*Chin area 52.9 50.3 females.
*Symphysis area 322.9 283.5
*Symphysis 15.5 14.4
Thickness
*Symphysis 33.1 30.1
height
*Minimum ramus
breadth
*Maximum ramus
breadth
*Mandibular *Mandibular angle
flexure upper *Upper mandibular Mandibular flexure
border Distance flexure angle angle presented no
*Mandibular (Mandibular flexure statistical difference
flexure lower upper border - between males and
Inci et al.24 border Distance Mean values not posterior plane of Mean values not females. Mandibular
*Mandibular provided mandibular ramus) provided angle values were
flexure depth * Mandibular flexure higher in females
vertical distance angle (Mandibular (P<0.001); all other
*Mandibular flexure upper border values were higher in
ramus flexure - Mandibular flexure males (P<0.001).
vertical height lower border)
*Maximum ramus
vertical height
*Upper ramus
vertical height
46
Sex difference in
virtual measurements
*Bicondylar 61.5 63.3
*Angle formed by were observed in the
breadth 128.4 121.7
bilateral gnathion angle formed by
Kano et al.27 *Bigonial width 102.9 95.8
and condyles bilateral gnathion and
*Gnathion - 125.2 117.4
*Mandibular Angle condyles. No mention
Condylus 126.0 129.8
about mandibular
angle results.
*Minimum
Ramus Breadth Mean measurement
*Maximum 36.46 34.24 values between male
Ramus Breadth and female showed
*Mandibular 46.71 44.08 statistically significant
Flexure Upper differences, with the
Border 26.01 22.92 exception of
*Mandibular *Mandibular Angle mandibular flexure
122.53 124.30
Flexure Lower 17.08 18.01 *Upper Mandibular angle, mandibular
52.52 50.01
Lin et al.29 Border Flexure Angle flexure depth and
*Mandibular 2.22 2.24 *Mandibular Flexure mandibular flexure
165.31 164.62
Flexure Depth Angle lower border. Males
*Mandibular 26.71 24.96 are larger than the
Ramus Flexure females except for
*Maximum 57.62 51.52 mandible angle
Ramus Vertical (p<0.05).
Height 30.92 26.56
*Upper Ramus
Vertical Height
*Coronoid
process-Condylar
process
*Condylar *Mental Tubercule-
process- Coronoid process-
Mandibular angle condylar process Distances Coronoid
*Mandibular *Coronoid process- process to Condylar
angle-Mental condylar process- process and Coronoid
Minier et tubercle Mean values not mandibular angle Mean values not process to Mandibular
al.30 *Condylar provided *Condilar process- provided angle
process-Mental mandibular angle- (R2=0.85);Condylar
tubercle Mental tubercule process to Mental
*Coronoid *Mandibular Angle- tubercle (R2=0.72).
process- Mental Tubercule-
Mandibular angle Coronoid Process
*Coronoid
process-Mental
tubercle
Femal
Male
e Bicondylar breadth
65.1
*Ramus lengthb 64.7 and minimum ramus
28.7
*Minimum ramus 27.96 breadth were
breadth significantly higher in
Karoshah et *Mandibular base Male Female males than in females.
al.31 length (gonion– *Gonial angle 122.8 121.1 Gonial angle in males
76.2
gnathion length) 83.1 was significantly
*Bigonial breadth greater than that in
*Bicondylar females.
104.8
breadth 100.8
108.9
99.6
aDistance between gonion and condilyon
bDistance from the highest point on the condyle to the gonion
*Pertaining to data demonstrated in this table only
47
Table 4. Measurements performed, and their main values, angles or morphometric parameters reported
(values or classification, if applicable) and statistical analysis results from publications that used CBCT in
the assessments
Values reported
Angles for the mandible
Mean values measured in
Measurements angles measured Statistical analysis
Authors provided mandible or
in mandible or morphologic Results*
morphometric features
parameters
Male Female Male Female
No significant
differences between sex
in age range of 7 to 17
*Mandible Ramus
4.79 4.66 years; significant
Motawei et Length
6.04 5.42 difference between
al.12 (07 – 17 years)
males and females for
(17 – 58 years)
the mean length of the
mandible ramus in the
age range of 17–58 years.
*Coronoid height
63.54 57.57
*Condilar height
66.65 59.98
*Mandibular ramus All measurements in
Okkesim and 53.91 48.00
height mandible ramus
Erhamza13 *Maximum ramus presented significant
38.41 35.16
breadth differences.
*Minimum ramus
31.68 28.97
breadth
Females presented a
more obtuse gonial
angle and a narrower
*Gonial angle
chin compared with
males. These two traits
*Chin
become more distinct
during growth.
Females = more
Considering the total
Fan et al.14 obtuse gonial More
sample, mandible was
angle and a obtuse
larger in males than in
narrower chin narrower
females at all ages. The
compared with
size difference became
males. These two
greater, as the size of
traits become
the mandible increased
more distinct
more rapidly in males
during growth.
than in females. The
growth rate is similar at
9–10 years for both
sexes.
*Linear distance
from the gonion
right to menton 86.8 82.6
Albalawi et Statistically significant
*Linear distance *Angle formed by
values were found for
al.15 from the gonion 49.5 47.7 gonion right to
129.9 126.7 differences for all
left to menton menton to gonion
variables studied (p =
*Linear distance left.
0.000)
from the gonion 47.7 46.6
right to gonion left
48
Males have mostly

higher mandibular
*Right Ramus measurements on
length 6.51 5.92 panoramic radiographs
*Left Ramus 6.35 5.83 *Right gonial 117.13 120.03 and CBCT except the
Tassoker et
length 4.11 3.84 angle gonial angle. Maximum
al.17
*Right maximum *Left gonial angle 118.02 119.41 ramus breadth
ramus breadth 17.42 16.38 presented statistical
*Bigonial width significance differences
when comparing
distinct age ranges.
*Type I: (no cortication)

Males with Type I and
Mean age: 14.14
Qualitative II cortication were
*Type II: Lower density in superior
assessment of older than females.
surface
Barak et al.19 mandibular Females presented more
Mean age: 16.11
condyle Type III cortication
*Type III: Similar or higher density in
cortication than males.
superior surface
Mean age: 19.39
No statistically
significant differences
*right mandibular were found between
102.01 105.25
Barbieri et notch angle mandibular incision
al.20 *left mandibular measurements in both
102.70 103.55
notch angle sexes, and in the
different age groups
studied.
*Area of
mandibular
Foramen
*Bigonial breadth 7.18 5.95
*Direct distance
between right and 103.39 95.83
left coracoid 102.01 97.11
*Height of
symphysis Significant differences
* Min-height of 32.52 29.64 were not observed in
mandibular notch 52.33 47.82 two variables: Tongue
*Min-breadth of side bone
*Mandibular angle
mandibular ramus 34.61 32.00 thickness of mandibular
(total sample)
*Buccal side bone foramen and vertical
(18 – 24 years) 121.52 125.44
thickness of 1.03 0.93 diameter of mandibular
(25- 30 years) 122.70 125.35
Mandibular foramen. Female
Zheng et al.21 (31- 40 years) 120.44 124.78
foramen presented higher values
(41-50 years) 122.45 126.19
* Tongue side bone 5.91 5.76 of Mandibular angle.
(51- 60 years) 120.92 125.42
thickness of Considering age ranges,
(61 – 70 years) 121.83 125.62
Mandibular mandibular angle and
120.80 125.27
foramen 2.23 2.23 the other variables did
*Vertical diameter not present differences
of Mandibular 3.84 3.50 when genders were
foramen compared.
*Horizontal
diameter of 38.70 37.17
Mandibular
foramen
*Vertical from 41.60 40.37
prosthion to
palatal
breadth
*Palatal breadth
49
*Bicondylar
129.70 121.80
breadth All the breadth
100.19 93.55
*Bigonial breadth dimensions described
89.81 85.72
Deng et al.22 *Biantegonial were significantly larger
notch breadth in males than in
49.45 47.32
*Bimental females.
foramina breadth
*Ramus length
54.36 49.41
*Gonion-gnathion All variables showed
70.37 67.14
length differences between sex
*Minimum ramus except minimum ramus
Gamba et al.25 28.70 28.91 *Gonial angle 121.28 119.83
breadth breadth. Gonial angle in
*Bigonial breadth male was larger than in
118.48 110.03
*Bicondylar female.
94.96 87.47
breadth
*Bicondylar
breadth 130.00 121.44
*Bigonial breadth 100.281 93.594
*Bi-antegonial 90.072 85.674
notch breadth
*Bi-mental 49.734 47.228
foramina breadth
*Distance between 15.297 14.006
mental foramen
and mandibular Mental Angle was the
*Mandibular angle 123.444 126.648
inferior border only measurement that
Dong et al.26 *Mental angle
*Maximum did not presented
72.909 71.974
mandibular ramus 45.413 41.996 statistical significance.
breadth
*Maximum
mandibular length 126.198 117.398
*Maximum
mandibular ramus 65.962 58.243
height
*Maximum
mandibular body 86.458 81.462
length
*Ramus length
61.67 54.72
*Min ramus
29.89 28.09 The mean values of
breadth
mandibular
71.86 67.73 *Gonial angle
*Gonion–gnathion measurements were
İlgüy et al.28 121.14 122.31
length greater for males than
females except for
*Bigonial breadth 100.33 94.77 gonial angle.
*Bicondylar 120.79 116.23

breadth
50
Figure 2. Risk of bias assessment according to Robvis
51
RESULTS breadth,9, 11, 22, 23, 25, 27, 30, 31 and bigonial breadth.9,
A total of 23 studies were included. 9-31 Eleven 21-23, 25, 27, 30, 31 Thus, means comparisons were
studies perform forensic assessments in MCT9-11, limited to some populations, as Brazilians, 25

16, 18, 23, 24, 26, 28-30 and twelve studies in CBCT.12-15, 17, Chinese (central or Han), 21, 22, 31 European
19-22, 25, 27, 31 Some author also evaluate other bones descendants, 27 Turkish, 17 French, 9 Israeli, 23
but mandible, such as the femur, 29 maxilla21 and Egyptian,30 Lybian and11 Korean.28 Meta-analysis
skul.l9, 10 The number of CT scans evaluated results are available on Figures 3 to 5.
ranged from 6020 to 65414 and the age of the Data provided by authors regarding to age was
patients that performed the CT scans ranged insufficient to perform statistical assessments.
from 630 to 10026 years old, except for a study
that included fetuses. 29 The origin of the a) Mandibular angle
population included in the samples were highly For mandible statistical analysis, it was included
heterogeneous and highly specific, as Libyan in statistical model four studies11, 23, 28, 30 which
from Tripoli11 or Turkish from Middle Anatolia.17 performed MCT and five that performed CBCT
Data about the year of publication, country of in separate. 17, 21, 25, 27, 31 Mandible angle presented
the assessments origin, bones included, type of significant differences between males and females
CT and sample features are available on Table 1. in the CBCT model but not presented in MCT
Age alone was the aim of a single study, which model. Figure 3A and 3B demonstrates the meta-
investigated mandibles of fetuses.29 Sex and age analysis graphics.
determinations was the objective of five studies, For MCT, the observed standardized mean
12, 16, 17, 19-21 although age determination was a differences ranged from -1.70 to 3.49. The
secondary data in two of them. 16, 21 The other estimated average standardized mean difference
studies investigated the possibility to determine based on the random-effects model was 1.40
sex using mandible bone data with different (95% CI: -1.46 to 4.27). The average outcome did
methodologies.9-11, 13-15, 17, 18, 20, 22-28, 30, 31 not differ significantly from zero (z = 0.96; p =
In Table 2, the methodology applied in the 0.34). The presence of heterogeneity was
studies, the results and the main conclusions are observed. Results are available on Figure 3A.
summarized. Most of the studies used quantitative For CBCT, the observed standardized mean
analysis in their methodologies, with linear or differences ranged from -1.45 to 3.92. The
angular measurements, 9, 11-18, 20-24, 26-31 although estimated average standardized mean difference
some of them used morphologic assessments10, 14, based on the random-effects model was 2.10
18, 19 or created nominal/qualitative classifications (95% CI: 0.32 to 3.87). The average outcome
for determining age or/and sex, 14, 18, 19 or even for differed significantly from zero (z = -2.31, p =
machine learning. 10 0.02). The presence of heterogenicity was
The landmarks, measures and/or classifications detected although studies have good scores on
applied by the authors that used MCT, as well as quality assessments. Results are available on
the statistical evaluations results of each Figure 3B.
investigation are available on Table 3. For CBCT
is available in Table 4. b) Bicondylar breadth
For bicondylar statistics analysis, both MCT9, 11,
Meta-analysis assessments 23, 30 CBCT 22, 25, 27, 31 four studies were included
The investigators used a highly heterogeneous and results confirmed significant differences
landmarks and measures or qualitative between male and females. Graphics are available
classifications, which limited the articles on Figure 4A and 4B.
included in meta-analysis. First, as it was For MCT mean differences ranged from 6.70 to
necessary to include in each assessment the same 9.84. The estimated average standardized mean
type of CT as measurements varies between CT difference based on the random-effects model
and CBCT.32 Secondly, it is also needed to include was 7.97 (95% CI: 6.29 to 9.65). The average
measurements using the exactly the same outcome differed significantly from zero (z = 9.28,
landmarks or sites of mandible. Considering the p < 0.0001). Even though there may be some
aforementioned, it was included both for MCT heterogeneity, the true outcomes of the studies
and CBCT for sex comparisons: mandibular angle are generally in the same direction as the
(gonial angle),11, 17, 21, 23, 25, 27, 28, 30, 31 bicondylar estimated average outcome. (Figure 4A)
52
For CBCT, the observed standardized mean differed significantly from zero (z = 10.32, p <
differences ranged from -8.56 to -4.91. The 0.0001). Even though there may be some
estimated average standardized mean difference heterogeneity, the true outcomes of the studies
based on the random-effects model was -7.34 are generally in the same direction as the
(95% CI: -8.37 to -5.94). The average outcome estimated average outcome. (Figure 4B)
Figure 3. Mandible angle meta-analysis results for multislice computed tomography (MCT) and cone-
beam computed tomography (CBCT)
Figure 4. Bicondylar breadth meta-analysis results for multislice computed tomography (MCT) and
cone-beam computed tomography (CBCT)
53
c) Bigonial breadth <0.00001). An examination of the studentized

For bigonial breadth analysis, three MCT9, 23, 30 residuals revealed that one study (Kharoshah et
studies and five for CBCT. 21, 22, 25, 27, 31 Both al.30) may be a potential outlier in the context of
assessments confirmed the statistical significant this model. Results are available on Figure 5A.
differences between male and females, as Considering CBCT, the observed standardized
demonstrated on Figure 5A and 5B. mean differences ranged from 5.56 to 8.45. The
For MCT, the observed standardized mean estimated average standardized mean difference
differences ranged from 4.00 to 7.42. The based on the random-effects model was 7.03 (95%
estimated average standardized mean difference CI: 6.16 to 7.89). The average outcome differed
based on the random-effects model was 6.96 significantly from zero (z = 15.97, p < 0.00001).
(95% CI: 5.99 to 7.93). The average outcome Results and metanalysis graph are available on
differed significantly from zero (z = 24.10, p = Figure 5B.
Figure 5. Bigonial breadth meta-analysis results for multislice computed tomography (MCT) and cone-
beam computed tomography (CBCT)
DISCUSSION methodologies, measurements, evaluations, and

The process of identifying humans from physical populations included.
features is not restricted to the identification of The minority of the studies aimed to associate a
individuals who are declared dead. It is also used mandibular assessment of any nature with age.12,
to identify asylum seekers (e.g., people without 17, 19, 20, 29 Minier et al.29 observed a significant
valid identification documents), unaccompanied correlation between a ge and mandible

minors,33 and victims and perpetrators of crimes measurements in fetuses and that the coronoid
and war atrocities (e.g., in criminal process to condylar process and coronoid process
prosecution).34 Hence, because living individuals to mandibular angle measurements had the
must sometimes be identified, the use of non- highest correlations with age. Other studies
invasive methods, such as CT, is essential. The estimated skeletal age based on the mandible
present review has shown that diverse methods dimorphism inherent to sex variation. On this
based on MCT or CBCT have been used to point, al l the col lected information wa s
determine age or sex for individual identification. complementary. Motawei et al.12 reported that
Essentially, the studies selected in this review mandible ramus length correlates with age,
were highly heterogeneous in terms of the particularly in the age range of 17–57 years old.
54
Also, Tassoker et al.17 verified that the only this structure actually shows dimorphism or
mandibular bone measurement that correlated whether the imaging technique influences the
with age was the maximum ramus breadth, which final results. Considering the meta-analysis
has lower values in 10–19 year-olds compared to assessment method, more CBCT-based studies
60–69 year-olds. were included in the statistical analysis, and this
When compared to the other facial bones, the may have influenced the statistical significance
mandible exhibits the greatest growth and observed.
morphological size and remodeling changes.35 The bicondylar breadth was found to be greater
Considering that the mandible is isolated, others in males than in females,9, 11, 18, 22, 23, 25-27, 30, 31 as well
have concluded that the mandible ramus is the as the bigonial breadth.21-23, 25-27, 30, 31 Statistically
structure that best represents the remodeling significant differences were found in al l
changes that occur in certain age ranges and that populations studied.
changes in the mandible ramus strongly correlate In contrast to the findings of other studies
with age.35, 36 The findings of the included studies included in the model, in a study with a Libyan
corroborate these conclusions.12, 17 population, Atef et al. 11 found that the
Furthermore, the size and shape of the mandible bicondylar breadth had smaller mean values in
are also used to predict an individual’s sex.35 females than males. The same observation was
Although sexual dimorphism is present at birth,37 made by İlgüy et al.,27 who studied a population
sex differences decrease rapidly during early life12, of European descendants. These findings lead to
37 and only resume during the phase of puberty to the question: Do these Libyan and European-
adulthood37 with the influence of sex hormones.12 descendant populations differ from the other
Thus, sex dimorphism is not only reflected in the populations studied, or were the selected
size of the mandible, but also in its shape.10, 14, 18 individuals not an appropriate representative
The mandible angle is one of the most-studied sample of the populations?
factors responsible for shape differences and In terms of sex differences, it was found that
more obtuse angles have been found in females.14 both bicondylar breadth and bigonial breadth
In MCT-based studies, the mandible angle was could be used to determine sex. However, a study
found to have higher mean values in females of an Egyptian population showed outlier
compared to males.9, 11, 23, 24, 28 An exception was results.30 Hence, conducting larger studies that
reported by Karoshah et al.,30 who found that the include distinct populations worldwide could
angles were greater in males than in females in an answer the question of whether bicondylar
Egyptian population. In CBCT-based studies, the breadth, bigonial breadth, and mandible angle
results were similar,17, 27, 31 except in Gamba et measurements correlate with sex.
al.’s25 study of a Brazilian population. In a study
that included different age groups of males and CONCLUSION
females, Zheng et al. 21 also did not observe Considering the studies included in this review,
statistically significant differences in mandible we conclude that mandible measurements are
angle between the sexes. useful for sex determination, as both the
Hence, this meta-analysis of studies that focused bicondylar and bigonial breadth have been found
on the mandible angle has shown that the results to have higher values in males than in females.
of MCT-based studies are dissimilar, with no Regarding the mandible angle, the meta-analysis
significant differences found between males and results confirm that sex differences can be
females, and that CBCT-based studies have detected using CBCT scans but not MCT scans.
found significant differences between males and In terms of age estimation, further studies are
females. These findings raise the issue of whether needed to prove that the hole of the mandible is
a reliable parameter for age estimation.
REFERENCES
1. Suzuki K, Nakano H, Inoue K, Nakajima Y, Mizobuchi S, sciences. A review of the Virtopsy and similar projects. Eur
Omori M, et al. Examination of new parameters for sex Radiol. 2008;18:273-82.
determination of mandible using Japanese computer tomography 3. Nardi C, Talamonti C, Pallotta S, Saletti P, Calistri L,
data. Dentomaxi$ofac Radiol. 2020;49:20190282. Cordopatri C, et al. Head and neck effective dose and
2. Bolliger SA, Thali MJ, Ross S, Buck U, Naether S, Vock P. quantitative assessment of image quality: a study to compare
Virtual autopsy using imaging: bridging radiologic and forensic
55
cone beam CT and multislice spiral CT. Dentomaxi$ofac Radiol. 19. Bayrak S, Halıcıoglu S, Kose G, Halıcıoglu K. Evaluation
2017;46:20170030. of the relationship between mandibular condyle cortication and
4. Knobloch K, Yoon U, Vogt PM. Preferred reporting items chronologic age with cone beam computed tomography. J
for systematic reviews and meta-analyses (PRISMA) statement Forensic Leg Med. 2018;55:39-44.
and publication bias. J Craniomaxi$ofac Surg. 2011;39:91-2. 20.Barbieri A.A. BD, Borges J.L.C. , Oliveira E. , Assis
5. Ouzzani M, Hammady H, Fedorowicz Z, Elmagarmid A. A.C.S. , Manhaes Junior L.R. , Lopes S.L.P.C.
Rayyan-a web and mobile app for systematic reviews. Syst Rev. Evaluation of the mandibular notch angle in three-dimensional
2016;5:210. virtual models as a parameter for determining age and sexual
6. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, dimorphism. Brazilian Dental Sciences. 2018;21:424 - 31.
Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 21. Zheng J, Ni S, Wang Y, Zhang B, Teng Y, Jiang S. Sex
statement: an updated guideline for reporting systematic reviews. determination of Han adults in Northeast China using cone
Syst Rev. 2021;10:89. b e a m c o m p u t e r t o m o g ra p h y. Fo r e n s i c S c i In t.
7. Sterne JAC HM, McAleenan A, Reeves BC, Higgins JPT. 2018;289:450.e1-.e7.
Chapter 25: Assessing risk of bias in a non-randomized study. 22. Deng M. a BRb, c , Dong H. d , Mu J. d , Lin W. d
In: Higgins JPT, Thomas J, Chandler J, Cumpston M, Li , Zhou Y. d. Sexual determination of the mandible breadth in
T, Page MJ, Welch VA (editors). Cochrane Handbook for a central Chinese population sample: a three-dimensional
Systematic Reviews of Interventions version 6.0 (updated July analysis. Australian Journal of Forensic Sciences. 2017;49:332 -
2019). C o c h ra n e, 2 0 1 9. Av a i l a b l e from 43.
www.training.cochrane.org/handbook . 2019. 23. Tunis T.S. SR, Cohen H. , Medlej B. , Peled N. , May
8. McGuinness LA, Higgins JPT. Risk-of-bias VISualization H. Sex estimation using computed tomography of the
(robvis): An R package and Shiny web app for visualizing risk- mandible. Inter national Jour nal of L egal Medicine.
of-bias assessments. Res Synth Methods. 2021;12:55-61. 2017;131:1691 - 700.
9. Gillet C, Costa-Mendes L, Rérolle C, Telmon N, Maret 24. Inci E, Ekizoglu O, Turkay R, Aksoy S, Can IO, Solmaz
D, Savall F. Sex estimation in the cranium and mandible: a D, et al. Virtual Assessment of Sex: Linear and Angular Traits
multislice computed tomography (MSCT ) study using of the Mandibular Ramus Using Three-Dimensional Computed
anthropometric and geometric morphometry methods. Int J Tomography. J Craniofac Surg. 2016;27:e627-e32.
Legal Med. 2020;134:823-32. 25. Gamba TeO, Alves MC, Haiter-Neto F. Mandibular sexual
10. Imaizumi K BE, Taniguchi K, Ogawa T, Nagata dimorphism analysis in CBCT scans. J Forensic Leg Med.
T, Kaga K, Hayakawa H, Shiotani S. 2016;38:106-10.
Development of a sex estimation method for sku$s using machine 26. Kano T, Oritani S, Michiue T, Ishikawa T, Hishmat AM,
learning on three-dimensional shapes of sku$s and sku$ parts. Sogawa N, et al. Postmortem CT morphometry with a
Forensic Imaging. 2020;22. proposal of novel parameters for sex discrimination of the
11. Atef M GA, Khalid AOA , Ismail MMR. Reliability of mandible using Japanese adult data. Leg Med ( Tokyo).
Three Dimensional Computed Tomography in Sex 2015;17:167-71.
Determination (om Mandible of a Sample of Libyan 27. İlgüy D, İlgüy M, Ersan N, Dölekoğlu S, Fişekçioğlu E.
Population in Tripoli. Bioscience Research. 2020;17:4253-60. Measurements of the foramen magnum and mandible in relation
12. Motawei SM Helaly AMN ael Aboelmaaty to sex using CBCT. J Forensic Sci. 2014;59:601-5.
WM Elmahdy HuO. Length of the ramus of the mandible 28. Lin C, Jiao B, Liu S, Guan F, Chung NE, Han SH, et al.
as an indicator of chronological age and sex: A study in a group Sex determination (om the mandibular ramus flexure of
of Eg yptians. Forensic Science International Reports. Koreans by discrimination function analysis using three-
2020;2:100066. d i m e n s i o n a l m a n d i b l e m o d e l s. Fo r e n s i c S c i In t.
13. Okkesim A, Sezen Erhamza T. Assessment of mandibular 2014;236:191.e1-6.
ramus for sex determination: Retrospective study. J Oral Biol 29. Minier M, Dedouit F, Maret D, Vergnault M, Mokrane
Craniofac Res. 2020;10:569-72. FZ, Rousseau H, et al. Fetal age estimation using MSCT scans
14. Fan Y, Penington A, Kilpatrick N, Hardiman R, of the mandible. Int J Legal Med. 2014;128:493-9.
Schneider P, Clement J, et al. Quantification of mandibular 30. Kharoshah MA, Almadani O, Ghaleb SS, Zaki MK,
s e x u a l d i m o r p h i s m d u r i n g a d o l e s c e n c e. J An a t. Fattah YA. Sexual dimorphism of the mandible in a modern
2019;234:709-17. Egyptian population. J Forensic Leg Med. 2010;17:213-5.
15. Albalawi AS, Alam MK, Vundavalli S, Ganji KK, Patil S. 31. Dong H, Deng M, Wang W, Zhang J, Mu J, Zhu G. Sexual
Mandible: An Indicator for Sex Determination - A Three- dimorphism of the mandible in a contemporary Chinese Han
dimensional Cone-Beam Computed Tomography Study. Contemp population. Forensic Sci Int. 2015;255:9-15.
Clin Dent. 2019;10:69-73. 32. Poeschl PW, Schmidt N, Guevara-Rojas G, Seemann R,
16. Bulut O, Freudenstein N, Hekimoglu B, Gurcan S. Ewers R, Zipko HT, et al. Comparison of cone-beam and
Dilemma of Gonial Angle in Sex Determination: Sexua$y conventional multislice computed tomography for image-guided
Dimorphic or Not? Am J Forensic Med Pathol. 2019;40:361-5. dental implant planning. Clin Oral Investig. 2013;17:317-24.
17. Tassoker M, Akin D, Aydin Kabakci AD, Sener S. 33. Schmeling A, Geserick G, Reisinger W, Olze A. Age
Comparison of cone-beam computed tomography and panoramic estimation. Forensic Sci Int. 2007;165:178-81.
radiography for mandibular morphometry. Folia Morphol 34. Schmeling A, Reisinger W, Geserick G, Olze A. Age
(Warsz). 2019;78:862-70. estimation of unaccompanied minors. Part I. General
18. Alias A, Ibrahim A, Abu Bakar SN, Swarhib Shafie M, considerations. Forensic Sci Int. 2006;159 Suppl 1:S61-4.
Das S, Abdullah N, et al. Anthropometric analysis of 35. de Oliveira FT, Soares MQ, Sarmento VA, Rubira CM,
mandible: an important step for sex determination. Clin Ter. Lauris JR, Rubira-Bullen IR. Mandibular ramus length as an
2018;169:e217-e23. indicator of chronological age and sex. Int J Legal Med. 2015;12:195-201.
56
36. Franklin D, Cardini A, O'Higgins P, Oxnard CE, Dadour 37. Coquerelle M, Bookstein FL, Braga J, Halazonetis DJ,
I. Mandibular morphology as an indicator of human subadult Weber GW, Mitteroecker P. Sexual dimorphism of the
age: geometric morphometric approaches. Forensic Sci Med human mandible and its association with dental development.
Pathol. 2008;4:91-9. Am J Phys Anthropol. 2011;145:192-202.
57

108 105 PB

Uploaded by

Copyright:

Available Formats

108 105 PB

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

108 105 PB

Uploaded by

Copyright:

Available Formats

ISSN 2219-6749

OFFICIAL PUBLICATION OF THE

THE JOURNAL OF FORENSIC ODONTO-STOMATOLOGY

Webmaster and Publishing Director

The role of orthodontics in children identification: a Aliki Rontogianni , 2

Canine sexual dimorphism in crown and root Mohammad Tajik, 12

Artificial intelligence in age and sex determination Sraddha Singh, 30

Are computed tomography images of the mandible Luciana Munhoz, 38

The role of orthodontics in children identification:

Copyright © 2024 International Organization ABSTRACT

Figure 5. Orthodontic plaster models

Clin Diagn Res. 2014;8(7):ZE01-3. identification. Acta odontol Venez. 2005;43:67–74.

Canine sexual dimorphism in crown and root

Copyright © 2024 International Organization

Variables Mean±SD P value

MD 7.12±0.47 6.63±0.51 <0.001*

BL 8.27±0.56 7.57±0.56 <0.001*

Maxilla TL 26.30±1.82 23.43±1.86 <0.001*

RL 19.16±1.84 17.01±1.67 <0.001*

TL 24.70±1.81 22.47±1.79 <0.001*

CEJMD 5.47±0.53 4.84±0.56 <0.001*

CEJBL 7.10±0.51 6.52±0.54 <0.001*

TL/RL 1.37±0.07 1.38±0.06 0.601

BL/MD 1.16±0.08 1.14±0.08 0.150

RL/BL 2.32±0.23 2.26±0.24 0.038*

RL/MD 2.70±0.28 2.58±0.28 0.002*

TL/BL 3.19±0.25 3.10±0.28 0.026*

TL/MD 3.70±0.31 3.55±0.35 0.001*

TL/CEJBL 3.50±0.28 3.41±0.30 0.026*

RL/CEJBL 2.56±0.26 2.48±0.27 0.040*

MD/CEJMD 1.28±0.09 1.36±0.12 <0.001*

CEJBL/ 1.36±1.35 1.42±0.15 0.003*

TL/CEJMD 4.74±0.47 4.82±0.53 0.0727

RL/BL 2.25±0.26 2.22±0.27 0.440

RL/MD 2.74±0.34 2.73±0.33 0.788

TL/BL 3.19±0.26 3.15±0.29 0.331

TL/MD 3.70±0.31 3.54±0.34 0.839

TL/CEJBL 3.50±0.28 3.40±0.30 0.577

RL/CEJBL 2.46±0.29 2.44±0.30 0.042

MD/CEJMD 1.18±0.12 1.21±0.13 0.041*

CEJBL/ 1.30±0.14 1.36±0.15 0.024*

TL/CEJMD 4.54±0.49 4.67±0.55 0.0751

RL/CEJMD 3.21±0.41 3.29±0.46 0.0660

The results of the independent T-test are shown in the table

TL2 0.463 0.857

RL2 -0.270 -0.475

CEJMD2 0.976 0.458

MD1 0.571 0.253

CEJMD1 0.434 0.217

Maxillary CEJMD 1.286 0.603

CEJBL 0.529 0.277

CEJMD 0.952 0.477

Mandibular MD 0.962 0.428

CEJBL 0.542 0.286

Maxillary MD/CEJMD 7.995 0.841

* Canonical Discriminant Function Coeﬃcient

Machine learning assisted 5-part tooth segmentation

Copyright © 2024 International Organization ABSTRACT

MATERIAL AND METHODS 35-35.99 5 5 5 5 5 5