International Journal of Legal Medicine (2023) 137:1117–1146

https://doi.org/10.1007/s00414-023-02960-z

REVIEW

A systematic overview of dental methods for age assessment

in living individuals: from traditional to artificial
intelligence-based approaches
Nicolás Vila-Blanco1,2,3 · Paulina Varas-Quintana4 · Inmaculada Tomás1,3,4 · Marı́a J. Carreira1,2,3

Received: 29 November 2022 / Accepted: 31 January 2023 / Published online: 14 April 2023
© The Author(s) 2023

Abstract
Dental radiographies have been used for many decades for estimating the chronological age, with a view to forensic
identification, migration flow control, or assessment of dental development, among others. This study aims to analyse the
current application of chronological age estimation methods from dental X-ray images in the last 6 years, involving a search
for works in the Scopus and PubMed databases. Exclusion criteria were applied to discard off-topic studies and experiments
which are not compliant with a minimum quality standard. The studies were grouped according to the applied methodology,
the estimation target, and the age cohort used to evaluate the estimation performance. A set of performance metrics was used
to ensure good comparability between the different proposed methodologies. A total of 613 unique studies were retrieved,
of which 286 were selected according to the inclusion criteria. Notable tendencies to overestimation and underestimation
were observed in some manual approaches for numeric age estimation, being especially notable in the case of Demirjian
(overestimation) and Cameriere (underestimation). On the other hand, the automatic approaches based on deep learning
techniques are scarcer, with only 17 studies published in this regard, but they showed a more balanced behaviour, with no
tendency to overestimation or underestimation. From the analysis of the results, it can be concluded that traditional methods
have been evaluated in a wide variety of population samples, ensuring good applicability in different ethnicities. On the other
hand, fully automated methods were a turning point in terms of performance, cost, and adaptability to new populations.

Keywords Dental radiology · Chronological age estimation · Forensic dentistry · Deep learning

Introduction adoption processes of undocumented children. In all these

cases, an expert performs a somatic maturity examination.
Chronological age is, together with biological sex and eth- The development status of bones has been used
nicity, the most important human feature to be considered in successfully to estimate chronological age. In this regard,
anthropological and forensic studies [1]. Besides, chrono- many skeletal parts have been used, such as pubic
logical age estimation is used daily in legal procedures symphysis, auricular surface, or sternal ribs [2]. Also, it is
where the birthdate of the involved subjects can not be ver- worth noting that there is not a single method based on
ified due to either the absence of birth certification or the bone development that outperforms others systematically, as
suspicion of false documentation. This applies to migra- the performance of each one depends on numerous factors.
tion controls or trials involving undocumented people since For instance, there are specific age estimation methods
the attainment of legal age has many implications accord- developed for subadults and others that work better in adults
ing to the laws of most countries. It is also important in the [3].
One of the most widely used body part in the field of age
 Inmaculada Tomás
estimation is the teeth, mainly because dental mineralisation
[email protected] has been reported to be less affected by external factors
(e.g. genetics or environment) than bone mineralisation [4].
 Marı́a J. Carreira In this regard, dental imaging techniques represented a step
[email protected]
forward because they allowed clinicians to assess bone
development with less invasive and faster procedures, and
Extended author information available on the last page of the article. thus enabled them to perform chronological age estimation.
1118 International Journal of Legal Medicine (2023) 137:1117–1146

The estimation of age from dental radiographic records EXCLUDE ( SUBJAREA , “BUSI” ) OR EXCLUDE
is based on the evaluation of some characteristics such as ( SUBJAREA , “CENG” ) OR EXCLUDE ( SUB-
JAREA , “CHEM” ) OR EXCLUDE ( SUBJAREA ,
the formation of jaw bones; the appearance of tooth germs,
“ENER” ) OR EXCLUDE ( SUBJAREA , “PSYC” )
the degree of crown completion and its eruption, the degree )
of resorption of deciduous teeth; the measurement of open
The query used in PubMed was:
apices in teeth; the volume of the pulp chamber and root (age estimation [Title/Abstract] OR age assessment
canals; the formation of physiological secondary dentin; the [Title/Abstract] OR age regression [Title/Abstract]
toot-to-pulp ratio; or the development and topography of the OR age determination [Title/Abstract]) AND (den-
third molar [5]. tal [Title/Abstract] OR tooth [Title/Abstract] OR
teeth [Title/Abstract] OR incisor [Title/Abstract] OR
It is worth noting that the panoramic X-rays (ortopanto- canine [Title/Abstract] OR premolar [Title/Abstract]
mographies or OPGs) provide the least invasive radiologic OR molar [Title/Abstract] OR mandib*
technique to estimate age, as it only requires a single image [Title/Abstract]) AND (x-ray [Title/Abstract]
to capture the whole dentition. Besides, other bone struc- OR radiograph* [Title/Abstract] OR radiolog*
[Title/Abstract] OR opg [Title/Abstract] OR
tures can be seen, such as the mandible, the nasal fossa, orthopantomograph* [Title/Abstract] OR panoramic
or the vertebrae, which are also helpful for further exami- [Title/Abstract] OR CT [Title/Abstract] OR MRI
nations. In the following, a review of the main methods to [Title/Abstract] OR CBCT [Title/Abstract]) AND
estimate the age of dental radiographs has been carried out. 2016[DP]: 2022[DP] AND full text [SB] AND
english [LA]
As it can be seen, the query is not strictly the same,
Material and methods as Scopus allowed also for excluding certain unwanted
subject areas, such as Veterinary or Arts. As a result, a
For the review purpose, a conducting protocol approved set of 537 studies were collected from Scopus and 336
by an expert reviewer and compliant with the PRISMA from Pubmed on February 24th, 2022, which in the end
guidelines for systematic reviews [6] has been established. represented a body of 613 unique works. The abstract
Scopus and PubMed databases have been used to retrieve of each work was reviewed to discard unwanted studies,
a collection of full-text studies on age estimation from according to the following exclusion criteria: (1) studies
dental radiographies published in the last 6 years (from not aimed at chronological age estimation in humans; (2)
2016 to 2022). This specific period was chosen for two non-radiological studies; (3) studies that use non-human
main reasons. First, the number of published studies samples; (4) studies relying on a sample smaller than 50
is sufficiently high to report significant conclusions. subjects or studies that do not report the sample size; (5)
Second, automatic methodologies in the field of dental age studies whose full text is not available.
estimation have been mainly used in this period, and not Regarding the collection of studies aimed at evaluating
before, and therefore including earlier years would have the age estimation methods proposed in the literature,
diluted their relevance in this review. Then, a study selection only those reporting at least one of the following metrics
process has been carried out, as seen in Fig. 1. were evaluated. In terms of numerical age estimation
The query used in Scopus was: studies, a statistic on the residual error (dental age
TITLE-ABS-KEY ( ( “age estimation” OR “age minus chronological age) and the absolute error—mean,
assessment” OR “age regression” OR “age determina- median, or standard deviation—, the standard error of
tion” ) AND ( dental OR tooth OR teeth OR mandib*
OR incisor OR canine OR premolar OR molar ) AND
the estimates, and/or the coefficient of determination R 2 .
( x-ray OR radiolog* OR radiograph* OR opg OR Methods geared toward age classification were required to
orthopantomograph* OR panoramic OR ct OR cbct report the accuracy, sensitivity, and/or specificity of the
OR mri ) ) AND ( LIMIT-TO ( DOCTYPE , “ar” ) classification results. Although dental development is less
OR LIMIT-TO ( DOCTYPE , “re” ) OR LIMIT-TO
affected by genetic or environmental factors than other
( DOCTYPE , “cp” ) ) AND ( LIMIT-TO ( PUB-
YEAR , 2022 ) OR LIMIT-TO ( PUBYEAR , 2021 ) bones, this process is still subject to variations, and so the
OR LIMIT-TO ( PUBYEAR , 2020 ) OR LIMIT-TO age estimation methods were usually assessed in different
( PUBYEAR , 2019 ) OR LIMIT-TO ( PUBYEAR populations and/or ethnic groups all over the world.
, 2018 ) OR LIMIT-TO ( PUBYEAR , 2017 ) OR
To reduce as much as possible the risk of bias in this work
LIMIT-TO ( PUBYEAR , 2016 ) ) AND ( LIMIT-
TO ( LANGUAGE , “English” ) ) AND ( EXCLUDE when comparing the results obtained by different methods,
( SUBJAREA , “BIOC” ) ) AND ( EXCLUDE the collected studies were analysed to detect evidence of
( SUBJAREA , “PHAR” ) OR EXCLUDE ( malpractice. As a result, five studies were discarded due
SUBJAREA , “PHYS” ) OR EXCLUDE ( SUB-
to the non-compliance with basic aspects such as good
JAREA , “VETE” ) OR EXCLUDE ( SUBJAREA ,
“ARTS” ) OR EXCLUDE ( SUBJAREA , “EART” wording or a comprehensive description of the experiments,
) OR EXCLUDE ( SUBJAREA , “MATE” ) OR as this could also indicate a problem in the peer review
International Journal of Legal Medicine (2023) 137:1117–1146 1119

Fig. 1 Flow diagram of the

study selection process

process. It is worth noting that only the most flagrant Children and young adults
cases were taken into account to minimise the bias that the
observer could introduce in this evaluation process. In the Age estimation via dentition analysis has reportedly led
end, a set of 286 studies was selected for further analysis. to better results when dealing with newborns to subjects
aged 22 to 25, that is, during tooth development. This
makes studies aimed at estimating the age of children and/or
Age estimation methods adolescents to be more numerous than those focusing on age
estimation in adults. Regarding the former, some methods
Tooth-based manual methods aim to assess specific development milestones (such as
dental eruption) to predict age [7, 40]. Though, they have
The studies retrieved in this work relied on a wide proven to lead to very limited estimates, as they rely on
variety of age estimation methodologies. However, as dental very quick changes, from which little information can be
formation is highly correlated with chronological age and, collected.
therefore, is a key indicator for age estimation, most Other methods aimed to assess the development of
methods are based on dentition analysis. In this regard, the the teeth over a longer period. That is the case for
first approaches were purely manual, that is, they required dental Atlases, which are graphic representations of dental
experts not only to retrieve the correspondent information development and eruption that provide an easy way to
from the X-ray image but also to translate this information estimate chronological age by comparing the status of the
into an age value. These approaches [7–39] are shown in dentition using radiological or osteological techniques to the
Fig. 2. charts provided in the Atlas [8–11]. Other authors went a
1120 International Journal of Legal Medicine (2023) 137:1117–1146

Fig. 2 Main manual methods

for estimating chronological age
from dental X-ray images

step further and developed dental scoring systems (DSS), stages, and Liliequist and Lundberg [21] proposed the
consisting of dividing the development period of each tooth use of seven stages, in a clear attempt to reduce the
into a set of developmental stages with associated scores and complexity of the method. Furthermore, some authors
using those scores to estimate the numerical chronological developed population-specific scoring tables on top of the
age. In this regard, the number of stages varied depending Demirjian et al. [14, 16, 17] and Gleiser and Hunt’s
on the specific system. For example, Gleiser and Hunt [18] systems [19, 20], while others mixed several staging
proposed 15 stages, Nolla [12] developed a division into systems to improve the overall estimation performance
11 stages, Demirjian et al. [13] reported eight alphabetical [22].
International Journal of Legal Medicine (2023) 137:1117–1146 1121

Cameriere et al. [23] introduced a different method for Non-numeric age estimation
estimating age, based on tooth measurements. Specifically,
the authors measured the open apices of the seven Besides age estimation methods developed for obtaining a
left permanent mandibular teeth. These measurements, numeric and continuous output, other authors focused on
previously normalised by tooth height, were highly and designing classification methods to estimate the probability
negatively correlated with chronological age. Furthermore, that a subject belongs to a specific age group. Most of these
the number of teeth with completely closed root apices studies relied on conventional numerical age estimation
was reportedly correlated with age. These findings led the methods and adapted them to be used as age group
authors to develop a regression formula that depends on the classifiers. This is the case, for example, of the study of
sex of the subject and the normalised measurements of the Sehrawat and Singh [41], which used the Kvaal et al.’s
seven teeth and the number of teeth whose root development method [26] to perform a classification into four groups.
is completed. However, the majority of these studies are focused on a
binary classification with two groups of subjects younger
Adults and older than a given threshold, which can be the legal age
of maturity or any other specific age with high relevance
Although the development of teeth ends once the third in legal procedures. In this regard, Mincer et al. [42] relied
molar is completely developed, some authors focused on on the staging system proposed by Demirjian et al. [13] to
other age-related changes that are radiologically observable assess the development of the third molar, with the objective
to estimate age in older subjects. In this regard, three of estimating the probability of being older than a certain
different families of methods can be identified. On the age for each stage. De Luca et al. [43] used measurements
one hand, some authors explored the use of specific of the open apices to establish a cut-off value of 0.08 over
measurements or ratios between them to perform age the normalised Cameriere measurement of the third molar,
estimations. For example, Kvaal et al. [26] proposed to above which the individual is considered older than the legal
measure dentin apposition indirectly through the assessment age. Other numeric age estimation methods were evaluated
of the dental pulp radiopacity. The researchers carried in age classification problems, such as Nolla’s [44] and
out several linear measurements of both the pulp and the Kohler et al.’s [45].
tooth and associated those measurements via linear ratios.
Cameriere et al. [30] proposed a similar idea, but they Age estimation on other radiologically observable
replaced the linear measurements with area assessments. structures
Another similar example is the Tooth Coronal Index (TCI),
studied by Ikeda et al. [27], which represents a height Although most age estimation methods from dental
ratio between the crown and the pulp cavity at the crown radiologic records are based on the analysis of the teeth,
level. there are other structures whose characteristics may also
On the other hand, a set of studies focused on the be useful for age estimation. In the period covered by this
visibility of some structures established staging systems systematic review (from 2016 to 2022), the number of
with which that visibility could be assessed. The struc- works is very limited and all of them rely on mandibular
tures most studied in this regard were the periodontal measurements. Some examples are the approach followed
ligament and the root pulp, with the staging systems pro- by Motawei et al. [38], who established a relationship
posed in this regard by Olze et al. [31, 33] standing between the length of the ramus and chronological age, or
out. the proposal by Acharya [39], in which the gonial angle was
Finally, some authors reported that a series of degen- used as the main age indicator.
erative changes can be assessed through a staging system
and therefore be used to estimate chronological age. In Automatic methods
this regard, Gustafson [35] set multiple evaluable criteria,
namely secondary dentin formation, periodontal recession, Recent advances in image processing have allowed for
attrition, apical translucency, cementum apposition, and automating dental age estimation methods to a greater
external root resorption. The degenerative stages proposed degree and have led to the development of numerous
in the original work, which were intended to be applied methodologies. In this regard, the authors explored the same
to extracted and ground teeth, proved to be applicable to objectives as those covered in the traditional methods, as can
radiographic images as well, as confirmed with the method- be seen in Table 1: numeric age estimation [46–55] and age
ologies proposed by other authors, such as Olze et al. [36] group classification [56–60], with the particular case of age
and Timme et al. [37]. thresholding [47, 61].
1122 International Journal of Legal Medicine (2023) 137:1117–1146

Table 1 Main automatic methods for estimating age. OPG, ortopan- numeric age; AGC, age group classification; AT, age thresholding; Mi ,
tomograph; FA, fully automatic; CNN, convolutional neural network; ith molar; P Mi , ith premolar
NAS, neural architecture search; SHN, stacked hourglass network; NA,

Reference Image FA Method Target Required teeth

Čular et al. [46] OPG  Active Appearance Model NA − Right mandibular M3

+ Radial Basis Network
Štern et al. [47] MRI χ Manual ROI crop NA+AT No specific tooth required
+ Regression CNN
De Back et al. [48] OPG  Bayesian CNN NA No specific tooth needed
Vila-Blanco et al. [49] OPG  Two-path CNN NA No specific tooth needed
Hou et al. [50] OPG  CNN designed through NAS NA No specific tooth needed
Pham et al. [51] CT  Thresholding mandible segmentation NA No specific tooth needed
+ 3D ResNet 34
Wallraff et al. [52] OPG  ResNet18 NA No specific tooth needed

Zheng et al. [53] CBCT χ Manual ROI crop NA −M1 in all four quadrants
+ Custom CNN segmentation
+ Level set segmentation refinement
+ Linear regression

Vila-Blanco et al. [54] OPG  SHN mandible segmentation NA No specific tooth needed
+ PDM shape characterisation
+ Ridge regression
Milošević et al. [55] OPG  VGG16 NA No specific tooth needed
+ Attention mechanism
De Tobel et al. [56] OPG χ Manual ROI crop AGC − Left mandibular M3
+ Staging CNN
Merdietio et al. [57] OPG  Segmentation CNN AGC − Left mandibular M3
+ Staging CNN
Banar et al. [58] OPG  Localization CNN AGC − Left mandibular M3
+ Segmentation CNN
+ Staging CNN
Kim et al. [59] OPG χ Manual ROI crop AGC − M3 in all four quadrants
+ Staging CNN
Kahaki et al. [60] OPG  Global fuzzy segmentation AGC − Left mandibular M1 , M2 and M3
+ Intensity projection
+ Staging CNN
Guo et al. [61] OPG  SE-ResNet 101 AT No specific tooth needed

Among the methods developed for numerical age esti- On the other hand, some methods were designed to
mation, some authors focused on subjects with developing perform an age group classification, with a variable number
dentitions (up to 20 or 25 years old), such as Čular et al. of target groups. While De Tobel et al. [56], Merdietio et al.
[46], De Back et al. [48], and Wallraff et al. [52], while oth- [57], and Banar et al. [58] treated the problem as a ten-
ers were targeted at a wider range of age cohorts, such as stage, Kim et al. [59] and Kahaki et al. [60] reduced the
Vila-Blanco et al. [49], Zheng et al. [53], Hou et al. [50], and number of stages to five. Moreover, Guo et al. [61] focused
Milošević et al. [55]. It is noticeable that Vila-Blanco et al. on the binary classification problem of legal age detection,
[54] also developed an automatic method to estimate age establishing the thresholds of 14, 16, and 18 years.
in children and subadults that is not focused on the entire Regarding the applied methodologies, one of the first
panoramic image or the dental area but the shape of the attempts to rely on image processing techniques was made
mandible. by Čular et al. [46], where the authors proposed the use of
International Journal of Legal Medicine (2023) 137:1117–1146 1123

an Active Appearance Model to localise the third molar and Milošević et al. [55], and Guo et al. [61] also developed one-
parameterise its shape and texture. In a second step, these stage methods that do not require the presence of specific
parameters are introduced into a neural network to estimate teeth to work.
the chronological age. As both steps do not need human
intervention, the method works automatically.
As in most of the topics involving image processing, Evaluation studies
deep neural networks (DNNs) helped not only to automatise
some tasks but also to improve their performance. The studies retrieved in this work were categorised
Regarding age estimation, De Tobel et al. [56] developed according to the age estimation methods they relied on. As
a staging system for the third molar based on modified it can be seen in Fig. 3a, where the ten most used methods
Demirjian stages and used a DNN to classify the third are represented, Demirjian et al.’s approach [13] has been
molar image crops into one of those stages. This method applied in more than a third of the studies (100 out of 286),
only required a minimum intervention of the expert to with some methods derived for it also in the first positions—
crop the region of interest to frame the third molar area. Willems et al.’s [14] and Chaillet and Willems’ [16] were
This approach was updated by Merdietio et al. [57] by applied in 45 and 7 studies, respectively. The first method
replacing the manual crop step with a DenseNet network, not aimed originally at estimating the numerical age is the
which allows estimation to run automatically. Banar et al. approach proposed by Cameriere et al. [62]. This method,
[58] developed a similar method, with a slightly more focused on the classification of subjects younger or older
complex third molar segmentation, in which the tooth is first than legal age, was used in 43 studies. On the other hand,
localised and then segmented. it is noticeable that 239 studies relied on OPGs to conduct
Kim et al. [59] followed a similar approach to that of the experiments, representing 84% of all the retrieved works
De Tobel et al. [56]. The authors also developed a two- (Fig. 3b). The rest of the studies used CT-based techniques
step approach which firstly requires a manual crop of the (such as CBCT or conventional CT) and, to a lesser extent,
third molar, although in this case the four third molars intraoral images, MRI, or the cephalometric view.
are required. In the second step, each of the four teeth is Regarding the performance of the age estimation
classified into different age groups, and the classifications methods, a maximum of one study was evaluated for each
are merged through a majority voting system. The authors population and each method, specifically that evaluated
established two different age group divisions: the first in the largest sample due to the greater significance of
grouped subjects younger than 20, subjects aged 20 to 49, the reported results. This ensured a good representation
and those over 50; the second split the middle group into of different ethnicities while avoiding overcrowded result
three subgroups, namely subjects aged 20 to 29, subjects tables. Following the same order as in the previous
aged 30 to 39, and those aged 40 to 49. section, the approach based on tooth eruption assessment
Although deep learning methods had already been proposed by Haavikko [7] was evaluated in a wide range
introduced in the studies mentioned above, De Back et al. of populations since its development, but it clearly lost
[48] proposed the use of a DNN, specifically a Bayesian popularity in comparison to other methods. As shown in
Convolutional Neural Network, as the only step to estimate Table 2, only four studies that met the inclusion criteria
chronological age. Therefore, the expert does not need to have been analysed, all focused on subjects younger than 16.
specify which features of the image should be taken into In terms of performance, these studies reported systematic
account, as the network focuses automatically on those underestimations given by residual errors (difference
regions which contributes the most to the age estimation. between estimated age and real age) with means ranging
Furthermore, the age estimation process can proceed even if from −0.22 to −1.35 years and standard deviations around
several teeth are missing. one year. The mean absolute errors yielded mean values of
Vila-Blanco et al. [49], following the clinical finding that 0.33 to 1.45 years.
dental development is different in boys and girls, developed The atlas-based methods listed in Fig. 2 were also applied
a method to automatically integrate sexual information into in the last few years, although only the London Atlas
the age estimation process. Thus, they proposed the use proposed by AlQahtani et al. [10] was tested in more than
of two identical CNNs, one for age estimation and the one population sample. As shown in Table 3, the work by
other for sex classification, so that the sex CNN learns Baylis and Bassed [63], which compared the three Atlas-
sexual dimorphic features and propagates them to the age based methods in a New Zealander population, reported
CNN at intermediate points to improve the age estimation a slight underestimation with the Schour and Massler
performance. Atlas [8] (−0.03 to −0.39 years of mean error), a slight
Similarly to that proposed by De Back et al. [48] and overestimation with the Blenkin and Taylor method [11]
Vila-Blanco et al. [49], Hou et al. [50], Wallraff et al. [52], (+0.07 to +0.34 years), and a noticeable overestimation
1124 International Journal of Legal Medicine (2023) 137:1117–1146

Fig. 3 a Number of studies applying the ten most assessed methods that used pulp-to-tooth volumetric ratios; b Number of studies relying
for age estimation in dental radiographic images in the period evalu- on the different types of radiological dental images used in the period
ated in this review (2016 to 2022), where PTVR represents the works evaluated in this review (2016 to 2022)

with the London Atlas [10] (+0.40 to +0.74 years). The dental age. In the analysed period of time, a set of 40
latter resulted in a systematic overestimation in the rest of studies using Demirjian et al.’s method [13] reported any
the populations in which it was applied. In terms of absolute of the required metrics in different populations, as shown
error, the mean values ranged from 0.43 to 1.43 years, while in Table 5. The range of ages was also wider than in the
the standard deviations ranged between 0.36 and 1.05. case of Nolla’s method [12], with subjects ranging from
Regarding the staging methods, the one proposed by two to 30, although most of them focused on the interval
Nolla [12] was used in eight different studies, as seen in between five and 23. Regarding the obtained results, a clear
Table 4, showing mean residual errors between −1.12 and overestimation can be seen, being the mean errors between
+0.54 years, and standard deviation values between 0.23 −0.58 and +2.13 years. Absolute errors indicate that the
and 3.30. The mean absolute errors ranged from 0.66 to error magnitude lies between 0.13 and 1.48 years, while the
1.10 years. Most of the studies were focused on subjects reported R 2 values were over 0.60 in any case.
between five and 15 years of age, although Berkvens et al. The modified Demirjian’s method developed by Willems
[64] conducted a study on subjects aged up to 30. et al. [14] led to numerous studies focused on testing its
The method developed by Demirjian et al. [13] is perhaps applicability in different populations. In this review, a set of
one of the most studied approaches for the estimation of 28 studies was analysed (Table 6). On average, the method

Table 2 Evaluation of the Haavikko’s method [7]

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Hedge et al. [81] Indian 1200 5–15 −0.22 0.82 – 0.71 – – – –

Kumaresan et al. [82] Malaysian 426 5–15 −1.31 1.10 – 1.45 – – – –
Benedicto et al. [83] Brazilian 1059 8–16 −1.35 1.05 – 1.42 – 1.29 – –
Sezer & Çarıkçıoğlu [84] Turkish 980 6–15 −0.49 0.85 – 0.33 0.26 – – –
International Journal of Legal Medicine (2023) 137:1117–1146 1125

Table 3 Evaluation of Atlas-based methods to estimate age in subadults. When two values are given, they correspond to the metrics obtained in
males/females

Evaluation work Population n Age Atlas E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Sezer and Çarıkçıoğlu [84]Turkish 980 6–15 London Atlas [10] +0.09 0.57 – 0.43 0.36 – – –
Baylis and Bassed [63] New Zealander 875 5–18 London Atlas [10] +0.40/+0.74 0.77/0.79 – 0.76/0.93 0.50/0.61 – – –
Schour and Massler [8] −0.39/−0.03 0.77/0.83 – 0.86/0.84 0.48/0.52 – – –
Blenkin and Taylor [11] +0.07/+0.34 0.77/0.94 – 0.77/0.89 0.47/0.55 – – –
Pavlović et al. [85] Portuguese 736 3–24 London Atlas [10] +0.09 1.45 – – – – – –
Ghafari et al. [86] Iranian 339 6–16 London Atlas [10] +0.09 – – 0.60 0.57 – – 0.92
Alsudairi and AlQahatni [87] Saudi 400 6–16 London Atlas [10] −0.59 1.46 – 1.16 1.05 – – –
Sharma and Wadhwan [88] Indian 335 5–16 London Atlas [10] +0.03 0.70 – 0.54 0.44 – – –
Gelbrich et al. [89] German 500 6–16 London Atlas [10] +0.30 – – 0.90 – – – –
da Silveira et al. [90] Brazilian 288 5–23 London Atlas [10] +0.55 – – 1.43 – – – –
Alkandiri et al. [91] Kuwaiti 375 5–15 London Atlas [10] +0.19/+0.60 – – – – – – –

was applied to a narrower age range, working most of the Table 8. The range of application, however, is narrower than
authors in the range between five and 16 years of age. in the Demirjian’s applications, as the subjects were in every
Although more investigations that show overestimation than case younger than 18. Unlike the systematic overestimation
underestimation—13 vs. 12, respectively—, this trend is of the Demirjian et al.’s method [13], Chaillet and Willems’
much less noticeable than in the case of the method by [16] tended to underestimate age, with mean errors between
Demirjian et al. [13]. The absolute errors also tended to −2.79 and −0.07 years. Absolute errors ranged from 0.66
decrease with this method, as the values lied between 0.61 to 1.14 years on average, with standard deviations between
and 1.16 years. Bedek et al. [15] proposed a modification 0.49 and 0.52 years.
of Willems et al.’s method [14], which was evaluated in an Finally, the modified Demirjian’s method developed
Indian population by Sheriff et al. [65], as it can be seen by Blenkin and Evans [17] was applied to two different
in Table 7. The results showed a notable underestimation populations of subjects aged six to 17 (Table 9), yielding
(up to −0.55 years of mean error), but the low standard errors with mean values ranging from − 0.05 to − 0.55
deviation values (0.05 to 0.06 years) indicated that the error years and standard deviations up to 1.04 years. The absolute
was consistent between all subjects. errors ranged from 0.61 to 0.91 years on average.
The modification of the Demirjian et al.’s method [13] The tooth staging criteria proposed by Gleiser and Hunt
proposed by Chaillet and Willems [16] was applied to four [18] led to the development of several methods, such as
different samples in the collected studies, as presented in those proposed by Moorrees et al. [19] and Kohler et al.

Table 4 Evaluation of the Nolla method [12]. When two values are given, they correspond to the metrics obtained on males/females

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Kumaresan et al. [82] Malaysian 426 5–15 +0.54 1.31 – 1.10 – – – –

Berkvens et al. [64] Canadian 361 8–30 – – – – – – – 0.88
Hedge et al. [92] Indian 1200 5–15 −0.13/−0.30 0.81/0.82 – – – – – –
Melo and Ata-Ali [93] Spanish 2541 7–21 −0.27/−0.16 0.50/0.23 – – – – – –
da Luz et al. [94] Brazilian 930 8–15 +0.09/+0.03 0.97/1.27 – 0.80/0.99 0.56/0.79 0.74/0.75 – –
Croatian 924 8–15 −0.13/−0.44 1.43/1.44 – 1.08/1.07 0.94/1.05 0.85/0.75 – –
AlQadi et al. [95] Yemeni 358 8–16 −0.59/−0.78 1.28/1.21 – – – – – 0.77/0.79
Yassin et al. [96] Saudi 458 5–11 −1.05/−1.12 3.07/3.30 – – – – – –
Han et al. [97] Chinese 2000 5–15 +0.18/−0.02 1.22/1.27 – 0.66/0.77 – – – –
Koç et al. [98] Turkish 916 6–14 −0.49 1.02 – 0.89 – – – –
1126 International Journal of Legal Medicine (2023) 137:1117–1146

Table 5 Evaluation of the Demirjian et al.’s method [13] and revisited Demirjian’s method [99]. When two values are given, they correspond to
the metrics obtained on males/females

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Gelbrich et al. [89] German 500 6–16 +0.11 – – 0.97 – – – –

Berkvens et al. [64] Canadian 361 8–30 – – – – – – – 0.87
Hedge et al. [92] Indian 1200 5–15 +0.19/+0.11 0.80/0.81 – – – – – –
Melo and Ata-Ali [93] Spanish 2541 7–21 +0.99/+0.72 0.39/0.56 – – – – – –
Yassin et al. [96] Saudi 627 5–12 +0.37/+0.69 1.49/1.17 – – – – – –
Amanullah et al. [100] Pakistani 300 9–26 – – – – – – – 0.64/0.60
Duangto et al. [101] Thai 1134 5–15 +0.11/+0.10 – – – – – 0.87/0.73 –
Cavrić et al. [102] Motswana 1760 6–23 +1.25/+0.72 1.11/1.02 – 1.36/0.96 0.96/0.80 – – –
Chiam et al. [103] Australian 230 2–15 −0.24/+0.04 0.84/0.82 – – – – – –
Bunyarit et al. [104] Malaysian 1236 5–16 +0.05/+0.04 0.83/0.89 – 1.24 – – – –
Alsaffar et al. [105] Maltese 200 4–26 +0.02/−0.02 – – – – – – –
Saade et al. [106] Lebanese 260 8–17 +0.74/+0.86 1.38/1.04 – – – – – –
Kelmendi et al. [107] Kosovar 1022 5–14 +0.20/+0.43 0.80/0.76 – 0.65/0.67 0.51/0.56 – – –
Almotairy et al. [108] Sweedish 107 9–11 +1.08 0.84 – – – – –
Agrawal et al. [109] British 150 8–19 +1.51/+2.13 0.96/1.35 – – – – – –
Esan et al. [110] South African 540 5–16 +0.85/+1.00 – – 1.10/1.10 – – – –
Sobieska et al. [111] Polish 1002 4–17 −0.32 – – – – – – –
Nemsi et al. [112] Tunisian 500 5–15 −0.38 0.93 – 0.77 0.64 – – –
Al-Balushi et al. [113] Omani 485 4–17 +0.10/+0.05 – – – – – – –
Kumagai et al. [114] Japanese 256 4–20 +0.09/+0.07 – – 1.07/1.00 – – – –
Liu et al. [115] Chinese 2519 8–23 −0.02 – – 0.13 – – 2.03 0.74
Khdairi et al. [116] German 1260 5–17 +0.27/+0.41 0.87/0.84 – – – – – –
Metsäniitty et al. [117] Somali 803 3–23 +0.26/+0.29 – – 1.03/1.11 – – – –
Ozveren et al.[118] Turkish 766 6–15 +1.04/+0.87 0.95/0.92 – – – – – –
Lopes et al. [119] Brazilian 403 7–13 +1.49/+1.47 – – – – – –
Birchler et al. [120] Finnish 100 6–15 +0.34 0.87 +0.32 – – – – –
Ranasinghe et al. [121] Sri Lankan 688 8–17 +0.19 0.87 – 0.70 0.55 – – –
Kermani et al. [122] Iranian 158 5–13 +0.67/+1.44 0.92/0.88 - 0.86/1.48 0.75/0.81 – – –
Ginzelova et al. [123] Czech 579 3–16 +0.06 1.60 – – – – – –
Alqadi et al. [124] Yemeni 357 8–16 −0.58 1.25 - - - - - 0.71/0.78
Moness et al. [125] Egyptian 160 3–10 +0.47/+0.33 – – – – – – –
Shen et al. [126] Taiwanese 799 8–16 +0.12/+0.21 0.80/0.95 – – – – – –
Memorando [127] Filipino 384 9–23 – – – 1.05/1.06 – – – –
Pinchi et al. [128] Italian 752 3–16 +0.41 0.88 – – – 0.67 – –
Subedi et al. [129] Nepalese 352 5–23 +0.01/+0.00 – – 1.02/1.23 0.86/1.42 – – 0.94/0.89
Galibourg et al. [130] French 3570 2–24 +0.71 0.07 – 1.11 0.05 – – 0.82
Karimi et al. [131] Kuwaiti 1393 3–26 −0.14/+0.33 1.23/0.84 – – – – – –
Rodrı́guez et al. [132] Mexican 182 6–15 +0.70/+0.51 – – 1.00/1.06 – – – –
Jayaraman et al. [133] American 600 6–17 −0.04/−0.07 0.51/0.41 – – – – – –
(Hispanic)
Shi et al. [134] Tibetan 1951 4-15 −0.46/−0.48 1.09/1.04 – 0.96/0.96 – – – –

[20]. Six studies applied the former in different samples of underestimating the age. In absolute terms, the error ranged
subjects aged from three to 30, as seen in Table 10, with between 0.63 and 1.42 years. On the other hand, Kohler
mean errors between −1.01 and +0.34 and so a tendency to et al.’s method [20] was applied to two different samples
International Journal of Legal Medicine (2023) 137:1117–1146 1127

Table 6 Evaluation of the Willems et al,’s method [14] to estimate age in subadults. When two values are given, they correspond to the metrics
obtained in males/females

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Kumaresan et al. [82] Malaysian 426 5–15 +0.54 1.28 – 0.99 – – – –

Gelbrich et al. [89] German 500 6–16 +0.38 – – 0.88 – – – –
Hegde et al. [92] Indian 1200 5–15 +0.09/+0.08 0.80/0.80 – – – – – –
Duangto et al. [101] Thai 1134 5–15 −0.37/−0.39 – – – – – 0.91/0.93 –
Saade et al. [106] Lebanese 260 8–17 +0.18/+0.08 1.24/1.12 – – – – – –
Sobieska et al. [111] Polish 1002 4–17 −0.38 – – – – – – –
Agrawal et al. [109] Nepali 150 8–19 −0.80/−1.23 1.13/1.45 – – – – – –
Kelmendi et al. [107] Kosovar 1022 5–14 −0.14/−0.24 0.77/0.75 – 0.61/0.64 0.49/0.46 – – –
Almotairy et al. [108] Sweedish 107 9–11 +0.46 0.83 – – – – – –
Nemsi et al. [112] Tunisian 500 5–15 −0.54 3.14 – 1.05 3.01 – – –
Ranasinghe et al. [121] Sri Lankan 688 8–17 −0.38 0.84 – 0.69 0.61 – – –
Shen et al. [126] Taiwanese 799 8–16 −0.22/+0.12 0.80/0.92 – – – – – –
Pinchi et al. [128] Italian 752 3–16 −0.25 1.02 – – – 0.76 – –
Galibourg et al. [130] French 3605 2–24 +0.22 0.08 – 0.93 0.04 – – 0.87
Rodrı́guez et al. [132] Mexican 182 6–15 +0.05/+0.00 – – 0.66/0.99 – – – –
Shi et al. [134] Tibetan 1951 4–15 −0.84/−1.00 1.03/1.04 – 1.06/1.16 – – – –
Lauc et al. [135] Bosnian 776 7–15 +0.57/+0.48 1.06/1.12 – – – – – –
Marinkovic et al. [136] Serbian 423 5–15 +0.63/+0.58 0.95/0.94 – – – – – –
Willems et al. [137] South African 986 4–15 −0.06 0.86 +0.00 0.69 0.52 0.57 – –
Metsäniitty et al. [138] Somali 808 4–18 −0.09 1.01 – 0.78 0.65 – – –
Kurniawan et al. [139] Indonesian 110 6–14 +0.15 0.92 – – – – – –
Ortega–Pertuz and Piña Venezuelan 458 6–18 +0.71 1.23 – – – – – –
D’Abreu [140]
Paz–Cortés et al. [77] Spanish 604 4–13 +0.26 0.91 – – – – – –
Pan et al. [141] Chinese 2367 5–16 −0.07/−0.24 0.92/1.03 – 0.70/0.79 – – – –
Çarıkçıoğlu and Turkish 1024 6–16 +0.23 0.80 – 0.66 0.51 – – –
Değirmenci [142]
Rocha et al. [143] Brazilian 1000 6–16 +0.18/−0.01 – – 0.78/0.79 – – – –
Yassin et al. [144] Saudi 1206 4–14 −0.39 1.48 – – – – – –
Cadenas et al. [145] Kenyan 1083 3–24 +0.00 1.30 +0.00 0.97 0.86 0.76 – –

of subjects up to 24 years of age, reaching systematic to an age underestimation in both cases, though it was more
underestimations in both cases and mean absolute errors up noticeable in the former, with a mean error of −0.58 years.
to 2 years. Absolute errors were very similar in both studies, with mean
The widely used tooth staging method proposed by values of 0.97 and 0.99 years and median values of 0.83 and
Liliequist and Lundberg [21] was used in two of the 0.81, respectively.
retrieved studies, in Brazilian and Croatian populations, The method proposed by De Tobel et al. [22], which
respectively. As it can be seen in Table 11, the method led mixed both Demirjian et al. [13] and Kohler et al.’s [20]

Table 7 Evaluation of Bedek et al.’s method [15] to estimate age in subadults. When two values are given, they correspond to the metrics obtained
in males/females

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Sheriff et al. [65] Indian 650 7–15 −0.55/−0.36 0.06/0.05 – – – – – –

1128 International Journal of Legal Medicine (2023) 137:1117–1146

Table 8 Evaluation of Chaillet and Willems’ method [16] to estimate age in subadults. When two values are given, they correspond to the metrics
obtained in males/females

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Kelmendi et al. [107] Kosovar 1022 5–14 −0.24/−0.35 0.85/0.74 – 0.72/0.66 0.52/0.49 – – –
Rodrı́guez et al. [132] Mexican 182 6–15 +0.07/−0.39 – – 0.78/1.14 – – – –
Hegde et al. [146] Indian 1200 5–15 −0.12 0.69 – – – – – 0.55
Bunyarit et al. [147] Malaysian 1569 5–18 −2.09/−2.79 0.90/0.99 – – – – – –

staging systems, was applied to a Belgian and Dutch sample not only depending on the measurement but also across
of subjects aged 14 to 26 as presented in Table 12. The the studies using the same measurement (from 5.66 to
authors reported mean errors closer to zero and thus an equal 25.85 years in the case of PTVR). This can also be seen
tendency to overestimating or underestimating age, while in the standard error metric, which lied between 4.66 and
the mean absolute error was between 1.70 and 2.00 years. 15.29 years. In terms of variance explained, the models
As seen in Table 13, Cameriere et al.’s method [23] moved between 1 and 97%. The specific method of Kvaal
based on the evaluation of open apices led to a systematic et al. [26], which is based on pulp-to-tooth linear ratios,
underestimation in most evaluated populations—13 out of yielded very different behaviour in the available studies.
14 reporting the residuals—, with a maximum mean error For example, while the study by Li et al. [67] reported
of −1.36 years, although it is noticeable that in the Rivera a noticeable overestimation, the work by Roh et al. [68]
et al. study [66], the mean error is negative while the median pointed out a great underestimation, both of them evaluated
error of male subjects is positive. The absolute errors ranged in samples of adults aged around 20 to 70. In the same way,
between 0.57 and 1.60 years on average. studies published on the evaluation of the Cameriere et al.’s
The proposed approaches for estimating chronological method [30] based on the pulp-to-tooth area ratio focused
age in adults produced systematically worse results than on the statistical analysis of the method and reported limited
their children-orientated counterparts, and the available information regarding the residual error of the model.
studies are much scarcer. Moreover, the studies that apply Only Dabbaghi and Kazemi [69] reported the mean error,
adult-based methods tended to report mostly the standard pointing out an overestimation of 8.83 years. Furthermore,
error and R 2 values instead of the residual and absolute error it is noticeable that the lowest absolute errors were obtained
measurements, in opposition to the previously presented in a multi-ethnic sample [70].
methods. In this regard, 35 studies were collected related to Methods that have associated radiographic visibility of
the evaluation of metric methods based on a set of linear several oral structures with chronological age usually do not
and volumetric tooth analysis, as seen in Table 14. The aim to estimate a numerical age value, so only two of them
most common approach is the pulp-to-tooth linear, area, reported estimation error metrics. As shown in Table 15,
and volumetric ratio (PTLR, PTAR, PRVR), used in eight the study of Chaudhary and Liversidge [71] pointed out an
works each, and the tooth-to-crown index (TCI) and the overall overestimate of 7.21 years for males and 6.87 years
pulp-to-crown volume ratio (PCVR), each one applied in for females, being the mean absolute error of 7.91 and 7.74
three studies. It is also worth noting that most of these works years in the same two scenarios. On the other hand, Timme
relied on 3D images (such as CT-based records) instead of et al. [72] did not report the error metrics, but a standard
flat X-rays, as they allow the volume of the different tooth error of 3.55 years and the percentage of explained variance
structures to be analysed accurately. (69%).
Huge variability in the results reported by these studies Methods based on the evaluation of degenerative tooth
is observed. For example, the mean absolute error varied changes, based on Gustafson’s criteria [35], are summarised

Table 9 Evaluation of Blenkin and Evans’ method [17] to estimate age in subadults

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Ranasinghe et al. [121] Sri Lankan 688 8–17 −0.55 1.04 – 0.91 0.75 – – –
Çarıkçıoğlu and Değirmenci [142] Turkish 1024 6–16 −0.05 0.77 – 0.61 0.47 – – –
International Journal of Legal Medicine (2023) 137:1117–1146 1129

Table 10 Evaluation of the methods based on Gleiser and Gunt criteria [18]. When two values are given, they correspond to the metrics obtained
on males/females

Evaluation work Population n Age Estimation method E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Alkandiri et al. [91] Kuwaiti 375 5–15 Moorrees et al. [19] −0.89/−1.01 – – – – – – –
Berkvens et al. [64] Canadian 361 8–30 Moorrees et al. [19] – – – – – – – 0.09
Metsäniitty et al. [117] Somali 803 3–23 Kohler et al. [20] −0.29/−0.30 – – 1.39/1.30 – – – –
Rodrı́guez et al. [132] Mexican 182 6–15 Moorrees et al. [19] −0.03/+0.34 – – 0.63/0.98 – – – –
Jayaraman et al. [133] Hispanic American 600 6–17 Moorrees et al. [19] −0.07 0.45 – – – – – –
Fulton and Liversidge [148] British 940 3–16 Moorrees et al. [19] −0.22 2.08 – 1.42 – – – –
Štepanovskỳ et al. [149] Czech 976 3–20 Moorrees et al. [19] – – – 0.64 – – – –
Sartori et al. [150] Brazilian 1062 15—24 Kohler et al. [20] −1.30 2.10 – 2.00 1.50 – – –

in Table 16. The results show again a high degree of same number of studies underestimating and overestimating
variability, with standard errors reaching a minimum of 0.69 age.
and a maximum of 10.92 years, even if in both cases the As mentioned in the previous section, some age
Gustafson criteria are assessed in a population of the same estimation methods were adapted to work as a binary
ethnic origin. Only two studies reported the mean absolute classifier for detecting people younger or older than the
error, with very different values in each one (3.34 and 3.68 legal age. The results obtained in this regard are presented
years in the first method and 11.08 years in the second). in Table 18. First, the methods based on tooth eruption
Finally, R 2 values were in the range 0.23–0.80. presented by Haavikko [7] and Olze et al. [40] were assessed
As mentioned in “Age estimation on other radiologically in the problem of 14-year-old detection. The former led
observable structures” section, the estimation of chronolog- to accuracy between 78 and 81%, while the latter yielded
ical age was also approached by mandibular bone analysis, better performance, with 83 to 86%. Also, the method
specifically by measuring ramus length [38] and gonial proposed by Olze et al. showed a more balanced behaviour,
angle [39]. The former produced a model that represented with similar sensitivity and specificity values.
62% of the data variance, while the latter led to an abso- There is only one study that evaluated an Atlas-based
lute error of 13.98 years. As it can be seen in Table 17, both method for binary age classification. Specifically, De
studies reported different metrics, so they are not directly Moraes et al. [73] used the London Atlas [10] for classifying
comparable. dental records according to the 18-year-old threshold.
The most widely used methods for numeric age Although the accuracy reached a reasonable value of 80%,
estimation were jointly analysed regarding the obtained the methods were heavily biased, as they produced a very
underestimation or overestimation. As it can be seen in high sensitivity—that is, they correctly detected subjects
Fig. 4, two of the six methods showed a clear pattern of older than 18— but very low specificity —it only correctly
overestimation, namely those proposed by Demirjian et al. classified half of the subjects younger than 18.
[13] and the London Atlas [10]. On the other hand, the Dental staging methods were used to a greater extent.
methods developed by Cameriere et al. [23] and Nolla [12] Regarding the Nolla method [12], it was applied to the
led to a systematic underestimation of age. Finally, the Portuguese and Montenegrin populations. In the first, the
methods based on linear and volumetric measurements of method obtained accuracies from 82 to 90%, depending
the teeth, as well as that proposed by Willems and Chaillet on the age threshold, while in the latter the accuracy was
[14] yielded a more balanced performance, with almost the 90% for males and 87% for females. It is noticeable that

Table 11 Evaluation of Liliequist and Lundberg’s method [21]

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Benedicto et al. [83] Brazilian 1059 8–16 −0.58 – – 0.97 – 0.83 – –

Da Luz et al. [94] Croatian 924 8–15 −0.05 1.27 – 0.99 0.81 0.81 – –
1130 International Journal of Legal Medicine (2023) 137:1117–1146

Table 12 Evaluation of De Tobel et al.’s method [22] to estimate age in subadults and young adults

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

De Tobel et al. [151] Belgian and Dutch 309 14–26 +0.00/+0.10 – – 1.70/2.00 – – – –

Nolla’s method was highly biased in both directions in methods of Moorrees et al. [19] and Kohler et al. [20].
the Portuguese population, as it obtained higher sensitivity The former was applied in a Portuguese sample with 14,
values with the thresholds of 14 and 21 years, and better 16, 18, and 21 thresholds, obtaining accuracies from 83
specificity values with the age thresholds of 16 and 18. to 90%. Again, the sensitivity and specificity values were
The Mincer et al.’s method [42] based on the devel- highly unbalanced, especially when using the 14-year-old
opment stages proposed by Demirjian et al. [13] led to threshold (92% of sensitivity and 59% of specificity). On
maximum accuracies of 91, 90, and 94 % when using the other hand, the Kohler et al.’s method reached an
the legal age thresholds of 12, 14, and 18 years, respec- accuracy of 91% in an Indian population and sensitivity and
tively. Although the sensitivity and specificity were almost specificity values of 87–80% and 87–90%, respectively, in
balanced in most studies, there are some cases where a a Russian sample.
significant bias was observed, such as the study by Mwe- The adaptation of Cameriere’s method for legal age
sigwa et al. [74] with a legal age of 16—sensitivity of 69% classification [62] was by far the most used method for
vs. specificity of 97%. Pinchi et al. [75] also tested the binary age classification. Among all the experiments carried
Willems et al.’s scores [14] in an Italian population, reaching out with this approach, 29 out of the 40 established an
a slightly worse result than in the case of the original scores, age threshold of 18 years. The accuracy values ranged
especially in the sensitivity values (74–78% vs. 80–82%). from 72 to 98%, although 35 studies yielded values greater
Gleiser and Hunt staging system [18] was also studied than 80%. As with most methods, there are some cases
on the problem of binary age classification via the derived where a great bias between sensitivity and specificity can

Table 13 Evaluation of Cameriere’s method [23] for age estimation in subadults. When two values are given, they correspond to the metrics
obtained on males/females

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Kumaresan et al. [82] Malaysian 426 5–15 −0.41 1.08 – 0.89 – – – –

Sharma and Indian 335 5–16 −0.60 1.32 – 1.11 0.93 – – –
Wadhwan [88]
da Luz et al. [94] Brazilian 930 8–15 −1.08/−1.03 1.13/1.14 – 1.27/1.26 0.91/0.88 1.06/1.13 – –
Croatian 924 8–15 −1.20/−1.19 1.25/1.36 – 1.38/1.39 1.05/1.16 1.16/1.11 – –
Marinkovic et al. [136] Serbian 423 5–15 −0.38/−0.38 0.93/0.92 – – – – – –
Çarıkçıoğlu and Turkish 1024 6–16 −0.51 0.90 – 0.77 0.68 – – –
Değirmenci [142]
Rivera et al. [66] Colombian 526 6–14 −0.08 0.69 +0.15/−0.28 0.57/0.57 0.38/0.41 – – –
Cameriere et al. [152] Italian 2630 4–17 – – – 0.72/0.73 0.60/0.61 – – –
Santana et al. [153] American - Indian 57 6–17 −1.36 1.47 −1.02 1.55 – 1.05 – –
American - European 173 6–17 −1.24 1.72 −1.02 1.60 – 1.21 – –
American - Hispanic 130 6–17 −1.24 1.52 −0.96 1.48 – 1.15 – –
Halilah et al. [154] German 1000 5–16 −0.64/−0.38 0.91/0.88 −0.36/−0.34 0.92/0.76 – 0.79/0.64 – 0.85/0.83
Angelakopoulos et al. South African - Black 970 6–14 – – – 0.70 – – – 0.82
[155] South African - White 974 6–14 – – – 0.58 – – – 0.874
AlShahrani et al. [156] Saudi 788 6–16 −0.26 1.47 – – – – – 0.49
Różyło–Kalinowska Polish 121 5–13 +0.17/+0.18 0.86/0.96 – 0.73/0.77 0.48/0.60 – – –
et al. [157]
Shen et al. [158] Chinese 748 5–13 −0.39 0.04 – 0.81 0.02 – – –
Table 14 Evaluation of methods based on evaluation of linear and/or volumetric dental PCLR, pulp-to-crown linear ratio; PCVR, pulp-to-crown volume ratio; PEVR, pulp-to-enamel
measurements. When two values are given, they correspond to the metrics obtained in volume ratio; PDVR, pulp-to-dentin volume ratio
males/females. TCI, tooth coronal indexes; PLR, pulp linear ratio; PTLR, pulp-to-tooth lin-
ear ratio; PTAR, pulp-to-tooth area ratio; PTVR, pulp-to-tooth volume ratio; PV, pulp volume;

Evaluation work Measurement Population n Age E (DA-CA) AE(|DA-CA|) SE R2

μ σ med. μ σ med.

Li et al. [67] PTLR (Kvaal et al.) Chinese 360 20–65 +8.6 11.80 – – – – 11.40 0.23
Roh et al. [68] PTLR (Kvaal et al.) South Korean 266 21–69 −9.01 11.58 – 12.58 7.54 – 10.70 0.47
Dabbaghi and PTAR (Cameriere et al.) Iranian 153 13-70 +8.83 0.52 – – – – 11.06 0.54
Kazemi [69]
Cameriere et al. [70] PTAR (Cameriere et al.) Turkish, Italian, Portuguese, Japanese and Mexican 891 20–86 – – – 2.43/2.49 – – – –
Jain et al. [159] TCI, PCLR Indian 180 15–70 +1.34 – – – – – – 0.83
Akay et al. [160] TCI British 250 18–60 – – – – – – 10.37 0.28
Gok et al. [161] TCI Turkish 9059 15–40 – – – – – – 7.25/7.02 0.08/0.06
International Journal of Legal Medicine (2023) 137:1117–1146

Herianti et al. [162] PLR Indonesian 113 11–60 – – – – – – 4.66 0.88

Hisham et al. [163] PTLR (Kvaal et al.) Malaysian 718 16–80 – – – – – – 15.29 0.06
Akay et al. [164] PTLR (Kvaal et al.) Turkish 134 16-71 – – – – – – – 0.39
PTVR – – – – – – – 0.52
de Miranda et al. [165] PTLR (Kvaal et al.) Brazilian 320 20-59 – – – 6.81 4.06 – – –
PTAR (Cameriere et al.) – – – 7.55 5.35 – – –
Limdiwala et al. [166] PTLR (Kvaal et al.) Indian 160 18–62 – – – – – – 12.47 0.01
Vossoughi et al. [167] PTLR (Kvaal et al.) Iranian 240 20–85 – – – 7.53 5.34 – 9.22 0.68
Farhadian et al. [168] PTLR, PTAR Iranian 300 14–60 – – – 4.12 – – – 0.86
Jambunath et al. [169] PTAR (Cameriere et al.) Indian 250 18–75 – – – 7.50 – – – 0.48
Lee et al. [170] PTAR (Cameriere et al.) Korean 402 20–78 – – – – – – 10.40 0.62
Kumar et al. [171] PTAR Indian 400 14–60 – – – – – – 12.00 0.17
Dehghani et al. [172] PTAR Iranian 271 16–64 – – – 6.07 1.70 – – –
Lee et al. [173] PTVR South Korean 205 20–77 – – – – – – – 0.52
Haghanifar et al. [174] PTVR Iranian 377 20–69 – – – – – – 7.21/7.68 0.55/0.50
Asif et al. [175] PTVR Malaysian 300 16–65 – – – 5.66 – – 5.84 –
Zhang et al. [176] PTVR Chinese 392 16–76 – – – 6.80/7.90 – – – 0.67/0.63
Muralidhar et al. [177] PTVR Indian 100 19–70 – – – – – – 7.21/7.68 0.55/0.50
Pires et al. [78] PTVR Portuguese 158 21–80 −21.36 16.51 – 25.85 – – – –
Kazmi et al. [178] PTVR Pakistani 719 15–65 – – – – – – – 0.46
Helmy et al. [179] PV Egyptian 187 21–50 – – – 4.61 – – 5.87 0.48
Ge et al. [180] PV Chinese 250 16–63 – – – – – 6.26 – 0.70
1131
1132 International Journal of Legal Medicine (2023) 137:1117–1146

0.97
0.95
0.37
be seen, the most significant example being the study by

0.78
0.42
R2

–
AlQahtani et al. [76], where the sensitivity was 51–52% and
the specificity was 100–97%.
Finally, Olze et al.’s [31] method based on the assessment

6.83

7.06
SE
of the root pulp visibility was evaluated in Turkish

–
–
–

–
and Indian samples. Only the latter study reported the
accuracy—77% in males and 80% in females—yielding
also a notable imbalance between sensitivity and specificity.
med.

– The two most widely applied methods for age thresh-

–
–
–
–
–
olding, namely those proposed by Mincer et al. [42] and
5.7 Cameriere et al. [62], were compared using a reference
value of 90% accuracy. As it is shown in Fig. 5, the method
AE(|DA-CA|)

–
–

–
–
–
of Mincer et al. obtained a performance better than the
reference when establishing an age threshold of 12 years
8.00

8.41

and three times out of four with a threshold of 18 years.

μ

–
–

When the threshold is set to 14 years of age, one study

obtained better performance than the reference, and another
work reported worse performance. Regarding the Cameriere
med.

et al’s method, all studies that set an age threshold of 12,

–
–
–
–
–
–

14, or 16 years reported accuracy values lower than the

reference, while studies applying a threshold of 18 years
σ

–
–
–
–
–
–

showed a more balanced performance, with 12 studies per-

forming better than the reference and 16 works reporting
worse results.
−0.29/−0.28
E (DA-CA)

Regarding the automatic approaches proposed for age

estimation, each method was tested in a single population. In
−0.11

those aimed at estimating a numerical age value (Table 19),

μ

–
–
–
–

the residual error was systematically closer to zero, being

the median between −0.07 and +0.12 years. Absolute error
varied depending on the age range of the assessed sample.
15–54
15–60
14–70
16–65
20–64
22–67
Age

As reported by Vila-Blanco et al. [49], the mean absolute

error in a Spanish sample ranged from 0.75 years in subjects
younger than 15 years to 2.84 years in subjects younger
than 90 years. In this case, the median absolute error was
416
200
107
110
414
120
n

as low as 0.64 years. Comparatively, the work by Čular

et al. [46] reported a mean absolute error of 2.28 years in
Croatian subjects between 10 and 25 years of age, being
Population

Malaysian
Egyptian

1.75 years on a German sample of subjects aged 5 to

Tunisian
Chinese
Spanish
Indian

25, as noticed by De Back et al. [48]. Performance was

significantly worse in the methods tested in older subjects,
as reported by Milošević et al. [55], Zheng et al. [53],
and Pham et al. [51], with mean absolute errors of 3.96,
Measurement

7.17, and 6.97–7.07 years, respectively. On the contrary,

the method of Hou et al. [50], which was tested in a large
PDVR
PCVR
PCVR
PCVR
PEVR
PCLR

population sample and very wide in terms of subject age

(from 0 to 93 years of age), led to a very low mean absolute
error, being of 1.64 years. It is also noticeable that the
(continued)

method of Vila-Blanco et al. [54], which relies only on the

Elgazzar et al. [182]
Molina et al. [183]

mandible shape instead of the whole dental image, yielded

Nemsi et al. [186]
Zhang et al. [185]
Singal et al. [181]
Evaluation work

Asif et al. [184]

a mean absolute error of 1.57 years, which is comparable or

even better than other methods relying on the whole dental
Table 14

image.
International Journal of Legal Medicine (2023) 137:1117–1146 1133

Table 15 Evaluation of methods to estimate the age of adults based on the radiographic visibility of several structures. When two values are given,
they correspond to the metrics obtained in males/females. PL, periodontal ligament; RP, root pulp

Evaluation work Observed structure Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Chaudhary and Liversidge [71] PL British 163 16–45 +7.21/+6.87 5.16/5.83 – 7.91/7.74 – – – –
Timme et al. [72] PL and RP German 1245 15–40 – – – – – – 3.55 0.69

Automatic methods focusing on classifying the subject’s methods developed by AlQahtani et al. [10], Nolla [12],
age into a defined set of age ranges led to different Willems and Chaillet [14], and Cameriere et al. [23] focused
results depending on the number of classes, as shown in on a very constrained group of patients aged two to 18,
Table 20. For example, the approaches of De Tobel et al. approximately, Demirjian et al.’s method [13] could be
[56], Merdietio et al. [57], and Banar et al. [58], which applied to a wider group of patients of even more than 25
proposed to use 10 different development stages, obtained years of age. The methods focused on post-developmental
classification accuracies between 51 and 60% and Kappa dental features, such as the one proposed by Kvaal et al.
values between 0.82 and 0.84. It should be noted that these [26] or the pulp-to-tooth volumetric ratios, have as their
three methods were assessed in the same Belgian sample. natural field of application those subjects aged 18 to 70,
On the other hand, Kim et al. [59] and Kahaki et al. approximately. On the other hand, the automatic methods
[60] reduced the classification problem to five age groups, have been proven to be applicable in a wider age range,
leading to significantly higher accuracy values of 82 and covering both the subjects with developing dentitions and
90%, respectively. the subjects with fully developed teeth.
Finally, Štern et al. [47] and Guo et al. [61] tested
their automatic age thresholding approach in Austrian and
Chinese populations, respectively, as it can be seen in Discussion
Table 21. It is worth noting that the latter sample consisted
of more than 10,000 OPGs. Guo’s method led to accuracy The oral cavity, and especially the teeth, has been used for
values between 93 and 96%, depending on the specific age decades as they show a high correlation with development
threshold, and was very consistent in terms of sensitivity patterns. In this regard, great efforts have been made
and specificity. On the other hand, the method by Štern et al. since the late nineteenth century to develop teeth evolution
reached a slightly worse accuracy value in the same scenario Atlases, with a view not only to the formation of teeth
(85% vs. 93%), and a notable imbalance between sensitivity concerning age but also to the sexual dimorphic patterns of
and specificity is observed. that development. The democratisation of radiology led to a
The applicability of the methods used in the studies number of improvements, one of them being the collection
included in this work was assessed in terms of the age of bigger databases to use in new studies, which in the end
of the subjects. As Fig. 6 indicates, there are notable increases the statistical significance of the findings. Another
differences among the proposed approaches. While the major revolution came with the arrival of computers, which

Table 16 Evaluation of methods for estimating age in adults based on Gustafson’s criteria [35]. When two values are given, they correspond to
the metrics obtained on males/females

Evaluation work Population n Age Estimation method E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Timme [37] German 2346 15–70 Olze [36] – – – – – – 6.75 0.80

Akay [160] British 250 18–60 Olze [36] – – – – – – 5.72 0.79
Sonjaya [187] Malaysian 400 18–74 Gustafson [35] −0.17 0.98 – – – – 0.69 0.60
Si [188] Chinese 1300 15–40 Olze [36] −0.47/−0.76 – – 3.43/3.68 – – 4.75 0.68
Dezem [189] Brazilian 503 15–70 Olze [36] – – – 11.08 – – – 0.23
Timme [37] – – – 11.08 – – – 0.23
Koh [190] Malaysian 284 20–70 Olze [36] – – – – – – 10.92 0.39
1134 International Journal of Legal Medicine (2023) 137:1117–1146

Table 17 Evaluation of methods to estimate the age based on the measurement of the mandible

Evaluation work Measurement Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Motawei et al. [38] Ramus length Egyptian 213 7–58 – – – – – – – 0.62

Acharya [39] Gonial angle Indian 100 18–89 – – – 13.98 – – 15.84 –

brought the possibility of acquiring X-ray images directly the age in both cases was through some modification of a
in a digital format and, therefore, to speed up the measuring numeric age estimation method [42–44].
processes. As tooth and bone development is reported to depend
With the aim of using the structures that are observable on factors such as ethnicity or environment, these methods
in the dental images to estimate the chronological age, have been evaluated to a greater or lesser extent in different
different approaches have been followed. Only three studies populations around the world. To assess the differences
collected in this work focused on the mandible bone [38, 39, between the different approaches and the population
54], while the rest relied on tooth analysis methods. Among samples used, this work retrieved a corpus of studies
the latter, two groups of works can be noticed, namely those applying age estimation methods in dental images published
aimed at estimating age in children and young adults [12, in the last 7 years (from 2016 to 2022). A total of 613 unique
13, 23] and those focused on adults [26, 31, 36]. Moreover, studies were obtained, of which 286 were selected after
some authors approached the age estimation problem as a applying the inclusion criteria. The first point to highlight
classification task instead of a regression problem. In this is the great difference in the number of studies that apply
regard, some studies tried to classify the age of the subjects each method. For example, methods such as those proposed
into two groups, usually with the main purpose of estimating by Schour and Massler [8] or Blenkin and Taylor [11]
the chances a person has to be older than the legal age. Other were applied in one single work in the evaluated period,
works generalised this idea and applied an age classifier which decreases the significance of the results to a large
with more than two target classes. The usual way to classify extent.

Fig. 4 Underestimation or
overestimation produced by the 100 Tendency
8
Overestimation →

methods for numeric age 28 Overestimation

estimation. Only those methods Underestimation
reporting the mean error in at
least five studies were included.
13
The methods based on linear 50 3
and volumetric measurements of
the teeth (L/V measurements in 2
the figure) were grouped for a
(% of works)

better representativity 1
0
1
6
← Underestimation

12
50 4

13
100
]

]
23

ts
13

4]
0]
en

[1
[1
.[

.[

2]
m
al

al

l.
s

[1

ta
la
re
et

et

At

lla
su

se
re

an

No
on
ea
ie

m
irji

nd
er

ille
m
m

Lo
V

W
De
Ca

L/

Method
Table 18 Evaluation of age thresholding methods. When two values are given, they correspond to males/females

Method Method work Evaluation work Population n Age Target Accuracy Sensitivity Specificity

Tooth eruption Haavikko [7] Pinchi et al. [75] Italian 501 11–15 14 78/81 88/72 70/86
Olze et al. [40] Thomas et al. [191] Indian 640 10–18 14 86/83 89/89 84/78
Atlas-based London Atlas [10] De Moraes et al. [73] Brazilian 1200 16–21 18 80 92 56
Nolla development stages Nolla [12] Pereira et al. [45] Portuguese 348 12–23 14 82 91 60
16 87 80 92
18 90 81 95
21 84 94 82
Antunovic et al. [192] Montenegrin 683 13–24 18 90/87 95/85 84/90
Demirjian development stages Willems et al. [14] Pinchi et al. [75] Italian 501 11–15 14 77/83 74/78 79/86
Mincer et al. [42] Mwesigwa et al. [74] Ugandan 1021 10–22 12 – 93 80
14 – 97 83
International Journal of Legal Medicine (2023) 137:1117–1146

16 – 69 97
18 – 88 88
Pinchi et al. [75] Italian 501 11–15 14 78/83 80/82 76/84
Lucas et al. [193] British 2000 16–26 18 93/91 92/91 98/92
Friedrich et al. [194] German 1804 15–24 18 – 74 73
Márquez–Ruiz [195] Mexican 135 14–23 18 – 85 71
Lizarbe et al. [196] Peruvian 208 14–22 18 86 88 85
Hassan et al. [197] Egyptian 350 14–24 18 – – 92/91
Sharma et al. [198] Indian 1062 14–23 18 93 93 92
Augusto et al. [79] Portuguese 1104 7–23 12 91 95 81
14 90 77 93
Melo et al. [199] Spanish 1386 10–26 18 94 94 95
Gleiser and Hunt staging system Moorrees et al. [19] Pereira et al. [45] Portuguese 348 12–23 14 83 92 59
16 87 82 91
18 90 80 95
21 84 92 82
Kohler et al. [20] Yellapurkar et al. [44] Indian 404 7–20 14 91 – –
Franco et al. [200] Russian 918 8–23 14 – 87/80 87/90
Root pulp visibility Olze et al. [31] Akkaya et al. [201] Turkish 463 14–34 18 – 89/83 91/67
21 – 86/73 88/93
Manthapuri et al. [202] Indian 760 12–20 16 77/80 61/65 96/97
1135
 1136

Table 18 (continued)

Method Method work Evaluation work Population n Age Target Accuracy Sensitivity Specificity

Cameriere’s method Cameriere et al. [23] Pinchi et al. [75] Italian 501 11–15 14 85/87 77/85 92/88
Antunovic et al. [192] Montenegrin 683 13–24 18 93/89 92/82 94/96
Lizarbe et al. [196] Peruvian 208 14–22 18 96 96 96
Hassan et al. [197] Egyptian 350 14–24 0 18 89 92 84
Augusto et al. [79] Portuguese 1104 7–23 12 89 94 74
14 87 87 88
Melo et al. [199] Spanish 1386 10–26 18 96 90 99
De Luca et al. [203] Colombian 288 13–22 18 91/94 92/95 91/94
Cavrić et al. [204] Motswana 1294 13–23 18 91/92 88/88 95/96
Zelic et al. [205] Serbian 598 13–24 18 95/91 96/86 94/98
Gulsahi et al. [206] Turkish 293 14–22 18 98/93 95/86 100/100
Franklin et al. [207] Australian 143 14–22 18 87/88 90/90 85/88
Dardouri et al. [208] Lybian 307 14–22 18 95/94 91/91 100/100
AlQahtani et al. [76] Saudi 300 14–22 18 76/72 51/52 100/97
Balla et al. [209] Indian 1330 15–22 18 90/88 91/86 90/90
Cameriere et al. [210] Chilean 822 11–22 14 86/80 84/81 90/80
16 80 79 81
18 83 71 88
Angelakopoulos et al. [211] South African 833 14–24 18 90 80 95
Różyło–Kalinowska et al. [212] Polish 982 15–24 18 87 85 92
Kelmendi et al. [213] Kosovar 1221 12–23 18 97/91 96/83 98/99
Spinas et al. [214] Italian 336 15–23 18 86 82 95
Doğru et al. [215] Dutch 360 14–22 18 89/83 84/73 95/96
Balla et al. [216] Indian 819 11–20 16 88/89 91/90 86/87
Kumagai et al. [217] Japanese 276 14–24 18 91/87 89/84 96/93
Chu et al. [218] Chinese 840 13–24 18 92 93/94 81/97
Moukarzel et al. [219] Lebanese 620 14–23 18 74/79 63/61 89/97
Ribier et al. [220] French 431 14–22 18 91/81 95/83 85/79
Scendoni et al. [221] Russian 571 14–24 18 97/95 96/93 98/98
Wang et al. [222] Chinese 671 10–20 16 89 78 97
Angelakopoulos et al. [223] African 1631 13–24 18 83 87 84
American 861 86 80 91
Asian 3919 73 61 91
European 3770 85 80 95
Yılancı et al. [224] Turkish 763 10–17 12 – 90/91 78/77
15 – 93/89 87/78
International Journal of Legal Medicine (2023) 137:1117–1146
International Journal of Legal Medicine (2023) 137:1117–1146 1137

Fig. 5 Performance of the

1

More accurate →
methods aimed at age 100
Tendency
thresholding with respect to an More accurate
accuracy reference value of 3
Less accurate
90%. Only those methods used
in at least five studies were 1
50 12
included

(% of works)
0 0 0 0 0
0

1
← Less accurate
1
50 16

1 3 3
100

Method: Mincer et al. [42] Method: Cameriere et al. [62]

12 14 16 18 12 14 16 18
Age thresold

Table 19 Results of the main automatic methods for numeric age estimation, evaluated in different populations. When two values are given, they
correspond to the metrics obtained in males/females

Evaluation work Population n Age E (DA-CA) AE (|DA-CA|) SE R2

μ σ med. μ σ med.

Čular et al. [46] Croatian 203 10–25 – – – 2.28 2.17 – – –

Štern et al. [47] Austrian 322 13–25 – – – 1.42 1.14 – – –
De Back et al. [48] Germa >12,000 5–25 – – – 1.75 – – – –
Vila–Blanco et al. [49] Spanish 2289 4.5–15 – – −0.07 0.75 0.57 0.64 – 0.84
4.5–20 – – +0.02 0.89 0.77 0.69 – 0.89
4.5–25 – – +0.07 1.17 1.11 0.85 – 0.90
4.5–30 – – +0.06 1.43 1.44 0.96 – 0.89
4.5–40 – – +0.04 1.80 2.12 1.10 – 0.87
4.5–89.2 – – +0.12 2.84 3.75 1.48 – 0.90
Hou et al. [50] Chinese 27,957 0–93 – – – 1.64 – – – –
Pham et al. [51] South Korean 814 20–70 – – – 6.97/7.07 –/0.51 – – –
Wallraff et al. [52] German 14,019 11–20 −0.30 1.41 – 1.08 – – – –
Zheng et al. [53] Chinese 180 10–60 – – – 7.17 – – – –
Vila–Blanco et al. [54] Spanish 260 5–17 – – – 1.57 1.21 – 2.0 0.80
Milošević et al. [55] Croatian 4,035 19–90 – – – 3.96 – 2.95 – –

Table 20 Results of the main automatic methods for age group classification, evaluated in different populations

Evaluation work Population n Age #Groups Accuracy Others

De Tobel et al. [56] Belgian 400 – 10 51 Kappa=82%

Merdietio et al. [57] Belgian 400 – 10 61 Kappa=84%
Banar et al. [58] Belgian 400 7–24 10 60 Kappa=82%
Kim et al. [59] South Korean 1586 – 5 90
Kahaki et al. [60] Malaysian 456 1–17 5 82
1138 International Journal of Legal Medicine (2023) 137:1117–1146

Table 21 Results of the main automatic methods for age thresholding, evaluated in different populations. When two values are given, they
correspond to the metrics obtained in males/females

Evaluation work Population n Age Threshold Accuracy Sensitivity Specificity

Štern et al. [47] Austrian 322 13–25 18 85 75 93

Guo et al. [61] Chinese 10,257 5–24 14 96 95 96
16 96 96 96
18 93 91 94

Regarding the performance obtained by the analysed multiple dental staging methods, showed a behaviour with
methods, it has been proved that the eruption-based methods no tendency to underestimation or overestimation, but
are useful for a very short period (until 15 years of greater absolute errors than those obtained with the methods
age, approximately), and the only method included in this it relies on individually. On the other hand, Cameriere’s
work, proposed by Haavikko et al. [7], led to a systematic method [23] based on the measurement of open apices
underestimation of the chronological age in every tested systematically underestimated the age, while the absolute
population. Similarly, Atlas-based methods assessed have errors were highly inconsistent compared to the previously
also been applied to very constrained population samples in mentioned methodologies.
terms of age, being the London Atlas method [10] the only Estimating chronological age in adults has proven to be
method evaluated in multiple populations. a much more difficult task, as confirmed in the studies
Dental staging approaches have proven to be highly retrieved in this work. None of the studies reported
relevant in this period, as they accounted for nine of the ten absolute errors lower than 2 years, and trends of under-
most commonly used methods in this work. This allowed or overestimation were also more pronounced [69, 78].
us to compare these methods to each other accurately and Regarding the methods using linear or volumetric dental
confirm the findings reported in previous works, such as the measurements, there is a clear improvement in terms
tendency of Nolla [12] and Demirjian et al.’s [13] methods of standard error values when volumetric information is
to underestimate and overestimate the age, respectively taken into account. It is also noticeable that adult-oriented
[77], or the balanced behaviour yielded by the Willems methods do not report as many performance metrics as
modification of Demirjian’s method [14]. The methods those targeted at children, which hinders the assessment
based on Gleiser and Hunt stages [18] achieved slightly of their performance. The most problematic cases are
worse results than Nolla’s [12] or Demirjian et al.’s [13], those involving the radiographic visibility methods and the
while both Moorrees et al.’s [19] and Kohler et al.’s mandible-based methods, where a direct comparison is not
[20] approaches showed a significant underestimation and possible.
greater absolute errors. It should also be noted that the The methods proposed for detecting if someone is
method proposed by De Tobel et al. [22], which combines younger or older than a predefined threshold (usually

Fig. 6 Application of the most

100
widely used methods for age Method
estimation regarding the age of
Demirjian et al. [13]
the subjects included in the
analysed studies. Only those Willems et al. [14]
methods applied in at least ten 75 Cameriere et al. [23]
studies were included
Number of works

Automatic
Cameriere et al. [62]
50 London Atlas [10]

PTVR
Nolla [12]
Kvaal et al. [26]
25

0 25 50 75 100
Age
International Journal of Legal Medicine (2023) 137:1117–1146 1139

the legal age) provided a more suitable environment for teeth to be present, these methods can work even if some
comparability purposes, as most of them reported the pieces are missing.
accuracy, sensitivity, and specificity metrics. The majority Although these automatic methods have been shown to
of studies yielded reasonably high accuracy, with values improve the performance and applicability of age estimation
over 80 or even 90%, suggesting that these methods could methods, their validation needs to be improved. The
be used either alone or combined with others to produce relative recency of deep learning techniques causes that
confident estimations. However, it is worth noting that the no automatic method has been tested in populations or
results are very dependent on the age cohort used to test the acquisition devices other than the original ones, which raises
methodology. If subjects are not well-balanced around the doubts about their generalisation to different scenarios.
threshold, that is, if most subjects are younger or older than However, the ease to compile new databases in this regard
the legal age, the performance will be abnormally good, as allows these methods to be easily adapted to different
the classifier will tend to assign the majority class to all situations through the application of specific domain-
subjects [79]. A similar problem can occur if the testing adaptation techniques, such as transfer learning or fine-
age cohort is too wide, as the further away the subjects are tuning [80].
from the age threshold, the easier it is to classify them and,
therefore, the better the classification performance.
The inclusion of modern image processing techniques for Conclusions
age estimation led to a set of improvements, being the first
one the time and cost saving by means of full or almost-full In this work, the current application of age estimation
automation. Moreover, the performance achieved by these methods in recent years has been studied, specifically
methods is clearly higher than that obtained with the manual those methods using radiological dental images. Although
methodologies. Although in some cases the errors were high classic methods have been thoroughly evaluated in many
due to the exclusion of children from the population sample populations of different ethnicities all over the world,
[51, 53, 55], the overall performance was remarkable, with automatic methods based on deep learning techniques led to
absolute errors lower than one year for patients younger than an improvement not only in terms of performance but also
20 [49] or 1.64 years in a huge sample of almost 28,000 regarding the applicability in a real scenario. This represents
images from patients aged zero to 93 years [50]. In this a turning point in the field of chronological age estimation,
regard, the applicability of these methods is better compared since the speed at which the estimations can be applied
to that of the classic manual approaches, with an almost can be significantly higher and the subjectivity inherent to
avoidance of underestimation and overestimation problems the observer analysis can be completely avoided. Future
and the subjectivity intrinsic to human-guided processes, as work in this area should involve a deeper assessment of
well as the possibility to run estimations in a wider age the proposed automatic methodologies, specifically their
cohort. evaluation in samples of different ethnicities, to improve
The studies that applied automatic methodologies to their generalisation capabilities.
perform binary age classification or multiple group age
Funding Open Access funding provided thanks to the CRUE-CSIC
classification are not as numerous as those aimed at
agreement with Springer Nature. This work has received financial
performing numeric age regression. Moreover, traditional support from Consellerı́a de Cultura, Educación e Ordenación
methods in this regard reached very good results, with Universitaria (accreditation 2019–2022 ED431G-2019/04 and Group
almost no room for improvement. However, the proposed with Growth Potential ED431B 2020–2022 GPC2020/27) and the
automated methodologies kept the same high-performance European Regional Development Fund (ERDF), which acknowledges
the CiTIUS-Research Center in Intelligent Technologies of the
level while providing a series of benefits such as economic University of Santiago de Compostela as a Research Center of the
and resource-saving [59, 61]. Galician University System.
Regarding the specific computational approaches fol-
Data availability Not applicable.
lowed by the automatic methods, most studies rely on fully
automatic and deep-learning-based solutions, which lead to
two key enhancements. As with every end-to-end method,
Declarations
the dataset only needs to be annotated with the expected out- Ethics approval and consent to participate Not applicable.
put, that is, the age, reducing the time of this process and,
thus, making it possible to compile bigger datasets of thou- Research involving human participants and/or animals Not applica-
sands of images. Furthermore, these methods do not rely on ble.
specific bone structures designated by an expert, but rather
Consent for publication Not applicable.
on the image parts that the algorithm considers to be most
relevant for that specific task. As there is no need for specific Conflict of interest The authors declare no competing interests.
1140 International Journal of Legal Medicine (2023) 137:1117–1146

Open Access This article is licensed under a Creative Commons 18. Gleiser I, Hunt EE (1955) The estimation of age and sex of
Attribution 4.0 International License, which permits use, sharing, praeadolescent children from bones and teeth. Am J Phys Anthr
adaptation, distribution and reproduction in any medium or format, as 13:479–488
long as you give appropriate credit to the original author(s) and the 19. Moorrees CFA, Fanning EA, Hunt EEH Jr (1963) Age variation
source, provide a link to the Creative Commons licence, and indicate of formation stages for ten permanent teeth. J Dent Res
if changes were made. The images or other third party material in this 42(6):1490–1502
article are included in the article’s Creative Commons licence, unless 20. Köhler S, Schmelzle R, Loitz C, Püschel K (1994) Development
indicated otherwise in a credit line to the material. If material is not of wisdom teeth as a criterion of age determination. Ann Anat -
included in the article’s Creative Commons licence and your intended Anat Anz 176(4):339–345
use is not permitted by statutory regulation or exceeds the permitted 21. Liliequist B, Lundberg M (1971) Skeletal and tooth develop-
use, you will need to obtain permission directly from the copyright ment: a methodologic investigation. Acta Radiol 11(2):97–112
holder. To view a copy of this licence, visit http://creativecommons. 22. De Tobel J, Hillewig E, Verstraete K (2017) Forensic age
org/licenses/by/4.0/. estimation based on magnetic resonance imaging of third molars:
converting 2d staging into 3d staging. Ann Hum Biol 44(2):121–
129
23. Cameriere R, Ferrante L, Cingolani M (2006) Age estimation in
References children by measurement of open apices in teeth. Int J Legal Med
120(1):49–52
1. Daniel F (2010) Forensic age estimation in human skeletal 24. Vadivel I, Ganapathy N, Jisha G, Keerthipriyadharshini T,
remains: current concepts and future directions. Leg Med Thangadurai M, Yoithapprabhunath TR (2018) Digitized mor-
12(1):1–7 phometric analysis using maxillary canine and mandibular first
2. Garvin HM, Passalacqua NV (2012) Current practices by molar for age estimation in South Indian population. Open Dent
forensic anthropologists in adult skeletal age estimation. J J 12:762
Forensic Sci 57(2):427–433 25. Bardal R, Soltani M (2021) Age estimation by pulp/tooth area
3. Christensen AM, Passalacqua NV, Bartelink EJ (2019) Age and width ratios in panoramic images of upper and lower canines
estimation, chapter 10, pages 307–349. Academic Press, second among the Iranian population. Braz Dent Sci 24(4)
edition. ISBN 978-0-12-815734-3 26. Kvaal SI, Kolltveit KM, Thomsen IO, Solheim T (1995) Age
4. Guy W (2001) A review of the most commonly used dental age estimation of adults from dental radiographs. Forensic Sci Int
estimation techniques. J Forensic Odontostomatol 19(1):9–17 74(3):175–185
5. Panchbhai AS (2011) Dental radiographic indicators, a key to age 27. Ikeda N, Umetsu K, Kashimura S, Suzuki T, Oumi M (1985)
estimation. Dentomaxillofac Radiol 40(4):199–212 Estimation of age from teeth with their soft x-ray findings. Jpn J
6. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Legal Med 39(3):244–250
Mulrow CD, Shamseer Larissa, Tetzlaff JM, Elie AA, Brennan 28. Rai A, Acharya AB, Naikmasur VG (2016) Age estimation by
SE et al (2021) The prisma 2020 statement: an updated guideline pulp-to-tooth area ratio using cone-beam computed tomography:
for reporting systematic reviews. Syst Rev 10(1):1–11 a preliminary analysis. J Forensic Dent Sci 8(3):150
7. Kaarina H (1970) The formation and the alveolar and clinical 29. Nudel I, Pokhojaev A, Hausman BS, Bitterman Y, Shpack NS,
eruption of the permanent teeth. an orthopantomographic study. May H, Sarig R (2020) Age estimation of fragmented human
Suom Hammaslaak Toim 66(3):103 dental remains by secondary dentin virtual analysis. Int J Legal
8. Schour I, Massler M (1944) The development of the human Med 134(5):1853–1860
dentition. J Am Dent Assoc 28:1153–1160 30. Cameriere R, Ferrante L, Cingolani M (2004) Variations in
9. Ubelaker DH, LaurinGuttenplanGrant. (1989) Human skeletal pulp/tooth area ratio as an indicator of age: a preliminary study.
remains: preservation or reburial? Yearb Phys Anthropol 32:249– J Forensic Sci 49(2):1–3
287 31. Olze A, Solheim T, Schulz R, Kupfer M, Schmeling A (2010)
10. AlQahtani SJ, Hector MP, Liversidge Helen M . (2010) Brief
Evaluation of the radiographic visibility of the root pulp in the
communication: the London Atlas of human tooth development
lower third molars for the purpose of forensic age estimation in
and eruption. Am J Phys Anthropol 142(3):481–490
living individuals. Int J Legal Med 124(3):183–186
11. Blenkin M, Taylor J (2012) Age estimation charts for a modern
32. Lucas VS, McDonald F, Andiappan M, Roberts G (2017) Dental
Australian population. Forensic Sci Int 221(1-3):106–112
12. Nolla CMN (1960) The development of the permanent teeth. J age estimation—root pulp visibility (rpv) patterns: a reliable
Dent Child 27:254–266 mandibular maturity marker at the 18 year threshold. Forensic
13. Demirjian A, Goldstein H, Tanner JM (1973) A new system of Sci Int 270:98–102
dental age assessment. Hum Biol 211–227 33. Olze A, Solheim T, Schulz R, Kupfer M, Pfeiffer H, Schmeling
14. Willems G, Olmen AV, Spiessens B, Carels C (2001) Dental age A (2010) Assessment of the radiographic visibility of the
estimation in Belgian children Demirjian’s technique revisited. J periodontal ligament in the lower third molars for the purpose
Forensic Sci 46(4):893–895 of forensic age estimation in living individuals. Int J Legal Med
15. Bedek I, Dumančić J, Lauc T, Marušić M, Čuković-Bagić I 124(5):445–448
(2020) New model for dental age estimation: Willems method 34. Lucas VS, McDonald F, Andiappan M, Roberts G (2017) Dental
applied on fewer than seven mandibular teeth. Int J Legal Med age estimation: periodontal ligament visibility (PLV)—pattern
134(2):735–743 recognition of a conclusive mandibular maturity marker related
16. Chaillet N, Willems G (2004) Dental maturity in Belgian to the lower left third molar at the 18-year threshold. Int J Legal
children using Demirjian’s method and polynomial functions: Med 131:797–801
new standard curves for forensic and clinical use. J Forensic 35. Gösta G (1950) Age determinations on teeth. J Am Dent Assoc
Odontostomatol 22(2):18–27 41(1):45–54
17. Blenkin MRB, Evans W (2010) Age estimation from the teeth 36. Olze A, Hertel J, Schulz R, Wierer T, Schmeling A (2012)
using a modified Demirjian system. J Forensic Sci 55(6):1504– Radiographic evaluation of Gustafson’s criteria for the purpose
1508 of forensic age diagnostics. Int J Legal Med 126(4):615–621
International Journal of Legal Medicine (2023) 137:1117–1146 1141

37. Timme M, Timme WH, Olze A, Ottow C, Ribbecke S, Dettmeyer by deep learning. Int J Comput Assist Radiol Surg 16(12):2215–
PR, Schmeling A (2017) Dental age estimation in the living 2224
after completion of third molar mineralization: new data for 55. Milošević D, Vodanović M, Galić I, Subašić M (2022) Auto-
Gustafson’s criteria. Int J Legal Med 131(2):569–577 mated estimation of chronological age from panoramic dental
38. Motawei SM, Helaly AMN, Aboelmaaty WM, Elmahdy K, x-ray images using deep learning. Expert Syst Appl 189:116038
Shabka OA, Liu H (2020) Length of the ramus of the mandible 56. De Tobel J, Radesh P, Vandermeulen D, Thevissen PW (2017)
as an indicator of chronological age and sex: a study in a group An automated technique to stage lower third molar development
of Egyptians. Forensic Sci Int: Reports 2:100066 on panoramic radiographs for age estimation: a pilot study. J
39. Acharya AB (2017) A digital method of measuring the gonial Forensic Odontostomatol 35(2):42
angle on radiographs for forensic age estimation. J Forensic 57. Boedi RM, Banar N, De Tobel J, Bertels J, Vandermeulen D,
Radiol Imaging 11:18–23 Thevissen PW (2020) Effect of lower third molar segmentations
40. Olze A, Niekerk PV, Schulz R, Schmeling A (2007) Studies of on automated tooth development staging using a convolutional
the chronological course of wisdom tooth eruption in a black neural network. J Forensic Sci 65(2):481–486
African population. J Forensic Sci 52(5):1161–1163 58. Banar N, Bertels J, Laurent F, Boedi RM, De Tobel J, Thevissen
41. Sehrawat JS, Singh M (2020) Application of Kvaal’s radiological P, Vandermeulen D (2020) Towards fully automated third molar
method for dental age estimation of Ajnala skeletal remains: a development staging in panoramic radiographs. Int J Legal Med
cross-validation study. Forensic Imaging 22:200401 1–11
42. Mincer HH, Harris E, Berryman HE (1993) The abfo study 59. Kim S, Lee Y-H, Noh Y-K, Park FC, Auh Q-S (2021) Age-group
of third molar development and its use as an estimator of determination of living individuals using first molar images
chronological age. J Forensic Sci 38(2):379–390 based on artificial intelligence. Sci Rep 11
43. De Luca R, Biagi S, Begnoni, Farronato G, Cingolani M, Merelli 60. Kahaki SMM, Nordin M, Ahmad NS, Arzoky M, Ismail W et
V, Ferrante L, Cameriere R (2014) Accuracy of Cameriere’s cut- al (2020) Deep convolutional neural network designed for age
off value for third molar in assessing 18 years of age. Forensic assessment based on orthopantomography data. Neural Comput
Sci Int 235:102–e1 Appl 32(13):9357–9368
44. Yellapurkar S, Srikant N, Naik DG, Shenoy N, Gundeti V 61. Guo Y-C, Han M, Chi Y, Long H, Zhang D, Yang J, Yang
(2018) Age estimation using 2nd molar development: an efficient Y, Chen T, Du S (2021) Accurate age classification using
indicator of adolescent age. Romanian J Leg Med 26(3):274–278 manual method and deep convolutional neural network based on
45. Pereira CP, Rodrigues A, Santos A, Salvado F, Santos R, orthopantomogram images. Int J Legal Med 135(4):1589–1597
Cameriere R (2021) Cut-off for the legal ages in the Portuguese 62. Cameriere R, Ferrante LD, De Angelis D, Scarpino F, Galli F
population by third maturity index: measures of accuracy. Arch (2008) The comparison between measurement of open apices of
Oral Biol 125:105089 third molars and Demirjian stages to test chronological age of
46. Čular L, Tomaić M, Subašić M, Šarić T, Sajković T, Vodanović over 18 year olds in living subjects. Int J Legal Med 122(6):493–
M (2017) Dental age estimation from panoramic x-ray images 497
using statistical models. In: Proceedings of the 10th international 63. Baylis S, Bassed R (2017) Precision and accuracy of commonly
symposium on image and signal processing and analysis, pp 25– used dental age estimation charts for the New Zealand
30. IEEE population. Forensic Sci Int 277:223–228
47. Štern D, Payer C, Giuliani N, Urschler M (2018) Automatic age 64. Berkvens ME, Fairgrieve SI, Keenan S (2017) A comparison
estimation and majority age classification from multi-factorial of techniques in age estimation using the third molar. Can Soc
mri data. IEEE J Biomed Health Inform 23(4):1392–1403 Forensic Sci J 50(2):74–83
48. De Back W, Seurig S, Wagner S, Marré B, Roeder I, Scherf 65. Sheriff SO, Medapati RHR, Ankisetti SA, Gurrala VRR, Haritha
N (2019) Forensic age estimation with Bayesian convolutional K, Pulijala S, Balla SB (2020) Testing the accuracy of bedek
neural networks based on panoramic dental x-ray imaging. In: others’s new models based on 1-to-7 mandibular teeth for age
2019 international conference on medical imaging with deep estimation in 7-15 year old south Indian children. J Forensic
learning (MIDL) Odontostomatol 38(2):22
49. Vila-Blanco N, Carreira MJ, Varas-Quintana P, Balsa-Castro 66. Rivera M, De Luca S, Aguilar L, Palacio LAV, Galić I, Cameriere
C, Tomás I (2020) Deep neural networks for chronological R (2017) Measurement of open apices in tooth roots in colombian
age estimation from OPG images. IEEE Trans Med Imaging children as a tool for human identification in asylum and criminal
39(7):2374–2384 proceedings. J Forensic Leg Med 48:9–14
50. Hou W, Liu L, Gao J, Zhu A, Pan K, Sun H, Zheng N (2021) 67. Li M-j, Chu G, Han M-q, Chen T, Zhou H, Guo Y (2019)
Exploring effective DNN models for forensic age estimation Application of the Kvaal method for age estimation using digital
based on panoramic radiograph images. In: 2021 international panoramic radiography of Chinese individuals. Forensic Sci Int
joint conference on neural networks (IJCNN), pages 1–8. IEEE 301:76–81
51. Pham CV, Lee S-J, Kim S-Y, Lee S, Kim S-H, Kim H-S 68. Roh B, Lee W-J, Ryu J-W, Ahn J-M, Yoon C, Lee S-S (2018)
(2021) Age estimation based on 3D post-mortem computed The application of the Kvaal method to estimate the age of live
tomography images of mandible and femur using convolutional Korean subjects using digital panoramic radiographs. Int J Legal
neural networks. PLos One 16(5):e0251388 Med 132(4):1161–1166
52. Wallraff S, Vesal S, Syben C, Lutz R, Maier A (2021) 69. Dabbaghi A, Kazemi M (2020) Age estimation in iranian
Age estimation on panoramic dental x-ray images using deep population using tooth/pulp ratio in mandibular premolars by
learning. In: Bildverarbeitung für die medizin 2021, pp 186–191. panoramic radiography. Avicenna J Dent Res 12(2):40–45
Springer 70. Cameriere R, De Luca S, Soriano Vázquez I, Kiṡ HC, Pigolkin Y,
53. Zheng Q, Ge Z, Du H, Li G (2021) Age estimation based on Kumagai A, Ferrante L (2020) A full bayesian calibration model
3D pulp chamber segmentation of first molars from cone-beam– for assessing age in adults by means of pulp/tooth area ratio in
computed tomography by integrated deep learning and level set. periapical radiography. Int J Legal Med:1–9
Int J Legal Med 135(1):365–373 71. Chaudhary MA, Liversidge HM (2017) A radiographic study
54. Vila-Blanco N, Varas-Quintana P, Aneiros-Ardao Á, Tomás I, estimating age of mandibular third molars by periodontal
Carreira MJ (2021) Automated description of the mandible shape ligament visibility. J Forensic Odontostomatol 35(2):79
1142 International Journal of Legal Medicine (2023) 137:1117–1146

72. Timme M, Timme WH, Olze A, Ottow C, Gladitz J, Pfeiffer teeth with the London Atlas of tooth development. J Forensic
H, Dettmeyer R, Schmeling A (2019) Examination of regressive Odontostomatol 38(1):39
features of third molars for the purpose of age assessment in the 89. Gelbrich B, Carl C, Gelbrich G (2020) Comparison of three
living by means of rescaled regression analyses. Int J Legal Med methods to estimate dental age in children. Clin Oral Investig
133(6):1949–1955 24(7):2469–2475
73. de Moraes Correia A, da Silva Barbosa D, da Silva Alcantara JA, 90. da Silveira Sousa AM, Jacometti V, AlQahtani S, da Silva RHA
da Costa Oliveira PM, de Barros Silva PG, Franco A, Bezerra (2020) Age estimation of Brazilian individuals using the London
TP (2020) Performance and comparison of the london Atlas Atlas. Arch Oral Biol 113:104705
technique and Cameriere’s third molar maturity index (I3M) for 91. Alkandiri F, Karimi A, Draft D, Lucas VS, Roberts G (2021)
allocating individuals below or above the threshold of 18 years. Dental age estimation: a comparison of three methods of
Forensic Sci Int 317:110512 estimating dental age in a population of Kuwaiti children and
74. Mwesigwa CL, Kutesa AM, Munabi IG, Kabenge CA, Buwembo adolescents. Forensic Sci Int: Reports 3:100214
W (2019) Accuracy of the lower third molar radiographic 92. Sapna H, Patodia A, Dixit U (2017) A comparison of the
imaging to estimate age among Ugandan young people. BMC validity of the Demirjian, Willems, Nolla and Häävikko methods
Res Notes 12(1):1–6 in determination of chronological age of 5–15 year-old Indian
75. Pinchi V, Pradella F, Vitale G, Rugo D, Nieri M, Norelli G-A children. J Forensic Leg Med 50:49–57
(2016) Comparison of the diagnostic accuracy, sensitivity and 93. Melo M, Ata-Ali J (2017) Accuracy of the estimation of dental
specificity of four odontological methods for age evaluation in age in comparison with chronological age in a Spanish sample
Italian children at the age threshold of 14 years using ROC of 2641 living subjects using the Demirjian and Nolla methods.
curves. Med Sci Law 56(1):13–18 Forensic Sci Int 270:276–e1
76. AlQahtani S, Kawthar A, AlAraik A, AlShalan A (2017) Third 94. da Luz LCP, Anzulović D, Benedicto EN, Galić I, Hrvoje brkić,
molar cut-off value in assessing the legal age of 18 in Saudi Biazevic MGH (2019) Accuracy of four dental age estimation
population. Forensic Sci Int 272:64–67 methodologies in Brazilian and Croatian children. Sci Justice
77. Cortés MMP, Rojo R, Garcı́a EA, Martı́nez MRM (2020) 59(4):442–447
Accuracy assessment of dental age estimation with the Willems, 95. Abdulrakeeb M, Alqadi M, Abuaffan AH (2020) The reliability
Demirjian and Nolla methods in Spanish children: comparative of Fishman and Nolla methods in prediction of chronological age
cross-sectional study. BMC Pediatr 20(1):1–9 of Yemeni children. Braz Dent Sci 23(3):9–p
78. Adriana CP, Pires AC, de Sousa Santos RFV, Pereira CP (2021) 96. S Yassin S, Alalmai BM, Huaylah SA, Althobati M, Alhamdi
Dental age assessment by the pulp/tooth area proportion in cone FA, Togoo R (2020) Accuracy of estimating chronological age
beam computed tomography: is medico-legal application for age from Nolla’s method of dental age estimation in a population
estimation reliable. J Forensic Odontostomatol 39(2):2 of Southern Saudi Arabian children. Niger J Clin Pract
79. Augusto D, Pereira CP, Rodrigues A, Salvado F, Santos R (2021) 23(12):1753–1753
Dental age assessment by I 2M and I 3M: Portuguese legal 97. Han M-q, Jia S-x, Wang C-X, Chu G, Chen T, Zhou H, Guo Y-c
age thresholds of 12 and 14 year olds. Acta Stomatol Croat (2020) Accuracy of the Demirjian, Willems and Nolla methods
55(1):45–55 for dental age estimation in a Northern Chinese population. Arch
80. Rongguang W, Chaudhari P, Davatzikos C (2022) Embracing Oral Biol 118:104875
the disharmony in medical imaging: a simple and effective 98. Koċ A, Özlek E, Talmaċ AG¨Ö (2021) Accuracy of the London
framework for domain adaptation. Med Image Anal 76:102309 Atlas, Willems, and Nolla methods for dental age estimation:
81. Sapna H, Shah K, Dixit U (2016) A comparative evaluation a cross-sectional study on Eastern Turkish children. Clin Oral
of the applicability of two adapted Häävikko methods for Investig 25(8):4833–4840
age estimation of 5-15 year old Indian children. J Forensic 99. Demirjian A, Goldstein H (1976) New systems for dental
Odontostomatol 34(2):21 maturity based on seven and four teeth. Ann Hum Biol 3(5):411–
82. Kumaresan R, Cugati N, Chandrasekaran B, Karthikeyan P 421
(2016) Reliability and validity of five radiographic dental-age 100. Amanullah A, Ullah U, Yunus S, Munim A (2016) Development
estimation methods in a population of Malaysian children. J stages of third molar tooth for estimation of chronological age in
Investig Clin Dent 7(1):102–109 children and young adult. Pak J Med Health Sci 10(3):750–754
83. Benedicto EN, Azevedo AC, Michel-Crosato E, Biazevic MGH 101. Duangto P, Janhom A, Mahakkanukrauh P, Iamaroon A (2016)
(2018) Validity and accuracy of three radiographic dental age New prediction models for dental age estimation in Thai children
estimation methods in Brazilians. Forensic Sci Int 283:128–135 and adolescents. Forensic Sci Int 266:583–e1
84. Sezer B, Ċarıkċıoġlu B (2022) Accuracy of the London Atlas, 102. Cavrić J, Vodanović M, Marušić A, Galić I (2016) Time of
Haavikko’s method and Cameriere’s European formula of dental mineralization of permanent teeth in children and adolescents in
age estimation in Turkish children. Leg Med 54:101991 Gaborone, botswana. Ann Anat - Anat Anz 203:24–32
85. Pavlović S, Pereira CP, de Sousa Santos VRF (2017) Age 103. Chiam S-L, Blenkin M, Taylor J, Moananui RT (2016) Validity
estimation in Portuguese population: the application of the of a modified Demirjian system based on an Australian dataset–
London Atlas of tooth development and eruption. Forensic Sci simple maturity score in age estimation. Aust J Forensic Sci
Int 272:97–103 48(5):571–582
86. Roshanak G, Ghodousi A, Poordavar E (2019) Comparison of 104. Bunyarit SS, Jayaraman J, Naidu MK, Ying RPY, Danaee M,
the accuracy of the London Atlas and Smith method in dental Nambiar P (2017) Modified method of dental age estimation of
age estimation in 5–15.99-year-old Iranians using the panoramic Malay juveniles. Leg Med 28:45–53
view. Int J Legal Med 133(1):189–195 105. Alsaffar H, Elshehawi W, Roberts G, Lucas V, McDonald
87. Alsudairi DM, AlQahtani SJ (2019) Testing and comparing the F, Camilleri S (2017) Dental age estimation of children and
accuracy of two dental age estimation methods on Saudi children: adolescents: validation of the maltese reference data set. J
measurements of open apices in teeth and the London Atlas of Forensic Leg Med 45:29–31
tooth development. Forensic Sci Int 295:226–e1 106. Saadé A., Baron P, Noujeim Z, Azar D (2017) Dental and
88. Sharma P, Wadhwan V (2020) Comparison of accuracy of age skeletal age estimations in Lebanese children: a retrospective
estimation in Indian children by measurement of open apices in cross-sectional study. J Int Soc Prev Community Dent 7(3):90
International Journal of Legal Medicine (2023) 137:1117–1146 1143

107. Kelmendi J, Vodanović M, Koçani F, Bimbashi V, Mehmeti B, 125. Amro M, Ali M, Ahmed WH, Khattab NM (2019) Applicability
Galić I (2018) Dental age estimation using four Demirjian’s, of Demirjian’s method for dental age estimation in a group of
Chaillet’s and Willems’ methods in Kosovar children. Leg Med Egyptian children. BDJ open 5(1):1–6
33:23–31 126. Shen C, Pan J, Yang Z, Mou H, Tao J, Ji F (2020) Applicability
108. Almotairy N, Pegelow M (2018) Dental age comparison in of 2 dental age estimation methods to Taiwanese population. Am
patients born with unilateral cleft lip and palate to a control J Forensic Med Pathol 41(4):269–275
sample using Demirjian and Willems methods. Eur J Orthod 127. Memorando JR (2020) Evaluation of mandibular third molar for
40(1):74–81 age estimation of Filipino population age 9–23 years. J Forensic
109. Agrawal NK, Hackman L, Dahal S (2018) Dental age assessment Odontostomatol 38(1):26
using Demirjian’s eight teeth method and Willems method in a 128. Pinchi V, Bianchi I, Pradella F, Vitale G, Focardi M, Toni I,
tertiary hospital. JNMA J Nepal Med Assoc 56(214):912–916 Ferrante L, Bucci A (2021) Dental age estimation in children
110. Esan TA, Schepartz LA (2018) Accuracy of the Demirjian and affected by juvenile rheumatoid arthritis. Int J Legal Med
Willems methods of age estimation in a black Southern African 135(2):619–629
population. Leg Med 31:82–89 129. Subedi N, Parajuli U, Paudel IS, Mallik M (2020) Demirjian’s
111. Sobieska E, Fester A, Nieborak M, Zadurska M (2018) eight teeth method for dental age estimation in Nepalese
Assessment of the dental age of children in the polish population population. J Nepal Health Res Counc 18(49):686–91
with comparison of the Demirjian and the Willems methods. Med 130. Galibourg A, Cussat-Blanc S, Dumoncel J, Telmon N, Monsarrat
Sci Monit 24:8315 P, Maret D (2021) Comparison of different machine learning
112. Nemsi H, ben Daya M, Mesrati MA, Masmoudi F, Maatouk F, approaches to predict dental age using Demirjian’s staging
Bouanène I, Aissaoui A (2018) Applicability of Willems method approach. Int J Legal Med 135(2):665–675
for dental age estimations in Tunisian children. Pediatr Dent J 131. Karimi A, Qudeimat MA, Lucas VS, Roberts G (2021) Dental
28(3):141–147 age estimation: development and validation of a reference data
113. Al Balushi S, Thomson WM, Al-Harthi L (2018) Dental age set for Kuwaiti children, adolescents, and young adults. Arch
estimation of Omani children using Demirjian’s method. Saudi Oral Biol 127:105130
Dent J 30(3):208–213 132. B Rodriguez A, Cucina EV, Hoyos R, Cardoso HFV (2021)
114. Kumagai A, Willems G, Franco A, Thevissen P (2018) Age A critical test of twelve methods for estimating age using
estimation combining radiographic information of two dental and radiographic staging of developing teeth on a sample of 6-to 15-
four skeletal predictors in children and subadults. Int J Legal Med year-old children from Mérida, Yucatán (México). Int J Legal
132(6):1769–1777 Med 135(6):2457–2467
115. Liu Y, Geng K, Chu Y, Xu M, Zha L (2018) Third molar min- 133. Jayaraman J, Mendez MJC, Gakunga PT, Roberts G (2022) Age
eralization in relation to chronologic age estimation of the han estimation of hispanic children in the United States: development
in Central Southern China. Int J Legal Med 132(5):1427–1435 and validation of dental reference dataset based on two staging
116. Khdairi N, Halilah T, Khandakji MN, Jost-Brinkmann P-G, systems. Leg Med 56:102033
Bartzela T (2019) The adaptation of Demirjian’s dental age 134. Shi L, Zhou Y, Ting Lu, Fan F, Zhu L, Suo Y, Chen Y,
estimation method on North German children. Forensic Sci Int Deng Z (2022) Dental age estimation of tibetan children and
303:109927 adolescents: comparison of Demirjian, Willems methods and a
117. Metsäniitty M, Waltimo-Sirén J, Fieuws S, Thevissen P (2019) newly modified Demirjian method. Leg Med 102013
135. Lauc T, Nakaš E, Latić-Dautović M, Džemidžić V, Tiro A, Rupić
Dental age estimation in somali children and sub-adults
I, kostić M, Galić I (2017) Dental age in orthodontic patients with
combining permanent teeth and third molar development. Int J
different skeletal patterns. BioMed Res Int 2017
Legal Med 133(4):1207–1215
136. Marinkovic N, Milovanovic P, Djuric M, Nedeljkovic N, Zelic
118. Ozveren N, Serindere G (2018) Comparison of the applicability
K (2018) Dental maturity assessment in serbian population:
of Demirjian and Willems methods for dental age estimation
a comparison of Cameriere’s European formula and Willems’
in children from the Thrace region, Turkey. Forensic Sci Int
method. Forensic Sci Int 288:331–e1
285:38–43 137. Willems G, Lee S-S, Uys A, Bernitz H, de Llano-Pérula MC,
119. Lopes LJ, Nascimento HAR, Lima GP, Dos Santos LAN, de Fieuws S, Thevissen P (2018) Age estimation based on Willems
Paula Queluz D, Freitas DQ (2018) Dental age assessment: which method versus new country-specific method in South African
is the most applicable method. Forensic Sci Int 284:97–100 black children. Int J Legal Med 132(2):599–607
120. Birchler FA, Kiliaridis S, Combescure C, Julku J, Pirttiniemi 138. Metsäniitty M, Waltimo-Sirén J, Ranta H, Fieuws S, Thevissen
PM, Vazquez L (2019) Dental age assessment on panoramic P (2018) Dental age estimation in somali children using the
radiographs: comparison between two generations of young Willems et al. model. Int J Legal Med 132(6):1779–1786
Finnish subjects. J Int Med Res 47(1):311–324 139. Kurniawan A, Agitha SRA, Margaretha MS, Utomo H, Chusida
121. Ranasinghe S, Perera J, Taylor J, Tennakoon A, Pallewatte A, A, Sosiawan A, Marini MI, Rizky BN (2020) The applicability
Jayasinghe R (2019) Dental age estimation using radiographs: of Willems dental age estimation method for Indonesian children
towards the best method for Sri Lankan children. Forensic Sci Int population in Surabaya. Egypt J Forensic Sci 10(1):1–4
298:64–70 140. Ortega-Pertuz AI, Piña-D’Abreu M (2020) Comparison of the
122. Kermani M, Yazdi FT, Haghighi MA (2019) Evaluation of the Willems I and Willems II methods for forensic age estimation in
accuracy of Demirjian’s method for estimating chronological age Venezuelan children. Span J Leg Med 46(1):4–11
from dental age in Shiraz, Iran: using geometric morphometrics 141. Pan J, Shen C, Yang Z, Fan L, Wang M, Shen S, Tao J, Ji F (2021)
method. Clin Exp Dent Res 5(3):191–198 A modified dental age assessment method for 5-to 16-year-old
123. Ginzelova K, Dostalova T, Eliasova H, Bruna R, Vinsu A Eastern Chinese children. Clin Oral Investig 25(6):3463–3474
(2019) Comparison of dental and chronological age based on 142. Burak Ç, Deġirmenci A (2021) Accuracy of the radiographic
cbct-generated panoramic images and reconstructed 3D images. methods of Willems, Cameriere and Blenkin–Evans on age
Anthropol Anz 76(1):49–56 estimation for Turkish children in the Northwest Anatolia region.
124. Alqadi MA, Abuaffan AH (2019) Sultan Qaboos Univ Med J. Aust J Forensic Sci 1–12
Validity of the Demirjian and Fishman methods for predicting 143. Rocha LT, Ingold MS, Panzarella FK, Santiago BM, de Oliveira
chronological age amongst Yemeni children 19(1):e26 RN, de Macedo Bernardino I, Makeeva I, Junqueira JLC, Mânica
1144 International Journal of Legal Medicine (2023) 137:1117–1146

S, Franco A (2022) Applicability of Willems method for age 160. Akay G, Karataṡ MS, Karadağ Ö, Gungor K, Üċok CÖ (2020)
estimation in Brazilian children: performance of multiple linear An application for Olze’s method and tooth coronal index for
regression and artificial neural network. Egypt J Forensic Sci age estimation of a Turkish adult population. Aust J Forensic Sci
12(1):1–9 52(6):699–710
144. Yassin SM, Alkahtani ZM, Khader MA, Almoammar S, Dawasaz 161. Gok E, Fedakar R, Kafa IM (2020) Usability of dental
AA, Asif SM, Togoo RA (2022) Applicability of Willems model pulp visibility and tooth coronal index in digital panoramic
of dental age estimation in a sample of Southern Saudi Arabian radiography in age estimation in the forensic medicine. Int J
children. Aust J Forensic Sci 1–17 Legal Med 134(1):381–392
145. de Llano-Perula MC, Thevissen E, Nyamunga P, Fieuws S, 162. Herianti VR, Oscandar F, Dardjan M (2021) A retrospective
Kanini M, Willems G (2021) Validating dental age estimation institutional study of human age determination by evaluating the
in Kenyan black children and adolescents using the Willems pulp length and width ratio of the maxillary lateral incisor on
method. Med Sci Law 61(3):180–185 panoramic radiographs in Indonesian subjects. Imaging Sci Dent
146. Hegde S, Patodia A, Shah K, Dixit U (2019) The applicability of 51(4):421
the Demirjian, Willems and Chaillet standards to age estimation 163. Hisham S, Abdullah N, Noor MHM, Franklin D (2019)
of 5-15 year old Indian children. J Forensic Odontostomatol Quantification of secondary dentin formation based on the
37(1):40 analysis of MDCT scans and dental OPGs in a contemporary
147. Bunyarit SS, Nambiar P, Naidu MK, Ying RPY, Asif MK (2021) Malaysian population. Leg Med 36:59–66
Dental age estimation of malay children and adolescents: Chaillet 164. Akay G, Gungor K, Gurcan S (2019) The applicability of kvaal
and Demirjian’s data improved using artificial multilayer methods and pulp/tooth volume ratio for age estimation of the
perceptron neural network. Pediatr Dent J 31(2):176–185 turkish adult population on cone beam computed tomography
148. Fulton AJ, Liversidge HM (2016) A radiographic study of images. Aust J Forensic Sci 51(3):251–265
estimating age by deciduous mandibular canine and molar root 165. de Miranda J, de Cássia Silva Azevedo A, Michel-Crosato E,
resorption. Ann Anat - Anat Anz 203:33–37 Biazevic MGH (2020) Age estimation in Brazilian adults by
149. Štepanovskỳ M, Ibrová A, Buk Z, Velemı́nská J (2017) Novel Kvaal’s and Cameriere’s methods. Brazilian Oral Res 34
age estimation model based on development of permanent teeth 166. Limdiwala PG, Shah JS, Pillai JP (2021) Application of modified
compared with classical approach and other modern data mining Kvaal’s age estimation method on extracted maxillary second
methods. Forensic Sci Int 279:72–82 premolar tooth: an ex-vitro study using cone beam computed
150. Sartori V, Franco A, Linden M-S, Cardoso M, de Castro D, tomography. J Int Oral Health 13(4):356
Sartori A, Silva C, Trentin M, De Carli J (2021) Testing 167. Vossoughi M, Movahhedian N, Ghafoori A (2022) The impact of
international techniques for the radiographic assessment of third age mimicry bias on the accuracy of methods for age estimation
molar maturation. J Clin Exp Dent 13(12):e1182 based on Kvaal’s pulp/tooth ratios: a bootstrap study. Int J Legal
151. De Tobel J, Phlypo I, Fieuws S, Politis C, Verstraete KL, Med 136(1):269–278
Thevissen PW (2017) Forensic age estimation based on 168. Farhadian M, Salemi F, Saati S, Nafisi N (2019) Dental age
development of third molars: a staging technique for magnetic estimation using the pulp-to-tooth ratio in canines by neural
resonance imaging. J Forensic Odontostomatol 35(2):117 networks. Imaging Sci Dent 49(1):19–26
152. Cameriere R, Pacifici A, Pacifici L, Polimeni A, Federici F, 169. Jambunath U, Balaji P, Poornima G, Vinitra V, Gupta A, Peeyush
Cingolani M, Ferrante L (2016) Age estimation in children by S (2016) Dental age estimation by radiographic evaluation of
measurement of open apices in teeth with Bayesian calibration pulp/tooth ratio in mandibular canines and premolars. J Indian
approach. Forensic Sci Int 258:50–54 Acad Forensic Med 38(4):416–419
153. Santana SA, Bethard JD, Moore TL (2017) Accuracy of 170. Lee J-H, Lee C, Battulga B, Na JY, Hwang JJ, Kim YH, Han SS
dental age in nonadults: a comparison of two methods for age (2017) Morphological analysis of the lower second premolar for
estimation using radiographs of developing teeth. J Forensic Sci age estimation of Korean adults. Forensic Sci Int 281:186–e1
62(5):1320–1325 171. Kumar NN, Panchaksharappa MG, Annigeri RG (2016) Dig-
154. Halilah T, Khdairi N, Jost-Brinkmann P-G, Bartzela T (2018) itized morphometric analysis of dental pulp of permanent
Age estimation in 5–16-year-old children by measurement of mandibular second molar for age estimation of Davangere popu-
open apices: North German formula. Forensic Sci Int 293:103–e1 lation. J Forensic Leg Med 39:85–90
155. Angelakopoulos N, De Luca S, Palacio LA, Coccia E, Ferrante L, 172. Dehghani M, Shadkam E, Ahrari F, Dehghani M (2018) Age
Pinchi V, Cameriere R (2019) Age estimation by measuring open estimation by canines’ pulp/tooth ratio in an Iranian population
apices in teeth: a new formula for two samples of South African using digital panoramic radiography. Forensic Sci Int 285:44–49
black and white children. Int J Legal Med 133(5):1529–1536 173. Lee S-M, Oh S, Kim J, Kim Y-M, Choi Y-K, Kwak HH, Kim
156. AlShahrani I, Yassin SM, Togoo RA, Tikare S, Abdul Khader Y-Il (2017) Age estimation using the maxillary canine pulp/tooth
MA, Alkahtani ZM (2019) Age estimation by measurement of ratio in Korean adults: a CBCT buccolingual and horizontal
open apices in tooth roots: study using Saudi Arabian samples. J section image analysis. J Forensic Radiol Imaging 9:1–5
Forensic Leg Med 62:63–68 174. Haghanifar S, Ghobadi F, Vahdani N, Bijani A (2019) Age
157. Różyło-Kalinowska I, Kalinowski P, Krasicka E, Galić I, Mehdi estimation by pulp/tooth area ratio in anterior teeth using cone-
F, Cameriere R (2020) The Cameriere method using cone-beam beam computed tomography: comparison of four teeth. J Appl
computed tomography (CBCT) scans for dental age estimation Oral Sci 27
in children. Aust J Forensic Sci 1–15 175. Asif MK, Nambiar P, Mani SA, Ibrahim NB, Khan IM, Lokman
158. Shen S, Liu Z, Wang J, Fan L, Ji F, Tao J (2021) Machine learning NB (2019) Dental age estimation in malaysian adults based on
assisted Cameriere method for dental age estimation. BMC Oral volumetric analysis of pulp/tooth ratio using CBCT data. Leg
Health 21(1):1–10 Med 36:50–58
159. Jain S, Nagi R, Daga M, Shandilya A, Shukla A, Parakh A, 176. Zhan M-j, Chen X-g, Shi L, Lu T, Fan F, Zhang K (2021) Age
Laheji A, Singh R (2017) Tooth coronal index and pulp/tooth estimation in western chinese adults by pulp–tooth volume ratios
ratio in dental age estimation on digital panoramic radiographs— using cone-beam computed tomography. Aust J Forensic Sci
a comparative study. Forensic Sci Int 277:115–121 53(6):681–692
International Journal of Legal Medicine (2023) 137:1117–1146 1145

177. Muralidhar NV, Nitin P, Kumareswar S, Pillai A (2021) Pulp subject over 18 years of age using the gold standard of chrono-
tooth ratio-based age estimation of adults using CBCT images. J logical age as the comparator. J Forensic Sci 61(5):1238–1243
Orofac Sci 13(2):114 194. Friedrich RE, Schmidt K, Treszl A, Kersten JF (2016) Predictive
178. Kazmi S, Shepherd S, Revie G, Hector M, Mânica S (2022) values derived from lower wisdom teeth developmental stages
Exploring the relationship between age and the pulp and tooth on orthopantomograms to calculate the chronological age in
size in canines. a CBCT analysis. Aust J Forensic Sci 1–12 adolescence and young adults as a prerequisite to obtain age-
179. Helmy MA, Osama M, Elhindawy MM, Mowafey B, Taalab adjusted informed patient consent prior to elective surgical
YM, ElRahman HAA (2020) Volume analysis of second molar procedures in young patients with incomplete or mismatched
pulp chamber using cone beam computed tomography for personal data. GMS Interdiscip Plast Reconstr Surg DGPW, 5
age estimation in Egyptian adults. J Forensic Odontostomatol 195. Márquez-Ruiz AB, Treviño-Tijerina MC, González-Herrera AR,
38(3):25 Sánchez B, Valenzuela A (2017) Three-dimensional analysis of
180. Ge Z-p, Yang P, Li G, Zhang J-z, Ma X-c (2016) Age estimation third molar development to estimate age of majority. Sci Justice
based on pulp cavity/chamber volume of 13 types of tooth 57(5):376–383
from cone beam computed tomography images. Int J Legal Med 196. Lizarbe RJQ, Adrianzén CS, Quezada-Márquez MM, Galić I,
130(4):1159–1167 Cameriere R (2017) Demirjian’s stages and Cameriere’s third
181. Singal K, Sharma N, Kumar V, Singh P (2019) Coronal pulp molar maturity index to estimate legal adult age in Peruvian
cavity index as noble modality for age estimation: a digital image population. Leg Med 25:59–65
analysis. Egypt J Forensic Sci 9(1):1–8 197. Hassan FM, Moawad AM, Samir W, Helaly YR, Abu-Taleb NS
182. Elgazzar FM, Elboraey MO, El-Sarnagawy GN (2020) The (2021) Mandibular third molar maturation stage as indicator for
accuracy of age estimation from pulp chamber/crown volume the legal adult age in an Egyptian sample. Homo
ratio of canines obtained by cone beam computed tomography 198. Sharma P, Wadhwan V, Sharma N (2018) Reliability of
images: an Egyptian study. Egypt J Forensic Sci 10(1):1–10 determining the age of majority: a comparison between
183. Molina A, Bravo M, Fonseca GM, Márquez-Grant N, Martı́n- measurement of open apices of third molars and Demirjian
de Las-Heras S (2021) Dental age estimation based on pulp stages. J Forensic Odontostomatol 36(2):2
chamber/crown volume ratio measured on CBCT images in a 199. Melo M, Ata-Ali F, Gonzalez JMM, Cobo T (2022) Demirjian
Spanish population. Int J Legal Med 135(1):359–364 and Cameriere methods for age estimation in a Spanish sample
184. Asif MK, Nambiar P, Mani SA, Ibrahim NB, Khan IM, of 1386 living subjects. Sci Rep 12(1):1–8
Sukumaran P (2018) Dental age estimation employing CBCT 200. Franco RPAV, Franco A, Turkina A, Arakelyan M, Arzukanyan
scans enhanced with mimics software: comparison of two A, Velenko P, Bortolami PB, Makeeva I, da Silva RHA
different approaches using pulp/tooth volumetric analysis. J (2021) Third molar classification using gleiser and hunt system
Forensic Leg Med 54:53–61 modified by khöler in Russian adolescents–age threshold of 14
185. Zhang Z-y, Yan C-x, Min Q-m, Li S-q, Yang J-s, Guo Y-c, Jin W- and 16. Forensic Imaging 25:200443
f, Li L-j, Xing P-F, Li J (2019) Age estimation using pulp/enamel 201. Akkaya N, Yılancı HÖ, Boyacıoġlu H, Göksülük D, Özkan G
volume ratio of impacted mandibular third molars measured on (2019) Accuracy of the use of radiographic visibility of root
CBCT images in a Northern Chinese population. Int J Legal Med pulp in the mandibular third molar as a maturity marker at age
133(6):1925–1933 thresholds of 18 and 21. Int J Legal Med 133(5):1507–1515
186. Nemsi H, Salem NH, Bouanene I, Jomaa S, Belhadj BM, Mosrati 202. Manthapuri S, Bheemanapalli S, Namburu LP, Kunchala S,
MA, Aissaoui A, Amor B, Chadly C (2017) Age assessment Vankdoth D, Balla SB, Bathala V, Kasabu AL (2021) Can root
in canine and premolar by cervical axial sections of cone-beam pulp visibility in mandibular first molars be used as an alternative
computed tomography. Leg Med 28:31–36 age marker at the 16 year threshold in the absence of mandibular
187. Sonjaya D, Nambiar P, Ramli H, Alias A (2021) Modified third molars: an orthopantomographic study in a South Indian
Gustafson’s method for age estimation from premolars: a study sample. J Forensic Odontostomatol 39(2):21
employing dental panoramic tomographs. J Int Dent Medical Res 203. De Luca S, Aguilar L, Rivera M, Palacio LAV, Riccomi G,
14(2):750–756 Bestetti F, Cameriere R (2016) Accuracy of cut-off value by
188. Si X-q, Chu G, Olze A, Schmidt S, Schulz R, Chen T, Pfeiffer H, measurement of third molar index: study of a Colombian sample.
Guo Y-c, Schmeling A (2019) Age assessment in the living using Forensic Sci Int 261:160–e1
modified Gustafson’s criteria in a Northern Chinese population. 204. Cavrić J, Galić I, Vodanović M, Brkić H, Gregov J, Viva S,
Int J Legal Med 133(3):921–930 Rey L, Cameriere R (2016) Third molar maturity index (i3m)
189. Dezem TU, Franco A, Palhares CEM, Deitos AR, da Silva RH, for assessing age of majority in a black African population in
Santiago BM, Ribeiro ILA, Daruge Junior E (2021) Testing botswana. Int J Legal Med 130(4):1109–1120
the Olze and Timme methods for dental age estimation in 205. Zelic K, Galic I, Nedeljkovic N, Jakovljevic A, Milosevic O,
radiographs of Brazilian subadults and adults. Acta Stomatol Djuric M, Cameriere R (2016) Accuracy of Cameriere’s third
Croat 55(4):390–396 molar maturity index in assessing legal adulthood on Serbian
190. Koh KK, Tan JS, Nambiar P, Ibrahim N, Mutalik S, Asif MK population. Forensic Sci Int 259:127–132
(2017) Age estimation from structural changes of teeth and 206. Gulsahi A, De Luca S, Cehreli SB, Tirali RE, Cameriere R (2016)
buccal alveolar bone level. J Forensic Leg Med 48:15–21 Accuracy of the third molar index for assessing the legal majority
191. Thomas LA, Thomas LR, Balla SB, Gopalaiah H, Kanaparthi A, of 18 years in Turkish population. Forensic Sci Int 266:584–e1
Sravanthi GS, Palla A, Galic I (2021) Above or below 14 years? 207. Franklin D, Karkhanis S, Flavel A, Collini F, De Luca S,
an orthopantomographic study based on chronological course of Cameriere R (2016) Accuracy of a cut-off value based on
eruption of mandibular premolars and second molars in a sample the third molar index: validation in an Australian population.
of South Indian children. Leg Med 48:101814 Forensic Sci Int 266:575–e1
192. Antunovic M, Galic I, Ksenija Zelic K, Nenad Nedeljkovic N, 208. Dardouri AAK, Cameriere R, De Luca S, Vanin S (2016) Third
Lazic E, Djuric M, Cameriere R (2018) The third molars for molar maturity index by measurements of open apices in a libyan
indicating legal adult age in Montenegro. Leg Med 33:55–61 sample of living subjects. Forensic Sci Int 267:230–e1
193. Lucas VS, Andiappan M, McDonald F, Roberts G (2016) Dental 209. Balla SB, Nambi NK, Nambi S, Chinnala B, Bojji M, Mula AP,
age estimation: a test of the reliability of correctly identifying a Kondakamalli S, Vankdoth D, Tadakamadla SK (2022) Legal
1146 International Journal of Legal Medicine (2023) 137:1117–1146

age determined by a new threshold value of third molar maturity 218. Chu G, Wang Y-h, Li M-j, Han M-q, Zhang Z-y, Chen T,
index in subjects with impacted mandibular third molars: an Zhou H, Guo Y-c (2018) Third molar maturity index (i3m) for
orthopantomographic study in South Indian adolescents. Int J assessing age of majority in northern Chinese population. Int J
Legal Med 136(1):251–259 Legal Med 132(6):1759–1768
210. Cameriere R, Palacio LAV, Pinares J, Bestetti F, Paba R, Coccia 219. Moukarzel M, Angelakopoulos N, De Luca S, Palacio LAV,
E, Ferrante L (2018) Assessment of second (i2m) and third (i3m) Aquilanti E, Coccia E, Pigolkin YI, Cameriere R (2020) Validity
molar indices for establishing 14 and 16 legal ages and validation assessment of the third molar maturity index (i3M ) in a Lebanese
of the Cameriere’s I3M cut-off for 18 years old in Chilean sample of adolescents and young adults. Aust J Forensic Sci
population. Forensic Sci Int 285:205–e1 1–12
211. Angelakopoulos N, De Luca S, Palacio LAV, Coccia E, Ferrante 220. Ribier L, Saint-Martin P, Seignier M, Paré A, Brunereau L,
L, Cameriere R (2018) Third molar maturity index (i3M ) for Rérolle C (2020) Cameriere’s third molar maturity index in
assessing age of majority: study of a black South African sample. assessing age of majority: a study of a French sample. Int J Legal
Int J Legal Med 132(5):1457–1464 Med 134(2):783–792
212. Różyło-Kalinowska I, Kalinowski P, Kozek M, Galić I, 221. Scendoni R, Zolotenkova GV, Vanin S, Pigolkin YI, Cameriere R
Cameriere R (2018) Validity of the third molar maturity index (2020) Forensic validity of the third molar maturity index (i3M )
i3m for indicating the adult age in the Polish population. Forensic for age estimation in a Russian population. BioMed Res Int 2020
Sci Int 290:352–e1 222. Wang M, Wang J, Pan Y, Fan L, Shen Z, Ji F, Tao J (2020)
213. Kelmendi J, Cameriere R, Koçani F, Galić I, Mehmeti B, Applicability of newly derived second and third molar maturity
Vodanović M (2018) The third molar maturity index in indicating indices for indicating the legal age of 16 years in the southern
the legal adult age in Kosovar population. Int J Legal Med Chinese population. Leg Med 46:101725
132(4):1151–1159 223. Angelakopoulos N, Galić I, Balla SB, Kiṡ HC, Jiménez LG,
214. Spinas E, De Luca S, Lampis L, Cameriere R (2018) Is the Zolotenkova G, Yusof MYPM, Selmanagić H, Pandey H, Pereira
third molar maturity index (I3M) useful for a genetic isolate CP (2021) Comparison of the third molar maturity index (I3M )
population? Study of a sardinian sample of children and young between left and right lower third molars to assess the age
adults. Int J Legal Med 132(6):1787–1794 of majority: a multi-ethnic study sample. Int J Legal Med
215. Doğru HB, Gulsahi A, Ċehreli BS, Galić I, van der Stelt 135(6):2423–2436
P, Cameriere R (2018) Age of majority assessment in dutch 224. Yılancı HÖ, Akkaya N, Bulut G, Göksülük D, Bulut H,
individuals based on Cameriere’s third molar maturity index. Cameriere R (2021) Assessment of second (I2M) and third (I3M)
Forensic Sci Int 282:231–e1 molar maturity indices individually and in combination (I2M+
216. Balla SB, Chinni SS, Galic A, Alwala M, Machani P, Cameriere 3M) and the sum of the seven mandibular teeth indices (s) for the
R (2019) A cut-off value of third molar maturity index for legal age thresholds of 12 and 15 years in a turkish sample. Aust
indicating a minimum age of criminal responsibility: older or J Forensic Sci 1–13
younger than 16 years. J Forensic Leg Med 65:108–112
217. Kumagai A, Takahashi N, Palacio LAV, Giampieri A, Ferrante Publisher’s note Springer Nature remains neutral with regard to
L, Cameriere R (2019) Accuracy of the third molar index cut- jurisdictional claims in published maps and institutional affiliations.
off value for estimating 18 years of age: validation in a Japanese
samples. Leg Med 38:5–9

Affiliations
Nicolás Vila-Blanco1,2,3 · Paulina Varas-Quintana4 · Inmaculada Tomás1,3,4 · Marı́a J. Carreira1,2,3

Nicolás Vila-Blanco
[email protected]
Paulina Varas-Quintana
[email protected]
1 Centro Singular de Investigación en Tecnoloxı́as Intelixentes
(CiTIUS), Universidade de Santiago de Compostela, Santiago de
Compostela, Spain
2 Departamento de Electrónica e Computación, Escola Técnica
Superior de Enxeñarı́a, Universidade de Santiago de Compostela,
Santiago de Compostela, Spain
3 Instituto de Investigación Sanitaria de Santiago de Compostela
(IDIS), Santiago de Compostela, Spain
4 Oral Sciences Research Group, Special Needs Unit, Department
of Surgery and Medical Surgical Specialities, School of Medicine
and Dentistry, Universidade de Santiago de Compostela, Santiago
de Compostela, Spain