Journal of Dental Sciences (2021) 16, 508e522

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.e-jds.com

Review article

Developments, application, and

performance of artificial intelligence in
dentistry e A systematic review
Sanjeev B. Khanagar a,b*, Ali Al-ehaideb a,b,c,
Prabhadevi C. Maganur d**, Satish Vishwanathaiah d,
Shankargouda Patil e, Hosam A. Baeshen f, Sachin C. Sarode g,
Shilpa Bhandi h

a
Preventive Dental Science Department, College of Dentistry, King Saud Bin Abdulaziz University for
Health Sciences, Riyadh, Saudi Arabia
b
King Abdullah International Medical Research Center, Riyadh, Saudi Arabia
c
Dental Services, King Abdulaziz Medical City- Ministry of National Guard Health Affairs, Riyadh, Saudi
Arabia
d
Department of Preventive Dental Sciences, Division of Pedodontics, College of Dentistry, Jazan
University, Jazan, Saudi Arabia
e
Department of Maxillofacial Surgery and Diagnostic Sciences, Division of Oral Pathology, College of
Dentistry, Jazan University, Jazan, Saudi Arabia
f
Consultant in Orthodontics, Department of Orthodontics, College of Dentistry, King Abdulaziz
University, Jeddah, Saudi Arabia
g
Department of Oral and Maxillofacial Pathology, Dr. D.Y.Patil Dental College and Hospital, Dr. D. Y.
Patil Vidyapeeth, Pimpri, Pune 411018, Maharashtra, India
h
Department of Restorative Dental Sciences, Division of Operative Dentistry, College of Dentistry,
Jazan University, Saudi Arabia

Received 2 June 2020; Final revision received 19 June 2020

Available online 30 June 2020

KEYWORDS Abstract Background/purpose: Artificial intelligence (AI) has made deep inroads into
Artificial intelligence dentistry in the last few years. The aim of this systematic review was to identify the develop-
dentistry; ment of AI applications that are widely employed in dentistry and evaluate their performance
Machine learning; in terms of diagnosis, clinical decision-making, and predicting the prognosis of the treatment.

* Corresponding author. Preventive Dental Science Department, College of Dentistry, King Saud Bin Abdulaziz University for Health Sci-
ences, King Abdullah International Medical Research Center, Ministry of National Guard Heath Affairs, Riyadh, Saudi Arabia.
** Corresponding author. Department of Preventive Dental Sciences, Division of Pedodontics, College of Dentistry, Jazan University, Jazan,
Saudi Arabia.
E-mail addresses: [email protected] (S.B. Khanagar), [email protected] (P.C. Maganur).

https://doi.org/10.1016/j.jds.2020.06.019
1991-7902/ª 2020 Association for Dental Sciences of the Republic of China. Publishing services by Elsevier B.V. This is an open access article under
the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Artificial intelligence in dentistry 509

Materials and methods: The literature for this paper was identified and selected by perform-
Computer-aided
ing a thorough search in the electronic data bases like PubMed, Medline, Embase, Cochrane,
diagnosis;
Google scholar, Scopus, Web of science, and Saudi digital library published over the past two
Deep learning
decades (January 2000eMarch 15, 2020).After applying inclusion and exclusion criteria, 43
models;
articles were read in full and critically analyzed. Quality analysis was performed using QUA-
Convolutional neural
DAS-2.
networks;
Results: AI technologies are widely implemented in a wide range of dentistry specialties.
Artificial neural
Most of the documented work is focused on AI models that rely on convolutional neural net-
networks
works (CNNs) and artificial neural networks (ANNs). These AI models have been used in detec-
tion and diagnosis of dental caries, vertical root fractures, apical lesions, salivary gland
diseases, maxillary sinusitis, maxillofacial cysts, cervical lymph nodes metastasis, osteopo-
rosis, cancerous lesions, alveolar bone loss, predicting orthodontic extractions, need for or-
thodontic treatments, cephalometric analysis, age and gender determination.
Conclusion: These studies indicate that the performance of an AI based automated system is
excellent. They mimic the precision and accuracy of trained specialists, in some studies it
was found that these systems were even able to outmatch dental specialists in terms of per-
formance and accuracy.
ª 2020 Association for Dental Sciences of the Republic of China. Publishing services by Else-
vier B.V. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

Introduction learn from data so they can resolve issues without

human input.
The exponential growth in science and technology has ⁃ Neural networks are a set of algorithms that compute
introduced different applications that are used daily, such signals via artificial neurons. The purpose of neural
as siri and alexa. These applications are found on the top of networks is to create neural networks that function like
artificial intelligence (AI) and its components. The term AI the human brain.
is mostly associated with robotics. It describes how tech- ⁃ Deep learning is a component of machine learning that
nology is used to develop a software or a machine that can utilizes the network with different computational layers
easily mimic human intelligence and perform specific in a deep neural network to analyze the input data. The
activities. purpose of deep learning is to construct a neural
John McCarthy, a mathematician coined the term arti- network that automatically identifies patterns to
ficial intelligence in 1955, and widely recognized as the improve feature detection.4
father of artificial intelligence. He chose this term to
explain the potential of machines to perform tasks that can Deep learning is also known as convolutional neural
fall in the range of “intelligent” activities.1 networks. They collect features from the abstracted layer
In the year 1956, John McCarthy organized a famous of filters and are primarily used to process large and com-
Dartmouth conference, which was formally on the research plex images.5
project artificial intelligence, this is when this discipline
originated. The conference sparked the crucial period,
from 1950s to 1970s, wherein extensive research was done
on AI.2
In 1978 Richard Bellman, an applied mathematician
defined artificial intelligence as the automation of activ-
ities associated with human thinking abilities, which in-
cludes learning, decision making and problem solving.3
In the modern day world, artificial intelligence refers to
any machine or technology that is able to mimic human
cognitive skills like problem solving. To understand AI, it is
important to know few of these key aspects (Fig. 1).

⁃ Artificial intelligence is termed as a capability of ma-

chines that exhibits a form of its own intelligence. The
aim here was to develop machines that can learn
through data so that they can solve the problems.
⁃ Machine learning is part of AI, which depends on algo-
rithms to predict outcomes based on a dataset. The Figure 1 Key aspects of artificial intelligence.
purpose of machine learning is to facilitate machines to
510 S.B. Khanagar et al

Application of AI in dentistry
Table 1 Description of the PICO (PZ Population, IZ
Intervention, CZ Comparison, OZ Outcome) elements.
AI has been used mainly in dentistry to make the process of
diagnosis more accurate and efficient, which is of atmost Research What are the developments, performance, and
importance in achieving best results to the treatments question application of artificial intelligence in
provided along with superior quality patient care. dentistry?
Dentists need to use all their acquired knowledge to Population Patients diagnostic images related to oral and
diagnose and decide the best treatment option. They are maxillofacial regions [clinical images,
also required to predict the prognosis where they need radiographs, CBCT, confocal laser
accurate clinical decision-making skills. However, in some endomicroscopy (CLE) Images, intraoral
cases, dentists do not have enough knowledge to make the fluorescence images, cephalometric
right clinical decision in a limited period. AI applications radiographs, near-infrared-light
can serve as their guide so that they can make better de- transillumination (NILT) images]
cisions and perform better. Intervention AI based models for diagnosis, treatment
Shortliffe,6 Chae et al.,7 Schleyer et al.,8 reported that planning, clinical decision making, predicting
dentists have become dependent on computer applications the need for treatment, and predicting the
to get insights for clinical decision making. The aim of this prognosis.
systematic review was to identify the development of AI Comparison Expert opinions, reference standards
applications that are widely employed in dentistry and to Outcome Measurable or predictive outcomes such as
evaluate their performance in terms of diagnosis, clinical accuracy, sensitivity, specificity, ROC Z
decision-making, and predicting the prognosis of the receiver operating characteristic curve, AUC Z
treatment. Area Under the Curve, ICC Z Intraclass
Correlation Coefficient, Positive/Negative
Predictive Values (PPV/NPV)

Materials and methods

Criteria for considering studies for this review
Data sources
Inclusion criteria
This systematic review was carried after referring the
guidelines for preferred reporting items for Systematic re-
1. The article must be focused on AI and its application
views and Meta-analyses extension for Diagnostic Test Ac-
should be related to dentistry.
curacy (PRISMA-DTA).9 The literature for this paper was
2. There must be some predictive or measurable outcomes
identified and selected by performing a thorough search in
so they can be quantified.
the electronic data bases like Pubmed, Medline, Embase,
3. There has to be a proper mention of datasets that are
Cochrane, Google scholar, Scopus, Web of science, and
used to assess a model.
Saudi digital library published over the past two decades
(January 2000eMarch 15, 2020) by using keywords such as
Exclusion criteria
artificial intelligence in dentistry, deep learning, machine
learning, artificial neural networks, convolutional neural
networks, and computer-aided diagnosis. This search was 1. The articles that are related to non-AI areas.
based on the PICO (problem/patient/population, inter- 2. Uistaploaded articles that were unpublished.
vention/indicator, comparison, and outcome) elements 3. Articles that consisted of only abstracts without the full
(Table 1). text.
4. Articles that were not written in English.

This criteria cut down the number of articles to 46. The

Resources selection name of the journals and names of the authors were hidden
and spread among the authors. A critical assessment was
Full-length articles were retrieved. Hand searching and carried out for all the articles by following QUADAS-2
electronic searching was performed to go through the (Quality Assessment and Diagnostic Accuracy Tool) guide-
journals. The required data for this review was lines, a tool for quality assessment of the studies on diag-
selected in two stages. In the first stage the articles nostic accuracy.10 3 more articles were excluded due to
were selected based on the title and abstracts related disagreement from the authors. Eventually, this systematic
to our research topic. The preliminary search resulted review performed qualitative synthesis on 43 articles
in 1268 articles that were appropriate enough to (Fig. 2).
address the paper’s aim. Due to duplication, 182 arti- All the articles were read fully. The years of these ar-
cles were removed. Hence, we retrieved 1086 articles ticles were taken in account to study the progress of AI
for the second stage of selection. Next, the following trends that were developed and evolved over the years in
criterion was applied. dentistry.
Artificial intelligence in dentistry 511

Figure 2 Flow chart for screening and selection of articles.

Figure 3 Trends of research on artificial intelligence in dentistry.

512 S.B. Khanagar et al

Results studies conducted on the cadaveric samples and extracted

specimens, under the applicability domain were considered
This systematic review included 43 research articles that as high risk for patient selection and index test domain
were analyzed for quantitative data. The analysis of the (Supplementary Table 2). Even though, comparable results
literature revealed that most of the studies were con- were obtained for the applicability arm of the QUADAS-2
ducted in the past 12 years (Fig. 3). The trends showed that (Figs. 4 and 5).
there is a gradual increase in the research related to arti-
ficial intelligence in dentistry. Discussion
The studies that were included in this systematic review
were mainly on the application of AI for detection and AI is modernizing the traditional aspects of dentistry. AI
diagnosis of dental caries, diagnosis of proximal dental based systems are often used for designing automated
caries, tooth detection and numbering, detection of verti- software programs that streamlines the diagnosis and data
cal root fractures, detection of apical lesions, locating management in dentistry.8 Mostly they are clinical decision
minor apical foramen, assessment of root morphology, support systems that assist and guide experts to make
diagnosis of salivary gland diseases, diagnosis of maxillary better decisions. These systems have been used for better
sinusitis, maxillofacial cysts, cervical lymph nodes metas- diagnosis, treatment planning and also for predicting the
tasis, detection of osteoporosis, predicting diagnosis of prognosis.54 The demand for these systems is booming due
orthodontic extractions, assessing need for orthodontic to their effectiveness in providing explanations and
treatments, determining the growth and development of reasoning.55
cervical vertebrae stages, cephalometric analysis, identi- AI has revolutionized in the field of dentistry and making
fying cephalometric landmarks, diagnosis of orthognathic the dentist’s task easier. The clinical decision support sys-
surgeries, assessing the impact of orthodontic treatment on tems that work on the AI technology are mainly designed to
the facial attractiveness, gender determination using provide expert support to the health professionals.56 Clin-
mandibular morphometric parameters, estimating the age ical decision support systems is defined as, any computer
based on third molar development, classifying the program that has been designed to help health pro-
cancerous tissues, predicting postoperative facial swelling fessionals in making clinical decisions, and also deals with
following extractions, detection of periodontal bone loss, the medical data or with the knowledge of medicine
detecting the degree of alveolar bone loss. necessary for interpreting such data.6
Most of the studies have used convolutional neural net- In this systematic review we analyzed studies on the
works (CNNs) and artificial neural networks (ANNs). Some of application of AI technology in dentistry and evaluated
the studies were based on bayesian network (BN) and their performance in terms of diagnosis, clinical decision-
probabilistic neural networks (PNNS). The purpose of these making, and predicting the prognosis of the treatment.
neural networks is to determine computer tomography (CT)
images, cone beam computed tomography (CBCT) images,
Application of AI technologies in the specialty of
lateral cephalometric radiographs, bitewing radiographs,
facial photos, panoramic radiographs (OPG). (Table 2). oral and maxillofacial radiology and diagnostics
The studies that have been analyzed in this systematic
review revealed that the AI technology has been widely Zhank et al.,25 reported the use of AI-based CNNs and
used in different specialties of dentistry. AI technology has evaluated effective teeth recognition by relying on the
been applied in 18 studies on oral & maxillofacial label tree along with cascade network structure. The model
radiology,11,22,25,26,32,34,36e41,44e48,51 11 studies on ortho- demonstrated a high precision of 95.8%. A study by Tuzoff
dontics and dentofacial orthopedics,12,15,24,28e31,42,49,50,53 5 et al.,36 showed very similar results when utilized an AI-
studies on endodontics,13,14,16,33,35 4 studies on powered CNNs model for teeth identification, which was
periodontics,19,23,27,43 2 studies from oral and maxillofacial then arranged numerically. This computer-aided diagnostic
surgery18,21 and 3 studies from forensic odontology.17,52,20 technique displayed a mean sensitivity of 0.987 and preci-
sion 0.9945. This output was similar to that of an expert.
Chen et al.,39 applied the CNNs to detect the teeth
Risk of bias assessment number in intra oral periapical films and then to identify
the tooth. The model demonstrated very high precision.
All the studies were based on assessing the diagnostic ac- The results indicated that AI technologies make it conve-
curacy of AI in various fields of dentistry. Hence, risk of bias nient for clinicians to do their job. They do not have to
assessment was assessed using QUADAS-2 assessment tool, enter the details manually. Using these automated systems
which is frequently used tool in the literature dentists can enter their dental charts digitally, resulting in
(Supplementary Table 1). The studies conducted on human higher efficiency.
beings for determining the reference standard were rated AI technology in the detection of dental caries has
as high risk. In the present analysis, 55.81% of studies re- demonstrated excellent results which were reflected in the
ported high risk of bias for the reference standard. Since study done by Lee et al.,26 who reported applying of CNN
the data feeding in AI technology was highly standardized, algorithms for detection and diagnosis of dental caries on
AI had no effect on the flow, and time frame in the final periapical radiographs. The result of the application
output, hence categorized as low risk categories. In the demonstrated considerably good performance. Similar re-
present systematic review, low risk of bias was reported in sults were seen in the study done by Casalegno et al.,32 who
index test (81%) and (74.41%) in flow and timings. Some used deep learning model designed for the detection and
 Artificial intelligence in dentistry
Table 2 Details of the studies that have used AI based models in various specialties of dentistry for diagnosis, treatment planning, clinical decision making, predicting the need
for treatment, and predicting the prognosis.
Serial Authors Year of Algorithm Objective of the study No. of Study factor Modality Evaluation accuracy/ Comparison if any Results Outcomes Authors suggestions/
no publication Architecture images/ average accuracy (þ)effective, recommendations
photographs ()non effective
for testing (N) neutral

1 Devito et al.11 2008 ANNs AI based model for 160 Tooth Decay Bite-wing ROC curve area of 25 examiners (þ) Effective This neural network could None
diagnosing the proximal Radiographs 0.884 improve the performance of
dental caries diagnosing proximal caries.
2 Xie et al.12 2010 ANNs ANN based AI model for 200 Tooth Lateral Accuracy of 80% Not mentioned (þ)Effective ANN was effective in None
deciding if extractions are malocclusion cephalometric determining whether
necessary prior to radiographs extraction or non-extraction
orthodontic treatment treatment was best for
malocclusion patients
3 Saghiri et al.13 2012 ANNs ANN based AI model for 50 Tooth Human cadavers Accuracy of 96% 2 Endodontists (þ)Effective The accuracy of ANN was more The ANN model is
determining the working than the endodontists accurate method for
length working length
determination
4 Saghiri et al.14 2012 ANNs ANN system for locating the 50 Tooth Human dried skull Accuracy of 93% Endodontists (þ)Effective ANN can useful for secondary ANN can be used for
minor apical foramen (AF) opinion for locating the AF on decision making
radiographs and it can be similar clinical
helpful in enhancing the scenarios
accuracy in determining the
working length
5 Jung et al.15 2016 ANNs Artificial Intelligence expert 156 Tooth Lateral Accuracy of 92% 1 Experienced (þ)Effective The success rates of the AI expert systems with
system for orthodontic malocclusion cephalometric orthodontists models were 92% for the neural network
decision-making of required radiographs system’s recommendations for machine learning
permanent tooth extraction extraction vs non extraction could be useful in
orthodontics
6 Johari et al.16 2017 PNNs Probabilistic Neural Network 240 Tooth CBCT and periapical Accuracy of 96.6, Not mentioned (þ)Effective The designed neural network None
(PNN) for diagnosing (VRFs) radiographs sensitivity of 93.3 can be used as a proper model
in intact and the teeth that and specificity of for the diagnosis of VRFs on
has undergone endodontic 100% CBCT images of endodontically
treatment treated and intact teeth; CBCT
images were more effective
than periapical radiographs.
7 Tobel et al.17 2017 CNNs An automated technique for 200 Tooth Panoramic Mean ICC was 0.95 2 observers (þ)Effective Deep CNN based AI system Further optimization
staging the development of radiographs (OPG) demonstrated similar results to is required to achieve
lower third molar. the results demonstrated by a fully automated
other trained examiners. system for estimating
the dental age.
8 Aubreville et al.18 2017 CNNs AI based automatic system 7894 Oral cavity Confocal laser AUC of 0.96 and a Not clear (þ)Effective This approach was found to None
for diagnosing (OSCC) oral endomicroscopy mean accuracy of outperform the state of the art
squamous cell carcinoma (CLE) images 88.3%, sensitivity in CLE image recognition
86.6%, specificity 90%
9 Imangaliyev 2017 CNNs CNN model for the automatic 427 Tooth Quantitative light- Predictive accuracy Reference models (þ)Effective CNN model prediction None
et al.19 classification of red induced of 0.89% performance was higher than
fluorescent dental plaque fluorescence other models.
images. images
10 Niño-Sandoval 2017 ANNs AI based model for 229 Anatomical Lateral Coefficients from Support vector (þ)Effective This model demonstrated high This model may be
et al.20 predicting the mandibular landmarks cephalograms 0.84 until 0.99 regression predictability ability the key for facial
morphology reconstruction
(continued on next page)

513
 514
Table 2 (continued )
Serial Authors Year of Algorithm Objective of the study No. of Study factor Modality Evaluation accuracy/ Comparison if any Results Outcomes Authors suggestions/
no publication Architecture images/ average accuracy (þ)effective, recommendations
photographs ()non effective
for testing (N) neutral

11 Zhang et al.21 2018 ANNs ANN for predicting 100 Face Data set Accuracy of 98.00% 1 Oral surgeon (þ)Effective This AI based model proved to None
postoperative facial swelling be an accurate in predicting of
following the extraction of the facial
impacted mandibular third swelling following the
molars. extraction of impacted
mandibular third molars.
12 Lee et al.22 2018 CNNs (DCNN)-Based Computer- 200 Face Panoramic (AUC) values 2 Experienced oral (þ)Effective The system that was based on None
Assisted Diagnosis (CAD) radiographs (OPG) obtained using SC- and maxillofacial DCNN was effective in
systems. Single-Column DCNN was 0.9763, radiologists detecting osteoporosis and also
DCNN (SC-DCNN), Single- SC-DCNN (Augment) demonstrated high agreement
Column with Data was 0.9991, MC- with the experienced oral and
Augmentation DCNN (SC- DCNN for 0.9987 maxillofacial radiologists
DCNN Augment) and
Multicolumn DCNN (MC-
DCNN).
13 Lee et al.23 2018 CNNs Diagnosing and predicting of 348 Tooth Intra oral periapical Mean predictive 3 calibrated board- (þ)Effective The DCNN based model was Further optimization
PCT using a computer radiographs accuracy of 78.9% certified effective and efficient in of the PCT dataset is
-assisted detection system periodontists diagnosing and predicting of required for
based on a deep CNN (PCT). improvement
14 Thanathornwong24 2018 Bayesian Bayesian network (BN) for 1000 Tooth Data sets AUC (0.91) 2 Experienced (þ)Effective This BN based system; and None
network predicting the need for malocclusion orthodontists demonstrated promising
(BNs) orthodontic treatment. results with high degree of
accuracy in the need for
orthodontic treatment.
15 Zhang et al.25 2018 CNNs Teeth recognition using label 200 Tooth Intra oral periapical Precision of 95.8% Reference models (þ)Effective This approach demonstrated a None
tree with cascade network radiographs high precision of 95.8% and
structure. recall of 96.1%.
16 Lee et al.26 2018 CNNs AI based deep learning 600 Dental caries Intraoral periapical Mean AUC of 0.890 4 calibrated board- (þ)Effective This DCNN based system None
system for detecting and radiographic images certified dentists algorithm performed
diagnosing dental caries considerably good in detection
dental caries on periapical
radiographs.
17 Yauney et al.27 2018 CNNs AI based system for 810 Periodontium Intraoral AUC of 0.677, Dentists (þ)Effective This automated process was Machine learning, can
automated oral health fluorescence precision of 0.271, effective in correlating poor be used for
screenings and cross images Recall of 0.429 periodontal health with automated diagnoses
correlations of oral-systemic systemic health outcomes and systemic health
health screenings for other
diseases
18 Kök et al.28 2019 ANNs AI algorithms for 300 Cervical Cephalometric Mean accuracy of 1 orthodontists (þ)Effective ANN could be the preferred None
determining the stages of Vertebrae radiographs 77.02% method for determining

S.B. Khanagar et al
the growth and development cervical vertebrae stages
by cervical vertebrae
29
19 Park JH et al. 2019 CNNs Comparing latest deep-CNN 283 Landmarks Cephalometric 5% higher accuracy Single shot multibox (þ)Effective You-Only-Look-Once model This model can be
based systems for identifying radiographs with (YOLOv3) than detector (SSD) outperformed in accuracy and used in clinical
cephalometric landmarks Single (SSD) computational time than the practice for
shot multibox detector identifying the
cephalometric
landmarks.
 Artificial intelligence in dentistry
20 Choi et al.30 2019 ANNs ANN based model for 316 Landmarks Lateral ICC 0.97e0.99 1 Experienced (þ)Effective This ANN based model This ANN based model
deciding on surgery/non- cephalometric orthodontists demonstrated higher success will be useful in
surgery and determining radiographs rate in deciding on surgery/ diagnosing of
extractions. non-surgery and was also orthognathic surgery
successful in deciding on the cases.
extractions.
21 Patcas. et al.31 2019 CNNs AI system for describing the 2164 Facial Facial photographs Not Clear Not mentioned (þ)Effective This CNN based AI system can None
impact of orthognathic landmarks be used for scoring facial
treatments on facial attractiveness and apparent
attractiveness and age age in patients under
appearance orthognathic treatments.
22 Casalegno et al.32 2019 CNNs AI based model for detecting 217 Dental caries Near-infrared ROC of 83.6 for Dental experts with (þ)Effective This CNN based model None
and localizing dental lesions transillumination occlusal and ROC of clinical experience demonstrated promising
in Near-Infrared (TI) imaging 84.6% for proximal results with increased speed
Transillumination (TI) images and accuracy in detecting
caries.
23 Fukuda et al.33 2019 CNNs CNN based AI system for 60 Tooth Panoramic Precision of 0.93 2 radiologists and 1 (þ)Effective The CNN based AI model is an None
detection of vertical root radiographs (OPG) Recall was 0.75 endodontist efficient tool in detecting VRFs
fracture (VRF) F- Measure of 0.83.
24 Kise et al.34 2019 CNNs AI system for detection of 100 Salivary glands Computed Accuracy of 96.0, 6 radiologists (þ)Effective The deep learning system Can be used as a
Sjögren’s syndrome (SjS) on tomography (CT) Sensitivity of 100% demonstrated a higher diagnostic support
CT, and comparing its images and specificity of diagnostic performance while interpreting CT
performance with 92.0% images
radiologists
25 Hiraiwa et al.35 2019 CNNs AI system for classifying root 760 Tooth Cone beam Accuracy of 86.9% 2 radiologists (þ)Effective The deep learning system None
morphologies of mandibular computed demonstrated high accuracy in
first molars tomography (CBCT) the differential diagnosis of a
Images single or extra root in the distal
roots of mandibular first
molars.
26 Tuzoff et al.36 2019 CNNs CNN based AI system for 222 Tooth Panoramic Precision of 0.9945 Dental experts (þ)Effective The performance of the this This system can
automatic teeth detection radiographs (OPG) and mean sensitivity system was comparable to the simplify the process of
and numbering of 0.987 level of performance of the filling digital dental
experts charts.
27 Ekert et al.37 2019 CNNs CNNs based AI system for 2001 Tooth Panoramic AUC of 0.85 (0.04) 6 Dentists (þ)Effective This deep CNN based AI system None
detecting apical lesions (ALs) radiographs (OPG) sensitivity 0.65 and was successful in detecting
specificity 0.87 apical lesions
28 Murata et al.38 2019 CNNs AI based system for 120 Maxillary Panoramic Accuracy of 87.5%, 2 experienced (þ)Effective The AI based deep learning The deep-learning
diagnosing of maxillary sinusitis radiographs (OPG) sensitivity of 86.7%, radiologists, 2 dental system demonstrated higher system can provide
sinusitis specificity of 88.3%, residents. diagnostic performance. diagnostic support for
AUC of 0.875 inexperienced
dentists
29 Chen et al.39 2019 CNNs CNN based tool package for 250 Tooth Intra oral periapical Precisions and recalls 3 Dentists (þ)Effective The results indicate that None
detecting and numbering the films exceed 90%, machines performance was
teeth IOU of 91% close to the level of a junior
dentist
30 Vinayahalingam 2019 CNNs CNN based AI system to 81 Tooth Panoramic Mean dice- Portable network (þ)Effective Deep-learning is an Further enhancement
et al.40 detect and segment the radiographs (OPG) coefficients for M3s graphics (PNG) files encouraging approach to of the algorithm is
approximate of inferior and IAN were 0.947 as gold standard segment anatomical structures advised to improve
alveolar nerve (IAN) to the  0.033 and 0.847  the accuracy
roots of lower third molars 0.099
(M3) on OPGs
(continued on next page)

515
 516
Table 2 (continued )
Serial Authors Year of Algorithm Objective of the study No. of Study factor Modality Evaluation accuracy/ Comparison if any Results Outcomes Authors suggestions/
no publication Architecture images/ average accuracy (þ)effective, recommendations
photographs ()non effective
for testing (N) neutral

31 Mallishery et al.41 2019 ANNs Machine learning to generate 500 Tooth Data Set Sensitivity of 94.96% 2 pre-calibrated (þ)Effective This study provides an option An AAE endodontic
an algorithm which can help endodontists for automation for increasing case difficulty
predict the difficulty level of the speed of decision-making assessment form
the case and decide on and referrals. when utilized along
referral with machine learning
can assist general
dentists in rapid
assessment of the
case difficulty
32 Patcas et al.42 2019 CNNs AI system for evaluating the 30 Face Frontal and profile Cleft cases (all Ps  15 laypeople, 14 ()Non Effective AI system scores were There is a need for
facial attractiveness of images 0.19), for control orthodontists, and 10 comparable with the scores of further refinement in
patients who have group (all Ps  0.02) oral surgeons the other groups for the cleft this AI based system
undergone treatment for patients, but the scores were
clefts and the facial lower for the controls
attractiveness of controls
and to compare these results
with panel ratings performed
by laypeople, orthodontists,
and oral surgeons
33 Krois et al.43 2019 CNNs Deep- CNN based system for 2001 Periodontium Panoramic Predictive accuracy 6 Experienced (N) Neutral CNN demonstrated similar Machine -learning
detecting periodontal bone radiographs (OPG) of 81% and were dentists results to that of the dentists in based technologies
loss similar to the detecting periodontal bone can reduce the
examiners loss. dentists’ diagnostic
efforts.
34 Ariji et al.44 2019 CNNs AI system for diagnosing 441 Cervical Computed Accuracy of 78.2%, Not clear (N) Neutral The diagnostic results of the This CNN based
metastasis of lymph node. lymph nodes tomography (CT) sensitivity of 75.4%, CNN based system were similar system is a valuable
images specificity of 81.0%, to the results of the for diagnostic
positive predictive radiologists. support.
value of 79.9%,
negative predictive
value of 77.1%, and
ROC of 0.80
35 Ariji et al.45 2019 CNNs Performance of deep 703 Cervical Computed Accuracy of 84.0% 4 Radiologists (þ)Effective The deep learning diagnostic This method is
learning classification in lymph nodes tomography (CT) performance in extra nodal expected to improve
diagnosing extranodal images extension was significantly diagnostic accuracy
extension of cervical lymph higher when compared with by further study with
node metastases in CT the performance of the increasing sample size
images radiologists of patients.
36 Hung et al.46 2019 CNNs AI based model for 5135 Root caries Data set Accuracy of 97.1%, Trained medical (þ)Effective This model perform well and Can be utilized by
predicting root caries precision of 95.1%, personnel can be allowed for clinical both dental and non-

S.B. Khanagar et al
sensitivity of 99.6% implementation dental professionals
and specificity of
94.3%
AUC of 0.997
37 Kim et al.47 2019 CNNs AI based (CNNs) for 200 Maxillary Waters’ view AUC of 0.93 for the 5 Radiologists (þ)Effective AI based (CNNs) demonstrated None
diagnosing maxillary sinusitis sinusitis radiographs temporal and 0.88 statistically significantly higher
for geographic AUC than radiologist in both
external test sets
 Artificial intelligence in dentistry
38 Schwendicke 2020 CNNs AI based (CNNs) to detect 226 Tooth decay NILT images The mean) AUC of 2 Experienced (þ)Effective The model demonstrated None
et al.48 caries lesions in near- 0.74, Sensitivity of dentists satisfying discriminatory ability
infrared-light 0.59 and specificity to detect caries lesions.
transillumination (NILT) of 0.76
images. PPV was 0.63 and
NPV was 0.73
39 Kunz et al.49 2020 CNNs An automated cephalometric 50 Landmarks Cephalometric Not clear 12 experienced (þ)Effective AI algorithm was able to None
X-ray analysis using a radiographs examiners analyze unknown
specialized (AI) algorithm cephalometric X-rays similar to
the quality level of the
experienced human examiners
40 Hwang et al.50 2020 CNNs Deep -learning based 283 Landmarks Cephalometric Not Mentioned Human examiners (þ)Effective This system accuracy in This system might be
automated system for radiographs identifying of cephalometric a viable option when
detecting the patterns of 80 landmarks similar to the human repeated
cephalometric landmarks examiners identification of
multiple
cephalometric
landmarks.
41 Lee et al.51 2020 CNNs Deep (CNNs), on the 136 Face Dental panoramic ROC of 0.858 Gold standard (þ)Effective This Deep (CNNs), could of use None
classification of specific radiographs (DPRs) reference models and reliable system for
features of osteoporosis automated screening of
osteoporosis patients.
52
42 Patil et al. 2020 ANNs ANN for gender 509 Mandible Panoramic Accuracy of 75% 1 experienced oral (þ)Effective ANN proved as a good tool for This automated
determination radiographs (OPG) and maxillofacial predicting the gender and can application is
radiologist be applied in the forensic promising for
sciences for near accurate identifying gender or
results. age with minimal
errors
43 Yu et al.53 2020 CNNs AI based skeletal diagnostic 5890 Anatomical Lateral Mean AUC of >95% 2 orthodontists (þ)Effective This model demonstrated None
system landmarks cephalograms excellent performance for
skeletal orthodontic diagnosis

ANNs Z Artificial Neural Networks, CNNs Z Convolutional Neural Networks, DCNNs Z Deep Convolutional Neural Networks, BN Z Bayesian Network, PNN Z Probabilistic Neural Network,
ROC Z Receiver Operating Characteristic curve, AUC Z Area Under the Curve, ICC Z Intraclass Correlation Coefficient, F Z F- measure, VRF Z Vertical Root Fracture, PTC Z Periodontal
Compromised Teeth, Positive/Negative Predictive Values (PPV/NPV).

517
518 S.B. Khanagar et al

Figure 4 Assessment of individual risk of bias domains.

Figure 5 Concerns regarding applicability.

localization of dental lesions in near-infrared trans- Hiraiwa et al.,34 reported of applying CNNs for detection
illumination (TI) images which revealed promising results. of sjögren’s syndrome (SjS) on CT images and compared the
Schwendicke et al.,48 reported using of near-infrared-light results with the performance of radiologists and showed a
transillumination (NILT) images for diagnosing dental higher diagnostic performance. In another study by Murata
caries and showed that the performance of this AI based et al.,38 the authors applied the deep learning system for
models was satisfactory. diagnosing maxillary sinusitis on panoramic radiography.
Devito et al.,11 used an AI based ANN model for diag- The diagnostic performance of this system was sufficiently
nosing proximal caries using the bitewing radiographs and high. These results were similar to the study conducted by
found a quite encouraging results. Study by Hung et al.,46 kim et al.,47 when compared the performance with expe-
reported AI technology for predicting root caries and rienced radiologists.
demonstrated excellent results. Ekert et al.,37 was suc- Ariji et al.,45 applied CNN system for deep learning
cessful in detecting apical lesion when they applied CNNs to image classification for diagnosing lymph node metastasis
detect apical lesions (ALs) on panoramic dental on (CT) Images and showed higher diagnosing accuracy,
radiographs. sensitivity, and specificity. These results were similar to the
Artificial intelligence in dentistry 519

study conducted by Hung et al.,46 who assessed the per- esthetics may further deteriorate creating a major prob-
formance of deep learning classification in diagnosing extra lem. This suggests that diagnosis is an important aspect for
nodal extension of cervical lymph node metastases in CT the dentist to analyze the problems of the patient accu-
images. Both the studies showed similar performance or rately. AI technology is striving to make dentists job much
even higher performance when compared to professional accurate and precise. Choi et al.,30 reported the use of new
radiologists. artificial intelligence model to decide the case for surgery/
Lee et al.,22 evaluated the efficiency and performance non-surgery using the lateral cephalometric radiographs.
of AI in diagnosis and detection of osteoporosis. In this He showed that the system was very effective with 96%
study deep convolutional neural network (DCNN) based success rate in diagnosing the surgery/non-surgery cases.
computer-assisted diagnosis (CAD) systems was applied for This model has shown promising results, hence can be
detection of osteoporosis, using panoramic radiographs and applied for the diagnosis of orthognathic surgery cases.
exhibited very promising results. This was well above in par Hagg et al.,59 stated that determining the chronological
with experienced oral and maxillofacial radiologists in age of the patient is not alone sufficient for estimating the
detecting osteoporosis. These results were similar to the actual growth time, hence various skeletal maturation in-
results done by Lee et al.,51 who used deep convolutional dicators have been developed for this task. Determination
neural networks (DCNN) for detecting osteoporosis in of the growth and development, and estimation of the
dental panoramic radiographs. skeletal maturation stages have been used to predict the
time of pubertal development, determining the growth rate
and for mainly estimating the remaining growth and
Application of AI technologies in the specialty of development potential of an individual as mentioned by
orthodontics and dentofacial orthopedics Flores-mir et al.60 These are usually determined by using
hand-wrist radiographs, cephalometric analysis, and with
Accurate diagnosis, treatment planning and prediction of the help of maturation stages of cervical vertebra. AI
prognosis are the key factors for successful orthodontic technology has also been applied for determining the
treatment. AI technology has been applied for deciding if growth and development by cervical vertebrae stages. Kok
extractions are necessary prior to the orthodontic treat- H et al.,28 showed a mean accuracy of 77.02%, using arti-
ment. In a study by Xie et al.,12 artificial neural network ficial intelligence algorithms for determining the growth
(ANN) model was applied for deciding if extractions are and development by cervical vertebrae stages when
necessary using lateral cephalometric radiographs. The applied on the cephalometric radiographs.
results were quite promising. Jung et al.,15 showed 92%
accuracy using AI expert system for deciding on permanent Application of AI technologies in the specialty of
tooth extraction, using lateral cephalometric radiographs. endodontics
The results of both the studies are suggestive that the AI
modes were effective and accurate in predicting the need
The success of root canal treatment mainly depends on
for extraction. These models can be used as a tool for
accuracy of working length determination. The prognosis of
making decisions in clinical practice. High accuracy was
the treatment can only be ensured when instrumentation
seen in the study by Thanathornwong,24 who suggested AI
terminates at the apical constriction.61 Saghiri et al.13 used
model based on bayesian network (BN) for assessing the
artificial neural network (ANN) system in determining the
need for orthodontic treatment.
working length and showed exceptional accuracy of 96%
Various studies have been conducted to demonstrate AI
which is higher than the accuracy compared to professional
technologies and its application in identifying cephalo-
endodontists. These results were similar to the study by
metric landmarks. Park et al.,29 compared the efficiency
Saghiri et al.,14 where they used the ANN system for
and accuracy of the updated deep-learning algorithms for
locating the minor apical foramen, with an accuracy of 93%.
automatic identification of cephalometric landmarks using
In endodontics, AI is used to diagnose vertical root
cephalometric radiographs. The results revealed that the
fractures. A study employed by Johari et al.,16 who used
system was extremely accurate in the computation of the
probabilistic neural network (PNN) for the diagnosis of
landmarks. Studies conducted by Kunz et al.,49 and Hwang
vertical root fractures. This PNN system displayed excellent
et al.,50 showed excellent accuracy in identifying the
performance with an accuracy of 96.6%. Similarly con-
landmarks similar to the trained human examiners using a
volutional neural network in detecting vertical root frac-
specialized artificial intelligence (AI) algorithm and deep
ture, showing a highly encouraging precision was conducted
learning based automated identification system respec-
by Fukuda et al.33 These assessments indicate that AI-based
tively. Yu et al.53 demonstrated excellent results with
models are incredibly effective when it comes to the
automated skeletal classification with lateral cephalometry
detection of vertical root fractures on CBCT images and
based on the AI Model. The results of the above mentioned
panoramic radiographs.
studies indicates that, these systems prove to be a viable
option for repeatedly identifying multiple cephalometric
landmarks. Application of AI technologies in the specialty of
Establishing of accurate diagnosis and treatment plan- periodontics
ning in orthognathic surgery is the most important step for
the success of the treatment.57 Arnet et al.,58 in his liter- Periodontal diseases are one of the most common oral
ature on facial keys to orthodontic diagnosis and treatment diseases affecting the mankind. It is a known fact that this
planning suggested that if diagnosis is incorrect, the patient is one of the main reasons for the early loss of teeth. It is
520 S.B. Khanagar et al

well reported by Lee et al.,62 that continuous progression of compared to the trained examiners. Patil et al.,52 used
the disease will eventually lead to the loss of teeth in the ANNs to determine gender using panoramic radiographs,
adults. Various studies have been done to ascertain AI the results were quite promising. This system is very useful
technology application to diagnose and predict periodontal as it automates and eases the method of identifying un-
diseases. Lee et al.,23 reported use of CAD system, based on known gender or age with minimal errors. Niño-Sandoval
a deep convolutional neural network (CNN) algorithm for et al.,20 reported an AI model based on ANNs for predicting
diagnosing and predicting the teeth that are compromised the mandibular morphology and demonstrated promising
with periodontal health. The outcome were quite accept- results. Hence AI can be used effectively in forensic
able with a mean predictive accuracy of 78.9%. Yauney dentistry.
et al.,27 used an AI based system based on CNNs for
correlating poor periodontal health with systemic health
outcomes and reported that, AI can be used for automated
Uses of AI in dentistry based on the conclusions
diagnoses and can also be useful for screenings for other
from the articles reviewed in the paper
diseases. Krois et al.,43 used CNNs to detect periodontal
bone loss (PBL) on panoramic dental radiographs. The re-
sults of this study were similar to that of the expert opin- ⁃ AI systems can assist the clinicians so they can offer
ions. This system can still help in reducing the dentist’s high-quality dental care to their patients.
diagnostic efforts. ⁃ Dentists can use AI systems as an ancillary tool for
increasing the accuracy of diagnosis, treatment plan-
ning, and predicting the treatment outcomes.
Application of AI technologies in the specialty of ⁃ Non-specialty dentists can receive diagnostic support via
oral and maxillofacial surgery the deep-learning systems.
⁃ Automated systems can save a lot of time and increase
It is estimated that every year there are around 657,000 the efficiency of the clinicians (for e.g. automatic
new cases detected of cancers of the oral cavity and completion of electronic dental records by identifying
pharynx and is also a reason for 330,000 deaths as noted in the tooth and numbering).
the article of oral cancer by WHO.63 AI technology has been ⁃ The use of these systems for secondary opinions can
used for detecting cancers. There is revolutionary devel- improve the accuracy of diagnosis.
opment and refinement of convolutional neural networks ⁃ These systems provide a great deal of value for forensic
that have demonstrated improved ability for automated diagnosis.
cancer detection as seen in the study by Xu et al.64
Aubreville et al.,18 showed extremely positive and
promising results when employed CNNs for an automatic
approach for diagnosing Oral Squamous Cell Carcinoma Conclusion
when used with confocal laser endomicroscopy images. The
study indicated that AI model will be helpful for early AI has revolutionized dentistry in the last few years. Studies
diagnosis. AI technology has also been used for predicting show that these AI-powered automated systems performed
postoperative facial swelling after extraction of teeth. extremely well in various scenarios. Few authors found
Zhang et al.,21 used an artificial intelligence model based them to be more accurate than even dental specialists.
on ANN for predicting the postoperative facial swelling Although these outcomes do not make them better than the
following the extraction of impacted mandibular 3rd mo- dentists, they do establish that AI can be considered for
lars. The model demonstrated excellent results and will be clinical applications. These systems bring terrific value to
of great importance for clinicians for predicting the prog- the table by improving the accuracy of diagnosis, enhancing
nosis of the treatment. clinical decision-making, and predicting the treatment
prognosis which can help the clinicians in rendering best
quality care to their patients. There are also documented
Application of AI technologies in the forensic studies that have also reported that these automated sys-
odontology tems are of greater value for screening the patients for
osteoporosis, oral cancer, and metastasis of the lymph
Forensic odontology is relatively new, but it has made a nodes. This is a priceless benefit because it can help pro-
stellar contribution to the field of dentistry. A dentist plays fessionals to diagnose cases in the early stages, which in
an important role when they have to identify people for turn can save many lives. Although AI is widely used in
child abuse, crime, sexual assault, mass calamities, and various fields of dentistry, some specialties such as pedo-
other legal issues. Their moral duty compels them to pro- dontics and oral pathology still lack the development and
vide justice to the victims and their families, especially application of AI technology.
when there is no other evidence other than the dental re-
mains. AI technology has been applied in this field and has
shown excellent results.
De Tobel et al.,17 used automated technique based on Conflicts of interest
CNNs for staging lower third molar development for esti-
mating the age of a person after applying on panoramic The authors have no conflicts of interest relevant to this
radiographs. The system showed remarkable results, when article.
Artificial intelligence in dentistry 521

Acknowledgements on panoramic radiographs for age estimation: a pilot study. J

Forensic Odontostomatol 2017;35:42e54.
18. Aubreville M, Knipfer C, Oetter N, et al. Automatic Classifica-
This is a non-funded research and there has been no tion of cancerous tissue in laser endomicroscopy images of the
financial support for this work that could have influenced oral cavity using deep learning. Sci Rep 2017;7:11979.
its outcome. 19. Imangaliyev S, van der Veen MH, Volgenant CMC, Keijser BJF,
Crielaard W, Levin E. Deep learning for classification of dental
plaque images. Springer, 2016:407e10.
Appendix A. Supplementary data
20. Niño-Sandoval TC, Guevara Pérez SV, González FA, Jaque RA,
Infante-Contreras C. Use of automated learning techniques for
Supplementary data to this article can be found online at predicting mandibular morphology in skeletal class I, II and III.
https://doi.org/10.1016/j.jds.2020.06.019. Forensic Sci Int 2017;281:187.e1e7.
21. Zhang W, Li J, Li Z, et al. Predicting postoperative facial
swelling following impacted mandibular third molars extrac-
tion by using artificial neural networks evaluation. Sci Rep
References 2018;8:12281.
22. Lee JS, Adhikari S, Liu L, Jeong HG, Kim H, Yoon SJ. Osteo-
1. Rajaraman V. John McCarthy Father of artificial intelligence. porosis detection in panoramic radiographs using a deep con-
Reson. 2014:198e207. volutional neural network-based computer-assisted diagnosis
2. National Research Council. Funding a revolution: government system: a preliminary study. Dentomaxillofacial Radiol 2019;
support for computing research. Washington, DC: The National 48:20170344.
Academies Press, 1999:302. 23. Lee JH, Kim DH, Jeong SN, Choi SH. Diagnosis and prediction of
3. Bellman Richard. Artificial intelligence: can computers think? periodontally compromised teeth using a deep learning-based
Thomson Course Technology, 1978:146. convolutional neural network algorithm. J Periodontal
4. Jef Akst. A primer: artificial intelligence versus neural net- Implant Sci 2018;48:114e23.
works. Inspiring Innovation: The Scientist Exploring Life, 2019: 24. Thanathornwong B. Bayesian-based decision support system
65802. for assessing the needs for orthodontic treatment. Healthc
5. Rabuñal JR, Dorado J. Artificial neural networks in real-life Inform Res 2018;24:22e8.
applications. IGI Global: Hershey 2005:166e346. 25. Zhang K, Wu J, Chen H, Lyu P. An effective teeth recognition
6. Shortliffe EH. Testing reality: the introduction of decision- method using label tree with cascade network structure.
support technologies for physicians. Methods Inf Med 1989; Comput Med Imag Graph 2018;68:61e70.
28:1e5. 26. Lee JH, Kim DH, Jeong SN, Choi SH. Detection and diagnosis of
7. Chae YM, Yoo KB, Kim ES, Chae H. The adoption of electronic dental caries using a deep learning-based convolutional neural
medical records and decision support systems in Korea. network algorithm. J Dent 2018;77:106e11.
Healthc Inform Res 2011;17:172e7. 27. Yauney G, Rana A, Wong LC, Javia P, Muftu A, Shah P. Auto-
8. Schleyer TK, Thyvalikakath TP, Spallek H, Torres-Urquidy MH, mated process incorporating machine learning segmentation
Hernandez P, Yuhaniak J. Clinical computing in general and correlation of oral diseases with systemic health. EMBC
dentistry. J Am Med Inf Assoc 2006;13:344e52. 2019:3387e93.
28. _
Kök H, Acilar AM, Izgi MS. Usage and comparison of artificial
9. McGrath TA, Alabousi M, Skidmore B, et al. Recommendations
for reporting of systematic reviews and meta-analyses of intelligence algorithms for determination of growth and
diagnostic test accuracy: a systematic review. Syst Rev 2017;6: development by cervical vertebrae stages in orthodontics. Prog
194. Orthod 2019;20:41.
10. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a 29. Park JH, Hwang HW, Moon JH, et al. Automated identification
revised tool for the quality assessment of diagnostic accuracy of cephalometric landmarks: part 1-comparisons between the
studies. Ann Intern Med 2011;155:529e36. latest deep-learning methods YOLOV3 and SSD. Angle Orthod
11. Devito KL, De Souza Barbosa F, Felippe Filho WN. An artificial 2019;89:903e9.
multilayer perceptron neural network for diagnosis of proximal 30. Choi HI, Jung SK, Baek SH, et al. Artificial intelligent model
dental caries. Oral Surg Oral Med Oral Pathol Oral Radiol with neural network machine learning for the diagnosis of
Endod 2008;106:879e84. orthognathic surgery. J Craniofac Surg 2019;30:1986e9.
12. Xie X, Wang L, Wang A. Artificial neural network modeling for 31. Patcas R, Bernini DAJ, Volokitin A, Agustsson E, Rothe R,
deciding if extractions are necessary prior to orthodontic Timofte R. Applying artificial intelligence to assess the impact
treatment. Angle Orthod 2010;80:262e6. of orthognathic treatment on facial attractiveness and esti-
13. Saghiri MA, Asgar K, Boukani KK, et al. A new approach for mated age. Int J Oral Maxillofac Surg 2019;48:77e83.
locating the minor apical foramen using an artificial neural 32. Casalegno F, Newton T, Daher R, et al. Caries detection with
network. Int Endod J 2012;45:257e65. near-infrared transillumination using deep learning. J Dent Res
14. Saghiri MA, Garcia-Godoy F, Gutmann JL, Lotfi M, Asgar K. The 2019;98:1227e33.
reliability of artificial neural network in locating minor apical 33. Fukuda M, Inamoto K, Shibata N, et al. Evaluation of an arti-
foramen: a cadaver study. J Endod 2012;38:1130e4. ficial intelligence system for detecting vertical root fracture on
15. Jung SK, Kim TW. New approach for the diagnosis of extrac- panoramic radiography. Oral Radiol 2019. https:
tions with neural network machine learning. Am J Orthod //doi.org/10.1007/s11282-019-00409-x.
Dentofacial Orthop 2016;149:127e33. 34. Kise Y, Ikeda H, Fujii T, et al. Preliminary study on the appli-
16. Johari M, Esmaeili F, Andalib A, Garjani S, Saberkari H. cation of deep learning system to diagnosis of sjögren’s syn-
Detection of vertical root fractures in intact and endodonti- drome on CT images. Dentomaxillofacial Radiol 2019;48:
cally treated premolar teeth by designing a probabilistic neural 20190019.
network: an ex vivo study. Dentomaxillofacial Radiol 2017;46: 35. Hiraiwa T, Ariji Y, Fukuda M, et al. A deep-learning artificial
20160107. intelligence system for assessment of root morphology of the
17. De Tobel J, Radesh P, Vandermeulen D, Thevissen PW. An mandibular first molar on panoramic radiography. Dentomax-
automated technique to stage lower third molar development illofacial Radiol 2019;48:20180218.
522 S.B. Khanagar et al

36. Tuzoff DV, Tuzova LN, Bornstein MM, et al. Tooth detection cephalometric analysis using a customized convolutional neu-
and numbering in panoramic radiographs using convolutional ral network. J Orofac Orthop 2020;81:52e68.
neural networks. Dentomaxillofacial Radiol 2019;48: 50. Hwang HW, Park JH, Moon JH, et al. Automated identification
20180051. of cephalometric landmarks: Part 2-Might it be better than
37. Ekert T, Krois J, Meinhold L, et al. Deep learning for the human? Angle Orthod 2020;90:69e76.
radiographic detection of apical lesions. J Endod 2019;45: 51. Lee KS, Jung SK, Ryu JJ, Shin SW, Choi J. Evaluation of transfer
917e22. learning with deep convolutional neural networks for screening
38. Murata M, Ariji Y, Ohashi Y, et al. Deep-learning classification osteoporosis in dental panoramic radiographs. J Clin Med 2020;
using convolutional neural network for evaluation of maxillary 9:392.
sinusitis on panoramic radiography. Oral Radiol 2019;35: 52. Patil V, Vineetha R, Vatsa S, et al. Artificial neural network for
301e7. gender determination using mandibular morphometric param-
39. Chen H, Zhang K, Lyu P, et al. A deep learning approach to eters: a comparative retrospective study. Cogent Engineering
automatic teeth detection and numbering based on object 2020;7:1723783.
detection in dental periapical films. Sci Rep 2019;9:3840. 53. Yu HJ, Cho SR, Kim MJ, Kim WH, Kim JW, Choi J. Automated
40. Vinayahalingam S, Xi T, Bergé S, et al. Automated detection of skeletal classification with lateral cephalometry based on
third molars and mandibular nerve by deep learning. Sci Rep artificial intelligence. J Dent Res 2020;99:249e56.
2019;9007. 54. Mendonça EA. Clinical decision support systems: perspectives
41. Mallishery S, Chhatpar P, Banga KS, Shah T, Gupta P. The pre- in dentistry. J Dent Educ 2004;68:589e97.
cision of case difficulty and referral decisions: an innovative 55. Begum S, Ahmed MU, Funk P, Xiong N, Von Schéele B. A case-
automated approach. Clin Oral Invest 2020;24:1909e15. based decision support system for individual stress diagnosis
42. Patcas R, Timofte R, Volokitin A, et al. Facial attractiveness of using fuzzy similarity matching. Comput Intell 2009;25:
cleft patients: a direct comparison between artificial- 180e95.
intelligence-based scoring and conventional rater groups. Eur 56. Khanna S. Artificial intelligence: contemporary applications
J Orthod 2019;41:428e33. and future compass. Int Dent J 2010;60:269e72.
43. Krois J, Ekert T, Meinhold L, et al. Deep learning for the 57. Proffit WR, Fields HW, Sarver DM. Contemporary orthodontics,
radiographic detection of periodontal bone loss. Sci Rep 2019: 5th ed. St Louis: Mosby, 2013.
8495. 58. Arnett GW, Bergman RT. Facial keys to orthodontic diagnosis
44. Ariji Y, Fukuda M, Kise Y, et al. Contrast-enhanced computed and treatment planning. Part I. Am J Orthod Dentofacial
tomography image assessment of cervical lymph node metas- Orthop 1993;103:299e312.
tasis in patients with oral cancer by using a deep learning 59. Hägg U, Taranger J. Maturation indicators and the pubertal
system of artificial intelligence. Oral Surg Oral Med Oral Pathol growth spurt. Am J Orthod 1982;82:299e309.
Oral Radiol 2019;127:458e63. 60. Flores-Mir C, Nebbe B, Major PW. Use of skeletal maturation
45. Ariji Y, Sugita Y, Nagao T, et al. CT evaluation of extranodal based on hand-wrist radiographic analysis as a predictor of facial
extension of cervical lymph node metastases in patients with growth: a systematic review. Angle Orthod 2004;74:118e24.
oral squamous cell carcinoma using deep learning classifica- 61. Baugh D, Wallace J. The role of apical instrumentation in root
tion. Oral Radiol 2020;36:148e55. canal treatment: a review of the literature. J Endod 2005;31:
46. Hung M, Voss MW, Rosales MN, et al. Application of machine 333e40.
learning for diagnostic prediction of root caries. Gerodontol- 62. Lee JH, Lee JS, Choi JK, Kweon HI, Kim YT, Choi SH. National
ogy 2019;36:395e404. dental policies and socio-demographic factors affecting
47. Kim Y, Lee KJ, Sunwoo L, et al. Deep Learning in diagnosis of changes in the incidence of periodontal treatments in Vorean:
maxillary sinusitis using conventional radiography. Invest a nationwide population-based retrospective cohort study
Radiol 2019;54:7e15. from 2002-2013. BMC Oral Health 2016;118.
48. Schwendicke F, Elhennawy K, Paris S, Friebertshäuser P, 63. WHO. Oral cancer. 2020. https://www.who.int/cancer/
Krois J. Deep learning for caries lesion detection in near- prevention/diagnosis-screening/oral-cancer/en/. [Accessed
infrared light transillumination images: a pilot study. J Dent 23 February 2020].
2020;92:103260. 64. Xu B, Wang N, Chen T, Li M. Empirical evaluation of rectified
49. Kunz F, Stellzig-Eisenhauer A, Zeman F, Boldt J. Artificial in- activations in convolutional network. 2015. https://arxiv.org/
telligence in orthodontics: evaluation of a fully automated abs/1505.00853. [Accessed 24 February 2020].