Journal of Contemporary Orthodontics 2023;7(4):281–286

Content available at: https://www.ipinnovative.com/open-access-journals

Journal of Contemporary Orthodontics

Journal homepage: https://www.jco-ios.org/

Original Research Article

Artificial intelligence-based automated model for prediction of extraction using
neural network machine learning: A scope and performance analysis
Mridula Trehan 1 *, Deeksha Bhanotia 1, Tarannum Alam Shaikh1 ,
Shivangi Sharma1 , Sunil Sharma1
1 Dept. of Orthodontics, NIMS Dental College & Hospital, Rajasthan, India

ARTICLE INFO ABSTRACT

Article history: Objective: To compare the Artificial Intelligence based model & conventional technique for prediction of
Received 01-10-2023 extraction in orthodontic treatment plan.
Accepted 08-12-2023 Materials and Methods: A comparative study was conducted on total 700 patients, who were divided
Available online 28-12-2023 into training set and testing set based on simple random sampling by means of computer generated random
numbers. The photographs of the 630 patients [training set] along with the treatment plan finalized for
them based on Arch Perimeter & Carey’s Analysis, was fed in the AI model [convolutional neural network
Keywords:
(ResNet-50)] in order to train it for the stipulated function of eventually predicting the treatment plan in the
Extraction
testing set [70 patients], based on the input of the right profile photographs. The accuracy of measurement
Orthodontic treatment planning of the parameters of these seventy test set patients by the machine learning model relative to the manual
Artificial intelligence method was compared eventually. Using the Statistical Package for Social Sciences, the acquired data
was statistically analyzed, and p <0.05 was deemed statistically significant. The normality of the data was
examined using the Shapiro-Wilks test and the Kolmogorov-Smirnov test. Depending on the collected data
and normality assessed, appropriate reliability was estimated.
Result: The analysis of 70 test patients showed that 65.12% of the total extraction cases and 62.96% of the
total non-extraction cases (as predicted by the AI model) were in agreement with the results of the model
analysis.
Conclusion: It is suggested that the present AI model can further be developed in order to improve the
accuracy of prediction.

This is an Open Access (OA) journal, and articles are distributed under the terms of the Creative Commons
Attribution-NonCommercial 4.0 International, which allows others to remix, and build upon the work non-
commercially, as long as appropriate credit is given and the new creations are licensed under the identical
terms.
For reprints contact: [email protected]

1. Introduction In recent years, orthodontics has made extensive use of

artificial intelligence to improve the efficiency and accuracy
The use of computers and other technologies to simulate of the diagnosis process. Since healthcare professionals
human behavior and mental processes is known as artificial are ultimately responsible for diagnosing patients and
intelligence, or AI. John McCarthy first described the term determining the best course of therapy, the advent of AI-
AI in 1956 as the science and engineering of making based models cannot completely replace them. 3 AI may
intelligent machines. 1 The goal of AI is to produce a be a helpful tool for making precise healthcare judgments
machine that is able to learn through data and can solve in a short amount of time. AI apps can help doctors
problems by itself. 2 make better judgments and hence perform better since the
results produced by AI are often quite accurate and can
* Corresponding author. thus be utilized, in certain situations, to prevent human
E-mail address: [email protected] (M. Trehan). errors. 4 They are dependable, faster, and have the potential

https://doi.org/10.18231/j.jco.2023.048
2582-0478/© 2023 Author(s), Published by Innovative Publication. 281
282 Trehan et al. / Journal of Contemporary Orthodontics 2023;7(4):281–286

to automatically complete tasks with the competence of on plaster study models and artificial intelligence-based
experienced clinicians. 2–5 automated model.
Plaster study models are an integral element of the
orthodontic patient record. They provide a progressive 2. Materials and Methods
record of care, which the practitioner uses as a diagnostic
A comparative study was conducted in National Institute
tool. Plaster model technique has better repeatability for
of Medical Sciences University to predict the need for
many of the parameters. 6 Plaster models may not be
extraction in orthodontic treatment planning using Artificial
an accurate replica of the real tooth dimensions due to
Neural Network modelling and conventional technique.
dimensional changes in the impression materials and stone
Data of seven hundred patients reporting to the Department
during preparation, even if they have been utilized for
of Orthodontics and Dentofacial Orthopaedics at the
the construction of appliances that fit precisely in the
institution who met the inclusion criteria were considered
mouth. 7 However, plaster models have been authenticated
for the study.
since many years, but their use is associated with several
problems, mainly storage, mutilation and loss. 8
A comprehensive diagnosis and treatment planning
leads to a successful orthodontic treatment. The degree of
obliquity and the mismatch between the arch form and tooth
dimensions is assessed using orthodontic model analysis
as one of the diagnostic and treatment planning tools. 9
This eases the three-dimensional (3D) documentation
of the dental arches in pre-treatment, progress and
post-treatment records. 10 Important components of an
orthodontic treatment planning include tooth size- arch
length discrepancies, maxillomandibular relationship, facial
profile, skeletal maturation, dental asymmetries, and patient
cooperation. 11 Tooth crowding and protrusions demand Figure 1: Artificial neural network model
rigorous attention during orthodontic planning and may
require the extraction of first or second premolars. 12
Strict observation and indications are necessary for
application of extraction modality. 13 It involves thorough
knowledge of the laws governing the movement of
teeth post extraction, the normal development of the
orofacial system and the eruption of teeth. Inconsiderate
extractions performed without thorough prior analysis lead
to irreparable damage rather than improvement of the
situation. 14 For most cases, extractions must precede fixed
mechanotherapy to achieve controlled closure of the spaces,
alignment of teeth in the dental arch and restoration of
proper occlusion. 15
The choice between extraction and non-extraction
treatment modality is an endless debate in orthodontics that
has seen many phases throughout the journey of time. The Figure 2: The extra-oral right profile photograph of the patients
model analysis of permanent dentition should be performed
meticulously so that different treatment possibilities can be
explored. 16 Table 1: Distribution of patients advised for extraction or
Due to the shortcomings of performing model non-extraction treatment modality based on model analysis,
analysis(es) manually, attempts have been made to utilize constituting the TRAINING SET for the AI model
technology in order to minimize human errors and thereby Treatment plan No. of patients (% of
aid in the enhancement of diagnostic accuracy. 17 Artificial the total training set
data)
Intelligence is certainly expected to present a lucrative deal
Indicated for extraction 284 (45.07%)
in this regard. This study was performed in order to compare
Not indicated for extraction 346 (54.92%)
the judgement for the need for extraction in orthodontic
Total 630(100%)
treatment planning between the conventional technique
based on Arch Perimeter and Carey’s Analysis performed
 Trehan et al. / Journal of Contemporary Orthodontics 2023;7(4):281–286 283

Table 2: Treatment plan derived using model analysis for patients

in Test Set
Treatment plan No. of patients (% of
the total test set data)
Employing extraction 38 (54.28%)
Not employing extraction 32 (45.72%)
Total 70(100%)

Table 3: Treatment plan predicted by the AI based model for the

Figure 3: Interpretation of diagnosis using software based on right patients comprising the Test Set.
profile photograph Treatment plan No. of patients (% of
the total test set data)
Employing extraction 43 (61.42%)
Not employing extraction 27 (38.57%)

Table 4: Interpretation of treatment decision in test group

Treatment decision In agreement Not in
predicted by AI with agreement with
model conventional conventional
method method
Extraction 28 ( 65.12%) 15 ( 34.88%)
Non extraction 17 (62.96%) 10 ( 37.04%)
Graph 1: Graph depicting model accuracy and model loss in
training set
Before the commencement of this study, approval
was obtained from the Institutional Ethics Committee,
NIMS University Rajasthan, Jaipur. Patients with normal
occlusion, patients with malocclusions except Angle’s Class
II div 2, Angle’s Class I type 1 M/O, individuals without a
prior history of extraction, orthodontic treatment, or partial
eruption of the permanent dentition up to the second molars,
as well as individuals without any dental abnormalities up
to the second molars, were included. And the exclusion
criteria led to elimination of the patients with skeletal
asymmetry and maxillofacial deformities, patients with
dental crowding, but normal soft tissue profile, patients
with Angle’s Class II div 2, Angle’s Class I type 1 M/O,
patients with missing teeth except congenitally missing third
molars, patients with impacted or unerupted teeth except
third molars & the patients with retained deciduous teeth.
Graph 2: The ROC graph for the ANN model A particular method was used to estimate the sample size
based on information from a prior study, and a sample size
of 700 was determined to be the appropriate number.
The conventional method comprises the assessment of
the need for extraction based on Arch Perimeter Analysis for
the maxillary arch and Carey’s Analysis for the mandibular
arch of the patient. The afore-mentioned model analysis(es)
were performed by measuring the mesio-distal diameters of
the teeth present in the upper and lower arches respectively,
on the plaster study models and recording the same on a
graph paper. The arch form was outlined from this graph
and the linear arch dimension was recorded. The available
Graph 3: Graphs depicting the validation accuracy of the linear arch dimension was then measured by adapting a
algorithm of the model 0.020 brass wire on the cast from the mesial marginal ridge
of the first molar on one side to the other, passing over the
284 Trehan et al. / Journal of Contemporary Orthodontics 2023;7(4):281–286

premolars through their greatest diameter and around and threshold of significance maintained at 5%. The normality
over the anterior ridge where the incisal edges of the teeth of the data was examined using the Shapiro-Wilks test
were intended to relocate. 11 Arch- length- tooth material and the Kolmogorov-Smirnov test. A suitable estimate of
discrepancy was then determined. The inferences for the dependability was made based on the data that had been
above analysis(es) being: If the arch discrepancy is lesser gathered and the normalcy test results.
than 2.5 mm, there is no need for extraction; the second
and first premolars, respectively, should be extracted if the 3. Result
difference is 2.5 to 5 mm and higher than 5 mm. 18
The dataset, for input feedback of the artificial Out of the 630 patients in the training group, 284 patients
intelligence based automated model utilized the extra- (i.e. 45.07% of the total training set data) were accepted
oral right profile photographs of the patients under for extraction treatment and 346 patients (i.e. 54.92% of
study to determine the requirement for extraction in the total training set data) were chosen for non-extraction
orthodontic treatment planning. A convolutional neural treatment based on Arch Perimeter & Carey’s Analysis.
network (ResNet-50), which is a type of ANN, was (Table 1)
used to predict the need for extraction. The images were Out of the 70 test set patients, 38 patients (i.e. 54.28%
preprocessed as required using horizontal and vertical of the total test set data) were indicated for extraction and
flipping and rescaling. They were re-sized to sixty-four 32 patients (i.e. 45.72% of the total test set data) were
by sixty-four size images and a batch-size of thirty-two not indicated for extraction based on conventional method.
was used to ingest images into the model. The model was (Table 2)
specifically designed to be a categorical classification model The right profile photographs of the 70 test set patients
with two predictive classes and was trained for twenty were uploaded on the AI model and the ANN model was
epochs. made to predict the treatment plan for the patients based
The model consists of two dense layers with one on the profile photographs. 43 patients (61.42% of the total
thousand and twenty-four and five hundred and twelve test set data) were predicted to undergo extraction line
nodes respectively, the ReLu activation function was applied of treatment and 27 patients (38.57% of the total test set
to add non-linearity to the model and a dropout of zero data) were indicated for non-extraction line of treatment.
point two was applied to prevent overfitting. With a learning (Table 3)
rate of 0.0001, the Adam optimizer was utilized together The treatment decision predicted by the AI based model
with category cross entropy loss. The model’s test accuracy for the 70 test set patients was eventually compared with the
ended up being around 65%. The data collected was entered results found in the model analysis for the same, to test the
into the artificial neural network (ANN) model and was ANN model for accuracy, as rightly suggested by the title of
checked for metrics including accuracy, precision, recall, the set.
true positive rate, and false positive rate. An ROC curve was Out of the 43 patients predicted as extraction cases by the
also made to analyse the accuracy. ANN model, the treatment plans for 28 cases (i.e. 65.12%
of the total extraction cases as predicted by the AI model)
The total data set (700) was divided into training set and
were in agreement with the results of the model analysis,
test set based on simple random sampling. 90% (630) of
whereas the treatment plan for 15 patients (i.e. 34.88% of
the total data constituted the training set.These 630 patients
the total cases predicted for extraction) did not conform
were selected to establish the training set based on simple
with the findings of the model analysis. Out of the 27 cases
random sampling by means of computer generated random
predicted for non-extraction treatment modality, 17 patients
numbers. Data of the remaining 10% (70) patients was
(i.e. 62.96% of the total non-extraction cases as predicted
aggregated to constitute the testing set.
by the AI model) were presented with a treatment decision
The photographs of the 630 patients along with the which was in agreement with the model analysis, however,
treatment plan finalised for them based on Arch Perimeter for 10 patients (i.e. 37.04% of the total cases predicted for
& Carey’s Analysis, was fed in the AI model in order to non-extraction treatment modality) the treatment decision
train it for the stipulated function of eventually predicting was in disagreement with the decision based on model
the treatment plan in the testing set, based on the input of analysis.(Table 4 ) The sensitivity for the prediction of
the right profile photographs. The accuracy of measurement treatment plan (either indicating or contraindicating the
of the parameters of these seventy test set patients by the need for extraction) of the AI based automated model was
machine learning model relative to the manual method was calculated as 64.04%.
compared eventually.
The Statistical Package for Social Sciences (SPSS, IBM 4. Discussion
version 20.0) was used to conduct statistical analysis on
the acquired data once it had been imported into Microsoft In a variety of clinical situations, artificial intelligence
Excel. Statistical significance was defined as p 0.05, with the approaches including Bayesian networks, artificial neural
 Trehan et al. / Journal of Contemporary Orthodontics 2023;7(4):281–286 285

networks, fuzzy expert systems, and hybrid intelligent The main limitation of the present study is that the
systems have been applied. 19 sensitivity is confined to 64.04%. Another limitation is that
The tremendous increase in the awareness and in the training group, the decision for extraction is purely
expectations towards dental health in recent times is based on Carey’s and Arch Perimeter Analysis without any
encouraging health care professionals to provide better consideration of soft tissue. Additionally, the AI model
standards of care. 20 Application of AI in orthodontics was unable to take into account patients with soft tissue
has the potential to revolutionalize the current system of anomalies, unusual extraction patterns & missing teeth. 29,30
practice. 21 These automated models simplify tasks and
provide results in no time, enabling the orthodontists to 5. Conclusions
become more efficient. 22 The present Artificial Neural
Network based model utilizes a profile view of the patient’s A comparative study was conducted to predict the need for
face to predict if there is need for extraction or not, in order extraction in orthodontic treatment planning using Artificial
to correct the malocclusion presented by the patient. 23 Neural Network modelling and conventional technique to
This valuable clinical aid will not only save time for the conclude that the ANN model presented a sensitivity of
orthodontists but can be used as an auxiliary support for 64.04% relative to the conventional method. The prediction
less experienced practitioners. 16,24 To determine whether accuracy was 65.12% for the extraction cases and 62.96%
extractions are required before receiving orthodontic for the non-extraction cases relative to the conventional
treatment, a decision-making expert system based on an method.
artificial neural network (ANN) might be helpful. 25 It is suggested that the present AI model can be
In the présent study, out of the six hundred and thirty developed so as to cover all the cases. In the future,
patients in the training group, cases selected for non- increasing the sample size of the dataset further, would give
extraction treatment modality were greater in number even better results because Deep Learning models generally
and the différence in distribution was not statistically perform better given a large sample size.
significant. 26
The prediction of accuracy of ANN model was 65.12% 6. Source of Funding
for extraction cases and 62.96% for non-extraction cases None.
and the AI model used in the present study had a
sensitivity of 64.04%, whereas Xie et al. (2010) used 7. Conflict of Interest
an ANN system to determine whether an extraction or
non- extraction treatment was good for juvenile patients None.
presenting malocclusion, and found the ANN worked with
eighty percent accuracy. 27 References
Study by Jung et al. (2016) suggested the success rates 1. Amisha MP, Pathania M, Rathaur VK. Overview of artificial
were ninety two percent in the training set and ninety three intelligence in medicine. J Fam Med Prim Car. 2019;8(5):2328–59.
percent in the test set. The learning set was divided into 2. González AM, Calatayud LR, Oliveira NG, Ustrell-Torrent JM.
Artificial intelligence in orthodontics: Where are we now? A scoping
training set and validation set to minimize overfitting and review. Orthod Craniofac Res. 2021;24(2):6–15.
to verify the fitness of the model. 28 3. Abizadeh N, Moles DR, JO’Neill, Noar JH. Digital versus plaster
In the study by Xie et al. 7 the age group of patients taken study models: how accurate and reproducible are they? J Orthod.
under study was limited to eleven to fifteen years old, unlike 2012;39(3):151–60.
4. Yılmaz H, Özlü FÇ, Karadeniz C, Karadeniz E. Time-Efficiency
our study in which age is no bar. So, our present model and Accuracy of Three-Dimensional Models Versus Dental Casts: A
ponders to a greater population than the above. At the same Clinical Study. Turk J Orthod. 2019;32(4):214–22.
time, the procedure in the ANN model developed by Xie et 5. Araújo TM, Caldas LD. Tooth extractions in Orthodontics: first or
al. 27 requires multiple inputs including cast measurements, second premolars? Dental Press J Orthod. 2019;24(3):88–98.
6. Bogataj J, Gantar B. Pomenzobnihekstrakcij v celjustniortopediji.
hard tissue cephalometrics and soft tissue cephalometrics, Role of tooth extractions in orthodontics. Zobo Drav Vestn.
whereas our present model gives the decision based on 1989;44(3):58–67.
the input of just a single and not so difficult to retrieve 7. Selvaraj M, Sennimalai K. Orthodontic model analysis in the
permanent dentition: A review of past, and current methods. IP Ind
parameter, the right profile photograph. We are utilizing an
J Orthod Dentofac Res. 2022;8(4):220–6.
essential diagnostic aid for planning the treatment compared 8. Weintraub JA, Vig PS, Brown C, Kowalski CJ. The prevalence
to the above study which utilizes a supplementary diagnostic of orthodontic extractions. Am J Orthod Dentofac Orthop.
aid, i.e. lateral ceph for diagnosis, which makes our study 1989;96(6):462–6.
9. Rudge SJ, Jones PT, Hepenstal S, Bowden D, Cardiff C. The reliability
superior. 28–35 of study model measurement in the evaluation of crowding. Euro J
Advantages of the present study are that the present AI Ortho. 1983;5(1):225–31.
model can assess the treatment plan purely based on single 10. Battagel JM. The Assessment of Crowding Without the Need to
parameter i.e., right profile pic and has no boundations of Record Arch Perimeter. Part I: Arches With Acceptable Alignment.
age limit. Brit J Orthod. 2016;23(2):137–44.
286 Trehan et al. / Journal of Contemporary Orthodontics 2023;7(4):281–286

11. Noroozi H. Orthodontic treatment planning software. Am J Orthod 28. Jung SK, Kim TW. New approach for the diagnosis of extractions
Dentofac Orthop. 2006;129(6):834–41. with neural network machine learning. Am J Orthod Dentofac Orthop.
12. Zarei A, El-Sharkawi M, Hairfield M, King G. An Intelligent System 2016;14(9):127–60.
for Prediction of Orthodontic Treatment Outcome. Int Join Conf Neur 29. Moon S, Mohamed AMC, He Y, Dong W, Yaosen C, Yang Y,
Net Proc. 2006;1(5):2702–6. et al. Extraction vs. Non extraction on Soft-Tissue Profile Change in
13. Burrow SJ. To extract or not to extract: A diagnostic decision, not a Patients with Malocclusion: A Systematic Review and Meta-Analysis.
marketing decision. Am J Orthod Dentofac Orthop. 2007;133(3):341– Biomed Res Int. 2021;4(3):7751516.
2. 30. Khanagar SB, Al-Ehaideb A, Vishwanathaiah S, Maganur CP, Patil
14. Yagi M, Ohno H, Takada K. Decision-making system for orthodontic S, Naik S, et al. Scope and performance of artificial intelligence
treatment planning based on direct implementation of expertise technology in orthodontic diagnosis, treatment planning, and clinical
knowledge. Annu Int Conf Eng Med Biol Soc. 2010;5(9):2894–901. decision-making - A systematic review. J Dent Scie. 2020;16(1):482–
15. Bootvong K, Liu Z, Mcgrath C, Hägg U, Wong RW, Bendeus M, et al. 92.
Virtual model analysis as an alternative approach to plaster model 31. Ryu J, Lee YS, Mo SP. Application of deep learning artificial
analysis: reliability and validity. Eur J Orthod. 2010;32(5):589–95. intelligence technique to the classification of clinical orthodontic
16. Konstantonis D, Anthopoulou C, Makou M. Extraction decision and photos. BMC Ora H. 2022;5(9):454.
identification of treatment predictors in Class I malocclusions. Prog 32. Albalawi F, Alamoud KA. Trends and Application of Artificial
Orthod. 2013;1(4):47. doi:10.1186/2196-1042-14-47. Intelligence Technology in Orthodontic Diagnosis and Treatment
17. Carey CW. Diagnosis and case analysis in orthodontics. Am A Orthod. Planning-A Review. App Sci. 2022;12(22):11864.
1951;4(7):149–61. 33. Hussain MA, Fatima S, Reddy KK, Ramya Y, Betha SP, Kauser A,
18. Yu X, Liu B, Pei Y, Xu T. Evaluation of facial attractiveness for et al. Artificial intelligence in orthodontics: A review. Int J Heal Scie.
patients with malocclusion: a machine-learning technique employing 2022;6(S2):9378–83.
Procrustes. Angle Orthod. 2014;84(3):410–6. 34. Jihed M, Dallel I, Tobji S. Adel Ben Amor. The Impact of Artificial
19. Moreira DD, Gribel BF, Torres GDR, Vasconcelos KF, Freitas DQ. Intelligence on Contemporary Orthodontic Treatment Planning - A
Reliability of measurements on virtual models obtained from scanning Systematic Review and Meta-Analysis. Sch J Dent Sci. 2022;9(5):70–
of impressions and conventional plaster models. Braz J Oral Sci. 87.
2014;13(4):297–302. 35. Ding H, Wu J, Zhao W, Matinlinna JP, Burrow MF, Tsoi JKH.
20. Auconi P, Scazzocchio M, Cozza P, Mcnamara JA, Franchi L. Artificial intelligence in dentistry-A review. Front Dent Med.
Prediction of Class III treatment outcomes through orthodontic data 2023;5(8):1085251.
mining. Eur J Orthod. 2015;37(3):257–67.
21. Kumar A. Bayesian Network Analysis: A New Approach to Diagnosis
and Prognosis. J Dent Med Sci. 2016;15(7):1–4. Author biography
22. Khanum A, Prashantha GS, Mathew S, Naidu M, Kumar A. Extraction
vs Non Extraction Controversy: A Review. J Dent Orofacia Res.
2018;14(1):41–9. Mridula Trehan, Dean and Principal, Professor and Head
23. Li P, Kong D, Tang T. Orthodontic Treatment Planning based on https://orcid.org/0000-0001-7329-3818
Artificial Neural Networks. Sci J Dent. 2019;7(9):2037–2037.
24. Lee KS, Ryu JJ, Jang HS, Lee DY, Jung SK. Deep Convolutional Deeksha Bhanotia, PG Student https://orcid.org/0009-0000-2969-8041
Neural Networks Based Analysis of Cephalometric Radiographs for
Differential Diagnosis of Orthognathic Surgery Indications. Ap Sci. Tarannum Alam Shaikh, PG Student
2020;10(6):2124.
25. Pan F, Yang Z, Wang J. Influence of orthodontic treatment with
premolar extraction on the spatial position of maxillary third molars in Shivangi Sharma, Master of Science
adult patients: a retrospective cohort cone-bean computed tomography
study. BMC Ora H. 2020;5(6):321. doi:10.1186/s12903-020-01314-0. Sunil Sharma, Professor Vice Chancellor, Professor
26. Xie X, Wang L, Wang A. Artificial neural network modeling for
deciding if extractions are necessary prior to orthodontic treatment.
Angle Orthod. 2010;80(2):262–8.
27. Kunz F, Eisenhauer AS, Zeman F, Boldt J. Artificial intelligence in Cite this article: Trehan M, Bhanotia D, Shaikh TA, Sharma S, Sharma
orthodontics: Evaluation of a fully automated cephalometric analysis S. Artificial intelligence-based automated model for prediction of
using a customized convolutional neural network. J Orofac Orthop. extraction using neural network machine learning: A scope and
2019;81(1):52–68. performance analysis. J Contemp Orthod 2023;7(4):281-286.