Evaluation of the efficiency of computerized algorithms to formulate a decision support system for deepbite treatment planning


This study aimed to evaluate the efficiency of a newly constructed computer-based decision support system (DSS) on the basis of artificial intelligence technology and designed to plan treatment for patients with a deep overbite.


With the help of information technology, a DSS was developed specifically for treatment planning of deepbite malocclusion. The program inputs were the components and the contributing factors used commonly by the orthodontic clinicians in deepbite diagnosis. The program outputs were the treatment planning options for deepbite treatment. A total of 357 decisions made by the algorithm were evaluated for accuracy by comparing them to the actual treatment changes of 51 patients with a well-treated deepbite.


The decisions made by the algorithm were precise, with 94.4% having a very good agreement with actual treatment changes determined using Cohen’s kappa coefficient.


The constructed DSS was shown to be an efficient tool for planning treatment of deep overbite malocclusion in the permanent dentition; thus, the artificial intelligence could be used to formulate a customized plan for orthodontic clinicians.


  • The efficiency of a computer algorithm to plan treatment for a deepbite was tested.

  • The sample comprised records of 51 patients with a well-finished and well-documented deepbite.

  • Conventional and computer algorithm-derived treatment planning methods were compared.

  • There was an agreement between the decisions made by the 2 methods.

  • The decision support system was an efficient tool for deepbite treatment planning.

Deepbite or deep overbite is defined as excessive vertical overlapping of the mandibular incisors by the maxillary incisors in centric occlusion. Normally, the incisal edges of the mandibular teeth should contact slightly at or above the cingulum plateau of the maxillary teeth. Because of differences in the crown lengths of the incisors, normal overbite is about 30% or one third of the clinical crown height of the mandibular incisors.

Deep overbite is one of the most common malocclusions that orthodontists come across during their daily practice. For a patient to have a deep overbite, the anterior overbite should be more than 5 mm. This condition is found in nearly 20% of children and 13% of adults, representing about 95.2% of the vertical occlusal problems. In orthodontics, various methods have been provided for deepbite correction via orthodontic tooth movement, such as extrusion of posterior teeth, intrusion of anterior teeth, proclination of incisors, leveling the curve of Spee, or combining more than one treatment option. The main objectives of planning the treatment are reporting the problems, coming up with proper solutions that would maximize benefits for the patient, and reducing treatment time and complexity of mechanics.

Most orthodontists make their decisions on the basis of previous clinical experience missing a standardized recipe leading to an unorganized treatment plan. This frequently causes significant variability in the treatment planning process, especially between qualified and less-experienced orthodontists. Deepbite does not only comprise a multitude of treatment options, but also it is considered a multifactorial problem, in which many components share in its etiology, in addition to other dental, skeletal, and soft tissue factors that could affect the treatment decision such as the incisors showing when smiling. Accordingly, there should be a more systemic treatment planning equation containing all those diagnostic variables, thus reducing the inconsistency during treatment planning. Based on this, another approach that could facilitate and refine the treatment planning process was required.

Artificial intelligence (AI) techniques have been widely used in the health care specialty. The constructed algorithms use the obtained perceptions to support clinical practice. In addition, an AI system can help to decrease diagnostic and therapeutic errors that are predictable in human clinical practice. In Dermatology, neural network algorithms performed skin cancer classification. As for radiology, computer programs were developed to detect pneumonia on chest x-rays in a simpler way than the conventional diagnostic methods done by radiologists.

Recently, AI and bioinformatics have been introduced in orthodontics in the form of computerized algorithms and softwares, providing orthodontists with fully automated and reliable identification of the cephalometric landmarks, , along with constructing algorithms to support practitioner’s decisions in the extraction or nonextraction dilemma.

This concept paved the way to formulate decision support systems (DSSs) in other aspects of orthodontic treatment planning. Hence, we attempted to design a system specialized in the treatment planning for patients with a deep overbite that considers all the components and clinical contributing factors and any other factors that were reported to be used commonly in this process. This study was designed to test the efficiency of computer algorithms in the process of deepbite treatment planning through testing the degree of agreement between the computer-derived treatment planning decisions and the actual treatment changes for a group of patients with well-finished deep overbites.

Material and methods

This study was conducted at the orthodontic departments of 2 universities. This study was approved by the Research Ethics Committee of the Faculty of Oral and Dental Medicine, in Future University in Egypt. All patients’ files were recruited from the orthodontic department, in which all the files were stored in numerical order and kept in a secured case.

The sample size required to achieve 85% (margin of error 85 ± 10 assuming alpha error with a significance of 95%) of agreement between the 2 methods (computerized and conventional) in terms of treatment planning decision (primary outcome), was 49 deepbite patients (343 decisions). This value was extracted from both previous studies , and expert opinions in orthodontics and AI.

The selected sample was allocated into either a conventional or computerized group. The conventional group was allocated by calculating the actual dental changes after treatment, whereas the computerized group was allocated by computer-derived treatment planning.

Blinding of the judges from the treatment planning results generated by the computerized algorithm was applied to eliminate bias.

Detailed patient history and comprehensive extraoral and intraoral (preoperative and postoperative) treatment records were recruited to ensure fulfillment of the inclusion criteria ( Table I ).

Table I
Inclusion and exclusion criteria for the included participants
Inclusion criteria
Patients with a history of deep overbite before orthodontic treatment
Patients with well-finished orthodontic treatment
Patients with a full set of permanent teeth erupted
Skeletal characteristics: patients with mild to moderate skeletal discrepancy
Well-documented patients with both preoperative and postoperative records
Exclusion criteria
Improperly finished orthodontic treatment
Patients with severe skeletal discrepancy
Poorly documented files

For extraction of input variables (etiologic and contributing factors), a list of the resultant diagnostic variables was made by the judges. As such, the factors that could affect a clinician’s decision were collected through a thorough analysis of all the diagnostic variables that contribute to the development of deepbite malocclusion. These variables were obtained from study models, cephalograms, and clinical examination.

These diagnostic variables were used 2 times. First, to feed the program (with the help of the engineer) in the process of the construction of the algorithm; second, to act as an input method for the computerized algorithm (by the orthodontist) in the process of running the program ( Table II ).

Table II
Etiologic and contributing diagnostic variables, with their extraction sources (inputs) and treatment modalities for deep overbite (outputs)
Source of extracting the diagnostic variables (inputs)
Diagnostic charts
Cephalometric x-rays
Angle classification of malocclusion
Vertical position of maxillary and mandibular incisors
Maxillary and mandibular anterior teeth inclination
Maxillary and mandibular posterior teeth vertical position
Lower anterior facial height
Mandibular plane angle
Interlabial gap
Nasolabial angle
Pretreatment photographs
Maxillary incisor shows on smile
Maxillary incisor displays at rest
Smile line
Diagnostic study casts
Maxillary and mandibular incisor clinical crown length
Depth of the curve of Spee
Treatment modalities for deep overbite (outputs)
Maxillary incisor intrusion
Mandibular incisor intrusion
Maxillary buccal segment extrusion
Mandibular buccal segment extrusion
Proclinning maxillary incisors
Leveling curve of Spee
Proclining mandibular incisors

For extraction of output variables (decisions), the judges agreed that the treatment options for the selected patients with deep overbite should be categorized into 7 treatment planning options ( Table II ). Each of them will take a percentage on the basis of the entered input data values formulating a customized treatment plan tailored for patients with a deepbite.

Setup of the DSS algorithm, with the help of an information technology engineer, was designed using a Java DSS comprising (1) inputs: diagnostic data in terms of the components and the contributing factors of deepbite malocclusion, (2) outputs: the 7-treatment planning option in the form of a percentage given for each outcome, and (3) courses of actions: using the diagnostic data to formulate a customized treatment plan for every patient.

The previously mentioned factors (components and contributing factors) were used to direct the algorithm toward planning the treatment in which each of the diagnostic variables’ values (inputs of the algorithm) had its impact on increasing or decreasing the value of one or more of the treatment planning options (outputs of the algorithm).

This impact value has a direct relationship with the amount of deviation of the diagnostic value away from its standard deviation. The more the deviation of the etiologic factor (component) or contributing factor from its standard deviation, the more its impact on directing the algorithm to resolve the corresponding problem. The display of the strength of every output (treatment option) would be represented by a percentage denoting its contribution in the deepbite treatment for each patient. This approach created a DSS comprising an input of 20 components and contributing factors providing an output of only 7 treatment options ( Fig 1 ). Machine learning was not done as DSS does not rely on software learning, but it rather depends on the originally established courses of actions between the inputs and outputs.

Fig 1
Screenshot of the algorithm showing the inputs (diagnostic data) and the outputs (treatment planning decisions).

To test the validity of the created system, a comparison with a gold standard was a must. Accordingly, a total of 1200 patients with permanent dentitions, aged from 16 to 22 years, were extracted. From which 442 patients were diagnosed with deep overbite malocclusion and accepted for fixed appliance orthodontic treatment. Those patients passed 2 steps of quality check.

First, the finishing quality of the patients was assessed with the help of 3 independent judges (Y.M., A.D., M.D.), according to rubrics that were set and approved by them ( Table III ). The rubrics were customized to fit testing the quality of any finished deep overbite patient considering all the dental, skeletal, and esthetic factors. A total of 130 patients only passed this step.

Table III
Measurements done to test the eligibility of patients to be included in the study (checklist)
Variables The included patients in the study should lie within the following normal ranges of the checklist variables
Extraoral assessment on posttreatment photographs
Maxillary incisors displaying at rest 3 ± 2 mm
Length of maxillary incisors showing on smile 10 ± 2 mm
Interlabial gap 0-3 mm
Smile arc It should be consonant
Nasolabial angle 120 ± 5°
Posttreatment study models
Overbite 1-3 mm
Curve of Spee 1.5-2.0 mm (posttreatment study models)
Quality of finishing 6 keys of normal occlusion
Posttreatment cephalometric x-rays
Maxillary anterior teeth inclination (SN-U1) 104 ± 6° (posttreatment cephalograms)
Mandibular Plane Angle (MMA angle) 27 ± 5° (posttreatment cephalograms)
Mandibular anterior teeth inclination (L1-Mp) 95 ± 5° (posttreatment cephalograms)
Only gold members can continue reading. Log In or Register to continue

Stay updated, free dental videos. Join our Telegram channel

Apr 19, 2021 | Posted by in Orthodontics | Comments Off on Evaluation of the efficiency of computerized algorithms to formulate a decision support system for deepbite treatment planning

VIDEdental - Online dental courses

Get VIDEdental app for watching clinical videos