AI-Assisted Versus Conventional Orthodontic Diagnosis and Treatment Planning (AI-OrthoDx)

Artificial Intelligence-Assisted Versus Conventional Orthodontic Diagnosis and Treatment Planning: A Prospective Paired Diagnostic-Agreement Study

This study compared an artificial intelligence-assisted orthodontic diagnostic workflow using WebCeph with a conventional clinician-based workflow.

The study included 120 new orthodontic patients treated at Ben Hafeed Dental Center in Sana'a, Yemen, between March 2024 and March 2026. For each participant, the researchers prepared a de-identified package of routine pretreatment records, including clinical findings, extraoral and intraoral photographs, panoramic and lateral cephalometric radiographs, and dental models or digital scans.

The same case package was independently assessed using both the artificial intelligence-assisted and conventional workflows. An independent panel of five expert orthodontists reviewed the cases and established the reference standard.

The main outcome was agreement with the expert panel for skeletal sagittal classification as Class I, Class II, or Class III. Other outcomes included orthodontic treatment-need grades, premolar extraction decisions, anchorage planning, appliance selection, and the time required to complete each diagnostic workflow.

Participation did not alter the routine orthodontic care provided to patients, and final clinical decisions remained the responsibility of the treating orthodontist.

Study Overview

• Status: Completed
• Conditions: Malocclusion, Angle Class II; Malocclusion, Angle Class III; Malocclusion, Angle Class I

Detailed Description

This was a prospective paired diagnostic-agreement study conducted at Ben Hafeed Dental Center, Sana'a, Yemen, from March 2024 to March 2026.

A total of 120 eligible new orthodontic patients were included. Each participant contributed one standardized, de-identified pretreatment case package. Every case was evaluated using both an artificial intelligence-assisted workflow and a conventional clinician-based workflow, allowing direct within-case comparison.

The artificial intelligence-assisted workflow used WebCeph Premium, software version 2.0.3 (AssembleCircle Corp., Seongnam-si, Republic of Korea). The lateral cephalometric radiograph was uploaded to the software for automated anatomical landmark identification and generation of angular and linear cephalometric measurements. The original automated landmarks and measurements were recorded without manual correction. The reviewer interpreted the automated findings together with the remaining standardized pretreatment records and entered the required diagnostic and treatment-planning decisions.

For the conventional workflow, two calibrated orthodontists independently reviewed each case using routine clinical, photographic, radiographic, cephalometric, and dental-model assessment. Conventional cephalometric evaluation was performed using manual landmark identification through a local Dolphin Imaging-based workflow. When the two conventional orthodontists disagreed on an endpoint, a third calibrated orthodontist adjudicated the disputed decision.

The reference standard was established by an independent panel of five experienced orthodontists. Each expert independently reviewed the de-identified cases while blinded to the artificial intelligence-assisted and conventional workflow results. Reference decisions were established through majority agreement, with additional review and structured adjudication when required.

The primary endpoint was skeletal sagittal classification, categorized as Class I, Class II, or Class III. Secondary endpoints included the Index of Orthodontic Treatment Need Dental Health Component, the Index of Orthodontic Treatment Need Aesthetic Component, premolar extraction decision, anchorage-plan category, and primary appliance modality. Workflow duration was measured from the beginning of case assessment until all required decisions had been entered.

The study included 71 females and 49 males. The mean participant age was 20.3 years, with a standard deviation of 3.9 years and a range of 14 to 33 years.

The protocol was approved by the Yemen Dental Association under approval number 207. Written informed consent was obtained from adult participants. For minors, written parental or legal-guardian permission and participant assent were obtained.

• Study Type: Observational
• Enrollment (Actual): 120

Contacts and Locations

Study Locations

• Yemen
  • Sanaa, Yemen
    • Ben Hafeed Dental Center

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child; Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

The study population consisted of new orthodontic patients presenting for assessment at Ben Hafeed Dental Center in Sana'a, Yemen, from March 2024 to March 2026. Eligible participants were aged 12 years or older and had complete, good-quality routine pretreatment orthodontic records. Each enrolled participant contributed one standardized, de-identified case package that was assessed using both the AI-assisted and conventional diagnostic workflows.

Description

Inclusion Criteria:

1. New orthodontic patients aged 12 years or older.
2. Availability of complete routine pretreatment records, including clinical examination data, standardized extraoral and intraoral photographs, panoramic and lateral cephalometric radiographs, and dental models or digital scans.
3. Pretreatment records of sufficient quality to permit comprehensive orthodontic diagnosis and treatment planning.
4. Provision of written informed consent by adult participants or written parental or legal-guardian permission, together with participant assent for minors, in accordance with applicable local requirements.

Exclusion Criteria:

1. Cleft lip and/or palate or another craniofacial syndrome that substantially altered the routine orthodontic diagnostic pathway.
2. A history of previous comprehensive fixed orthodontic treatment.
3. Incomplete, poor-quality, or non-standardized records that prevented comprehensive orthodontic diagnosis and treatment planning.
4. Refusal to participate or subsequent withdrawal of consent.

Study Plan

How is the study designed?

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Exact Agreement With the Expert Panel for Skeletal Sagittal Classification	Proportion of participant cases in which the skeletal sagittal classification assigned by each diagnostic workflow exactly agreed with the expert-panel reference classification. Classification was categorized as Class I, Class II, or Class III. Agreement was reported separately for the AI-assisted and conventional workflows using percentage agreement and unweighted Cohen's kappa with 95% confidence intervals. Paired classification accuracy was compared using McNemar's test.	At baseline pretreatment assessment; one case evaluation per workflow, completed within 1 day.

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Agreement With the Expert Panel for IOTN Dental Health Component Grade	Agreement between each workflow and the expert-panel reference grade for the Index of Orthodontic Treatment Need Dental Health Component, graded from 1 to 5. Outcomes included exact agreement, agreement within one grade, weighted Cohen's kappa with 95% confidence intervals, and absolute grade deviation.	At baseline pretreatment assessment; one case evaluation per workflow, completed within 1 day.
Agreement With the Expert Panel for IOTN Aesthetic Component Grade	Agreement between each workflow and the expert-panel reference grade for the Index of Orthodontic Treatment Need Aesthetic Component, graded from 1 to 10. Outcomes included exact agreement, agreement within one grade, weighted Cohen's kappa with 95% confidence intervals, and absolute grade deviation.	At baseline pretreatment assessment; one case evaluation per workflow, completed within 1 day.
Diagnostic Accuracy for the Premolar Extraction Decision	Accuracy of each workflow in classifying whether premolar extraction was indicated, using the expert-panel decision as the reference standard. Outcomes included sensitivity, specificity, positive predictive value, negative predictive value, overall accuracy, and unweighted Cohen's kappa with 95% confidence intervals.	At baseline pretreatment assessment; one case evaluation per workflow, completed within 1 day.
Agreement With the Expert Panel for Anchorage-Plan Category	Proportion of participant cases in which the anchorage-plan category assigned by each workflow agreed with the expert-panel reference decision. Categories were conventional anchorage, reinforced anchorage, or skeletal anchorage. Agreement was evaluated using percentage agreement and unweighted Cohen's kappa.	At baseline pretreatment assessment; one case evaluation per workflow, completed within 1 day.
Agreement With the Expert Panel for Primary Appliance Modality	Proportion of participant cases in which the primary appliance modality selected by each workflow agreed with the expert-panel reference decision. Categories were fixed appliance, clear aligner, or functional appliance. Agreement was evaluated using percentage agreement and unweighted Cohen's kappa.	At baseline pretreatment assessment; one case evaluation per workflow, completed within 1 day.
Time Required to Complete the Diagnostic and Treatment-Planning Workflow	Total time in minutes from opening the standardized case package until all required diagnostic and treatment-planning decisions were entered. For the conventional workflow, the mean of the two initial orthodontists' assessment times was used. Third-rater adjudication time was excluded. The within-case paired difference between workflows was calculated.	During the baseline diagnostic assessment; measured continuously from case opening to completion, within 1 day.

Collaborators and Investigators

• Sponsor: Ben Hafeed Dental Center

Study record dates

Study Major Dates

• Study Start (Actual): March 3, 2024
• Primary Completion (Actual): March 31, 2026
• Study Completion (Actual): March 31, 2026

Study Registration Dates

• First Submitted: July 2, 2026
• First Submitted That Met QC Criteria: July 2, 2026
• First Posted (Actual): July 8, 2026

Study Record Updates

• Last Update Posted (Actual): July 8, 2026
• Last Update Submitted That Met QC Criteria: July 2, 2026
• Last Verified: July 1, 2026

More Information

Terms related to this study

• Keywords: Artificial intelligence; Cephalometry; Clinical decision support; Orthodontic diagnosis; Orthodontic treatment planning; Diagnostic agreement; WebCeph; Workflow efficiency; Index of Orthodontic Treatment Need; Premolar extraction decision
• Additional Relevant MeSH Terms: Stomatognathic Diseases; Tooth Diseases; Malocclusion; Malocclusion, Angle Class I; Malocclusion, Angle Class II; Malocclusion, Angle Class III
• Other Study ID Numbers: BHDC-AIORT-2024-01; YDA-207 (Other Identifier: Yemen Dental Association)

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No