Eur J Orthod. 2026 Jun 2;48(4):cjag050. doi: 10.1093/ejo/cjag050.
ABSTRACT
BACKGROUND: Artificial intelligence has been gaining popularity in all fields of dentistry. Orthodontic screening is needed to categorize patients for treatment eligibility and urgency of care in public orthodontic clinics. However, screening is time consuming due to high demand. This is the first study to investigate the use of artificial intelligence-supported teleorthodontics for orthodontic screening and triage.
OBJECTIVES: The objective of this study was to investigate the validity of teleorthodontics and artificial intelligence (TAI) in orthodontic screening and triage in comparison to face-to-face (F2F) screening.
TRIAL DESIGN: This study was designed as a single-centre crossover randomized controlled trial.
METHODS: A total of 255 patients referred for public orthodontic treatment were randomized into two sequences: control, F2F triage first, and test, TAI triage first. A total of 178 participants completed the trial (age range: 7-38 years) with 95 participants enrolled initially to the control and 83 to the test sequence, respectively. For TAI triage, patients submitted intraoral scans using Dental Monitoring™ (DM™), extraoral photos, and an online patient history survey. After a 2-month washout period, participants were re-triaged with the other method.
OUTCOMES: The primary outcome was the validity of TAI triage in referral acceptance or rejection based on a minimum Index of Orthodontic Treatment Need (IOTN) Dental Health Component (DHC) threshold of ≥3. Secondary outcomes were diagnostic validity of TAI for all IOTN grades, at referral acceptance threshold IOTN ≥ 4, and triage duration comparison.
RANDOMIZATION: Patients were randomized using a permuted randomized block design (allocation ratio 1:1).
BLINDING: Investigators and participants could not be blinded to sequence allocation.
RESULTS: Referral acceptance or rejection at IOTN ≥ 3, TAI triage had a sensitivity of 1, a specificity of 0.67, and an overall diagnostic accuracy of 0.98, with three referrals incorrectly rejected by TAI. Artificial intelligence could not detect OB and OJ correctly for some patients and did not measure important traits, including crossbite, contact point displacement, and functional shift. Teleorthodontic triage duration was 2.9 times faster than F2F.
LIMITATIONS: This study was conducted in a public orthodontic clinic, and results apply to this setting when using IOTN and the hybrid method used in this investigation.
CONCLUSIONS: Teleorthodontics combined with DM™ is a valid and reliable tool for orthodontic screening of patients with mild and severe malocclusions but cannot be used to confidently assign IOTN-DHC grade for patients with borderline malocclusion severity yet. The duration of screening is significantly shorter using TAI.
REGISTRATION: Australian New Zealand Clinical Trials Registry ID: ACTRN12623000327684.
PMID:42400943 | DOI:10.1093/ejo/cjag050
