Development of a fully AI-based three-dimensional workflow for personalized target set-ups in orthodontics: First results
Introduction To present a fully automated artificial intelligence (AI) workflow to generate personalized orthodontic target set-ups and compare the quality of the occlusal outcome with manual set-ups, using the American Board of Orthodontics Model Grading System (ABO MGS). Material and methods The AI workflow takes an intraoral scan and performs tooth segmentation, landmark detection, tooth completion, and tooth alignment to generate a 3D-printable personalized set-up without human intervention. The occlusal outcome was evaluated on twenty non-extraction full jaw cases treated with customized fixed appliances by comparing the AI set-ups to manual set-ups made by experienced laboratory technicians. Both set-up series were scored with the ABO MGS for alignment and rotations, marginal ridges, buccolingual inclination, overjet, occlusal contacts, occlusal relationships, and interproximal contacts. The ‘alignment and rotations’ category was additionally analysed separately for the anterior (canine to canine) and posterior (premolars to second molars) segments. Paired comparisons used the two-sided Wilcoxon signed-rank test (α = 0.05). Results Based on the total ABO MGS scores, the AI set-ups (mean: 14.95) and manual set-ups (mean: 12.20) showed high-quality occlusal outcomes, which were both far below the ABO passing score. AI set-ups showed significantly more penalty points for alignment and rotations (median: 8 vs. 5; P < 0.001). This difference was confined to the posterior segments; anterior alignment was near-ideal for both methods. Other categories did not differ significantly ( P = 0.082–0.886). Conclusions For the first time, a fully automated AI-based 3D workflow generated personalized orthodontic target set-ups for non-extraction cases with a clinically acceptable, high-quality occlusal outcome, in approximately four minutes. Posterior alignment remains the only shortcoming of the AI set-ups and is the obvious target for further algorithm refinement or brief manual correction.
Authors
- Dirk Wiechmann
- Niels van Nistelrooij (ORCID: https://orcid.org/0009-0002-9123-9756)
- Shankeeth Vinayahalingam (ORCID: https://orcid.org/0000-0002-2679-3841)
Institutions
- University of Bern (CH)
- Radboud University Nijmegen (NL)
- Radboud University Medical Center (NL)
- Medizinische Hochschule Hannover (DE)
Publication Details
- Journal
- International Orthodontics
- Published
- 2026-11-21
- DOI
- https://doi.org/10.1016/j.ortho.2026.101269
- Primary Topic
- Orthodontics and Dentofacial Orthopedics
- Type
- article
- Field-Weighted Citation Impact
- 0.00