AI algorithms for predicting orthodontic treatment outcomes based on craniofacial morphology

Background and Aims: Orthodontic treatment planning relies on cephalometric analysis (X-ray measurements) to assess jaw and tooth positions. This process is time-consuming and subject to clinician variability. Artificial intelligence (AI) offers potential for automation and standardization. This study aims to develop neural networks to predict three clinical decisions: (۱) necessity of tooth extraction, (۲) specific teeth for extraction, and (۳) orthodontic appliance selection, based on craniofacial morphology. Methods: Retrospective analysis of ۳۰۲ anonymized orthodontic patient records (۲۰۱۸–۲۰۲۳) was conducted. Data included lateral cephalograms, intraoral photographs, and ۳D dental models. Twenty-four validated cephalometric parameters (e.g., ANB angle, overjet, crowding index) were extracted. Three neural networks (multilayer perceptrons) were trained: Binary classification (extraction yes/no), Multi-class classification (teeth extraction pattern), Appliance selection (fixed vs. removable). Data were partitioned into training (۷۰%), validation (۱۵%), and test sets (۱۵%). Missing data (۵%) were imputed via k-nearest neighbors. Model performance was evaluated using accuracy, AUC-ROC, sensitivity, and specificity. Results: Extraction necessity: Accuracy = ۸۹.۳% (AUC = ۰.۹۵, sensitivity = ۸۸.۱%, specificity = ۹۰.۴%) Teeth selection: Accuracy = ۷۹.۸% (Top-۳ accuracy = ۹۲.۱%) Appliance setup: Accuracy = ۸۷.۶% Key predictive features included anterior crowding index (ACI), ANB angle discrepancy, and incisor inclination. AI reduced planning time by ۶۸% compared to manual analysis. Conclusion: AI models demonstrated clinically relevant accuracy in orthodontic decision-making, with potential to reduce subjectivity and improve efficiency. Limitations include dataset size and single-center data. Future work will focus on multicenter validation, integration with ۳D imaging, and ethical AI auditing. Keywords: Artificial Intelligence, Orthodontics, Teeth Alignment, Neural

کلیدواژه ها:

Artificial Intelligence ، Orthodontics ، Teeth Alignment ، Neural