Artificial Intelligence Frameworks for CBCT Image Processing in Clear Aligner Fabrication: Mini Review

Abstract

Clear aligner therapy increasingly relies on accurate digital models to improve the predictability of complex tooth movements; however, conventional workflows based on intraoral scans lack root and alveolar bone information that is critical for biomechanical planning and risk assessment. Cone-beam computed tomography (CBCT) can provide this anatomical detail but is limited in routine use by time-consuming segmentation, artifact management, and registration processes. This mini review synthesizes current evidence on artificial intelligence (AI), particularly deep learning–based frameworks, applied to CBCT image processing for clear aligner fabrication and digital orthodontic workflows. A targeted literature search identified studies evaluating AI-driven CBCT segmentation, multimodal fusion with intraoral scans, artifact handling, and clinically relevant applications such as root-aware planning, midpalatal suture maturation staging, and automated assessment of orthodontically induced root resorption. Across predominantly retrospective and laboratory-based studies published between 2021 and 2025, deep learning models—most commonly U-Net–based architectures—demonstrated high segmentation accuracy (often exceeding 90%) while substantially reducing processing time from hours to minutes. Multimodal CBCT–intraoral scan fusion emerged as a key advance for generating anatomically complete crown–root–bone models that may enhance aligner planning and monitoring. Despite promising technical performance, clinical translation remains constrained by small datasets, heterogeneous reference standards, limited external validation, and a lack of prospective outcome-focused studies. Overall, AI-enabled CBCT processing shows strong potential to streamline digital orthodontic workflows and improve anatomical fidelity in clear aligner therapy, but further multi-center validation and clinical effectiveness studies are required before widespread adoption.