Artificial Intelligence-Driven Automated Incisor Root-Alveolar Bone Metric Measurement

To develop an artificial intelligence(AI)-driven system for precise, robust, and fully automated quantitative assessment of root-alveolar bone relationships in the incisor region on cone beam computed tomography (CBCT) images. This study employed a "segment-first, measure-later" strategy for automated root-alveolar bone analysis using deep learning. It involved constructing a dataset of 1,800 annotated CBCT sagittal images of maxillary/mandibular incisor regions. Three deep learning models (U-Net, ResUnet, TransUnet) were trained for segmentation, evaluated via the Intersection over Union (IoU), Dice Similarity Coefficient (DSC), and 95% Hausdorff Distance (HD95). The Gradient-weighted Class Activation Mapping (Grad-CAM) heatmaps visualized model decision-making rationales. The optimized segmentation model was integrated into a Python framework to enable automated quantification of labial/palatal alveolar bone height and thickness. Consistency between automated and manual measurements was validated using the Intraclass Correlation Coefficient (ICC) and Bland-Altman analysis. Additionally, the time required for processing and measuring images was compared between the automated and manual methods. Compared with manual measurements, the proposed ResUnet-based automated system exhibited excellent consistency (ICC=0.94) in the evaluation of palatal alveolar bone thickness in the maxillary incisor region. The system achieved a 97.7% reduction in computational time for assessing alveolar bone height and thickness on sagittal CBCT images. These findings demonstrated the potential clinical applicability of the AI-driven system in streamlining clinical workflows through automated metric quantification of root-alveolar bone relationships within the incisor region.