Search papers, labs, and topics across Lattice.
This study introduces CephViT, a Vision Transformer-based model designed for the automated localization of 2D lateral cephalometric landmarks, addressing the challenges of manual annotation in craniofacial assessments. The model was trained on a public dataset and demonstrated a mean radial error of 1.28 mm with a detection rate of 92.0%, showcasing its effectiveness in accurately identifying landmarks. Furthermore, the classification of skeletal malocclusion using landmarks derived from digitally reconstructed radiographs (DRRs) achieved comparable accuracy to that of manually annotated references, indicating the model's practical applicability in clinical settings.
Automated landmark localization can achieve nearly identical classification accuracy for skeletal malocclusion as traditional manual methods, streamlining craniofacial assessments.
Manual cephalometric landmark annotation is important for craniofacial assessment but is labor-intensive and difficult to scale. We introduce CephViT, a Vision Transformer-based model for automated 2D lateral cephalometric landmark localization, and evaluate its use in downstream skeletal malocclusion classification. CephViT was trained and benchmarked on a public lateral cephalogram dataset, achieving a mean radial error of 1.28 +/- 1.42 mm and a successful detection rate of 92.0% at 3.0 mm. Because the private evaluation cohort consisted of 3D CBCT scans, lateral cephalogram-like digitally reconstructed radiographs (DRRs) were generated from each volume and used as 2D inputs to the landmark localization model. Landmark coordinates were normalized into a common coordinate frame, and skeletal malocclusion classification was performed using landmarks shared between the reference and DRR-based pipelines. Classification performance using DRR-localized landmarks was comparable to that obtained using manually annotated reference landmarks, with accuracies of 70.0% and 68.3%, respectively. These results support the feasibility of automated cephalometric analysis on CBCT-derived DRRs for skeletal malocclusion assessment.
