Automatic Cephalometric Landmark Localization on CBCT-Derived Digitally Reconstructed Radiographs for Skeletal Malocclusion Classification

Benjamin Hou; Konstantinia Almpani; Janice S. Lee; Zhiyong Lu

cs.CVAug 17, 2026

Automatic Cephalometric Landmark Localization on CBCT-Derived Digitally Reconstructed Radiographs for Skeletal Malocclusion Classification

Authors: Benjamin Hou, Konstantinia Almpani, Janice S. Lee, Zhiyong Lu

Organizations: Division of Intramural Research, National Library of Medicine · Craniofacial Anomalies and Regeneration Section, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD, USA.

Abstract

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.

Figures & tables

Panel of Figure 1 — Figure 1: Overview of the study pipeline. From L-to-R: CBCT volume, lateral DRR, cephalometric landmark localization using CephViT, landmark heatmap output. The localized landmark coordinates were subsequently normalized and used for downstream skeletal malocclusion classification.

Method	MRE $\pm$ SD (mm)	SDR (%)
		2.0 mm	2.5 mm	3.0 mm	4.0 mm
Khan et al. [ 12 ]	1.92 $\pm$ 7.85	78.54	85.72	89.64	94.49
Khalid et al. [ 9 ]	1.87 $\pm$ 4.01	75.17	82.43	88.78	93.01
Khan et al. [ 11 ]	1.69 $\pm$ 3.36	81.18	87.28	90.82	94.82
CephViT (ours)	1.28 $\pm$ 1.42	82.10	87.90	92.00	95.70

Table 1: Cephalometric landmark detection performance on the Aariz test set.

Figure 2: Projection and normalization of CBCT-derived reference landmarks. Left; 3D CBCT landmark configuration and lateral projection onto a 2D cephalometric plane. Right; all projected subjects after 3D anatomical normalization, with mean landmark positions shown in red.

Figure 3: 2D normalization of DRR-localized cephalometric landmarks before (left) and after (right) similarity-based alignment to a common coordinate system.

	Support	Reference landmarks	DRR-localized landmarks	$\Delta$	95% CI for $\Delta$	Paired $p$
Accuracy	300	0.683	0.700	+0.017	[-0.050, 0.083]	0.696
Macro F1	300	0.642	0.647	+0.005	[-0.066, 0.074]	0.894
Weighted F1	300	0.688	0.708	+0.020	[-0.045, 0.084]	0.555
Class I	99	0.667	0.691	+0.024	[-0.071, 0.119]	–
Class II	99	0.768	0.794	+0.026	[-0.052, 0.107]	–
Class IIIA	27	0.431	0.364	-0.067	[-0.223, 0.080]	–

Table 2: Five-fold cross-validated malocclusion classification.

Panel of Figure 6 — Figure S1 : Craniofacial malocclusion classes

Explore similar work

Morphological Decoupling-Based Skeletal Classification for Clinical Assessment of Malocclusion
Sep 9, 2026Zhichun Jin, Zhicheng He, Hao Xu +4Graph Neural Networks Occlusion Reasoning
The Affiliated Stomatological Hospital of Nanjing Medical University, Nanjing, China · State Key Laboratory Cultivation Base of Research, Prevention and Treatment for Oral Diseases, Nanjing Medical University, Nanjing, China · Jiangsu Province Engineering Research Center of Stomatological Translational Medicine, Nanjing, China +1
Optimising for the Unknown: Domain Alignment for Cephalometric Landmark Detection
Oct 6, 2024Julian Wyatt, Irina VoiculescuMedical Image Localization Domain Adaptation
Department of Computer Science, University of Oxford
LM-PCVMNet: Pediatric Cervical Vertebral Maturation Analysis with Deep Fusion of Landmarks and Metadata
Sep 11, 2026Peng Wang, Wanzhen Song, Anli Wang +3Medical Imaging Medical Image Classification
College of Cryptology and Cyber Science, Nankai University, Tianjin, 300350, China · College of Engineering, Yanbian University, Yanji, 133002, Jilin, China · Haihe Lab of ITAI, Tianjin, 300350, China +1

Sep 9, 2026J/K move · Enter open · S save

Morphological Decoupling-Based Skeletal Classification for Clinical Assessment of Malocclusion

Zhichun Jin, Zhicheng He, Hao Xu, Dongyang Li+3

The Affiliated Stomatological Hospital of Nanjing Medical University, Nanjing, China · State Key Laboratory Cultivation Base of Research, Prevention and Treatment for Oral Diseases, Nanjing Medical University, Nanjing, China · Jiangsu Province Engineering Research Center of Stomatological Translational Medicine, Nanjing, China+1

Graph Neural Networks Occlusion Reasoning Multimodal Classification Multimodal Learning

Open paper AskPDF arXiv

Malocclusion skeletal grading is a fundamental task in orthodontics, critical for diagnosis and treatment planning. Traditionally, cone-beam computed tomography (CBCT) is used for visual measurement, and the reconstructed lateral cephalograms are handed over to expert dentists for diagnosis. However, manual review is time-consuming, labor-intensive, and subject to inter-operator variability. Therefore, an automatic CBCT-based system is needed for reliable malocclusion skeletal grading. In this case, we develop TeethGNN, a novel graph-based framework designed to combine CBCT image features with morphological information for accurate and efficient malocclusion grading. TeethGNN utilizes a decoupled learnable decoder to directly predict key morphological indicators from CBCT images, eliminating the need for manual measurements. These morphological features are then fused with image features using a graph neural network (GNN), which effectively models the relationships between the modalities. To further enhance robustness and calibration, we introduce a collaborative calibration strategy. This strategy combines multi-scale graph adversarial perturbation for explicit calibration and nonlinear topological graph calibration for implicit confidence adjustment. Extensive experiments and ablation studies on our collected clinical dataset demonstrate that our malocclusion measurement system achieves 77.08% in accuracy and 89.61% in AUC, outperforming the compared state-of-the-art methods. These results validate the effectiveness of graph-based multimodal fusion and collaborative calibration in improving malocclusion grading performance. Our system shows strong potential for advancing computer-aided orthodontic diagnosis, providing an accurate and reliable solution for vision-based clinical measurement and diagnosis.

Abstract

Explore similar work

Morphological Decoupling-Based Skeletal Classification for Clinical Assessment of Malocclusion

Optimising for the Unknown: Domain Alignment for Cephalometric Landmark Detection

LM-PCVMNet: Pediatric Cervical Vertebral Maturation Analysis with Deep Fusion of Landmarks and Metadata