Related Experiment Video
Updated: Jan 6, 2026
10:23
Author Spotlight: Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
3.6K
Accuracy comparative study of automatic landmarking and diagnostic models on lateral cephalograms
Wen-Qing Bu1,2, Zhan-Yi Shi3, Zhi-Qiang Tian3
1Key Laboratory of Shaanxi Province for Craniofacial Precision Medicine Research, College of Stomatology, Xi'an Jiaotong University, Xi'an, China.
Progress in Orthodontics
|November 27, 2025
Summary
Deep learning in cephalometric analysis shows automatic landmarking models outperform diagnostic models. While landmarking offers precision, diagnostic models provide speed, suggesting hybrid approaches for clinical needs.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Dental Diagnostics
Background:
- Deep learning is increasingly used in cephalometric analysis.
- Current automatic landmarking models require clinical validation.
- Automatic diagnostic models show promise but lack comparative evidence.
Purpose of the Study:
- To compare the performance, generalization, and transferability of automatic landmarking and diagnostic models.
- To evaluate model accuracy on external datasets.
Main Methods:
- Two deep learning models were evaluated: automatic landmarking and automatic diagnostic.
- Performance was assessed using northern Chinese population data and the IEEE ISBI 2015 Grand Challenge dataset.
Main Results:
- The automatic landmarking model demonstrated superior performance.
- Accuracy reached 90.80% on the IEEE dataset for the landmarking model.
Conclusions:
- Automatic landmarking models provide precise measurements.
- Automatic diagnostic models offer faster results.
- Hybrid models could combine the strengths of both approaches for clinical application.
