Related Experiment Video
Updated: Nov 21, 2025
10:23
Author Spotlight: Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
3.3K
Automatic Cephalometric Landmark Identification System Based on the Multi-Stage Convolutional Neural Networks with
Min-Jung Kim1, Yi Liu2, Song Hee Oh3
1Department of Orthodontics, Graduate School, Kyung Hee University, Seoul 02447, Korea.
Sensors (Basel, Switzerland)
|January 15, 2021
Summary
This study developed an automated cephalometry system using multi-stage convolutional neural networks (CNNs). The system achieved high accuracy in identifying 15 landmarks on combined cephalogram datasets.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Radiology
Background:
- Cephalometry is crucial for diagnosing and treating craniofacial abnormalities.
- Manual landmark identification is time-consuming and subject to inter-observer variability.
- Automated systems can improve efficiency and consistency in cephalometric analysis.
Purpose of the Study:
- To develop and validate a fully automated cephalometry landmark identification system.
- To leverage multi-stage convolutional neural networks (CNNs) for enhanced accuracy.
- To assess the system's performance using a combined dataset of synthesized cephalograms.
Main Methods:
- A multi-stage CNN architecture was employed for landmark identification.
- A combined dataset was created using 430 lateral and 430 MIP lateral cephalograms synthesized from cone-beam computed tomography (CBCT) data.
- Fifteen standard cephalometric landmarks were manually identified for training and validation, achieving high intra-examiner reliability (ICC = 0.99).
Main Results:
- The automated system demonstrated a mean radial error (MRE) of 1.03 mm.
- The standard deviation (SD) of the landmark identification was 1.29 mm.
- The system's prediction accuracy was evaluated on the synthesized CBCT-derived cephalograms.
Conclusions:
- The developed automated cephalometry system shows promising accuracy for landmark identification.
- The type of image data used may influence the prediction accuracy of CNN-based automated systems.
- Further research is needed to optimize performance across diverse imaging modalities.
