Related Experiment Video
Updated: May 10, 2026
Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
A new automated 3d facial soft tissue landmarking method via deep learning
Han Bao1, Zhidong He2, Jiasong Wu3
1Department of Orthodontics, The Affiliated Stomatological Hospital of Nanjing Medical University, Nanjing 210029, China; State Key Laboratory Cultivation Base of Research, Prevention and Treatment for Oral Diseases (Nanjing Medical University), Nanjing 210029, China; Jiangsu Province Engineering Research Center of Stomatological Translational Medicine (Nanjing Medical University), Nanjing 210029, China.
This study introduces FST-Net, a deep learning tool for automatically detecting 3D facial soft tissue landmarks and measurements. FST-Net shows promise in improving accuracy and efficiency for orthodontic diagnosis and treatment planning.
Area of Science:
- Medical Imaging
- Computer Vision
- Orthodontics
Background:
- Accurate 3D facial soft tissue landmark detection is crucial for orthodontic diagnosis and treatment planning.
- Manual landmarking is time-consuming and prone to inter-observer variability.
Purpose of the Study:
- To propose and evaluate a novel deep learning (DL)-based facial soft tissue network (FST-Net).
- To automate the detection of 34 commonly used 3D soft tissue landmarks and eight measurements.
- To assess the accuracy and efficiency of FST-Net compared to manual methods.
Main Methods:
- A deep learning model (FST-Net) was developed using feature fusion and local coordinate regression.
- 297 patient cases were randomly allocated into training, validation, and testing sets.
- The model's performance was evaluated based on landmark localization error and measurement accuracy.
Main Results:
- The mean radial error (MRE) for all 34 landmarks was 1.80 ± 1.81 mm.
- High operator reproducibility landmarks (Group 1) achieved an MRE of 1.25 ± 0.90 mm.
- Successful detection rates (SDRs) within 4 mm were 90.54%, and the average intra-class correlation coefficient (ICC) for measurements was 0.909.
Conclusions:
- FST-Net demonstrates significant potential for accurate and efficient 3D soft tissue landmark detection in orthodontics.
- The DL approach offers improvements for diagnosis and treatment planning, reducing manual workload.
- Further optimization is needed for landmarks dependent on subjective judgment (Group 2).
