Related Experiment Video
Updated: Jul 19, 2025
10:23
Author Spotlight: Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
2.8K
Using a New Deep Learning Method for 3D Cephalometry in Patients With Hemifacial Microsomia.
Meng Xu1, Bingyang Liu2, Zhaoyang Luo3
1From the Cleft Lip and Palate Center.
Annals of Plastic Surgery
|August 11, 2023
Summary
This study introduces a novel deep learning approach for 3D cephalometry in hemifacial microsomia (HFM) patients. The graph convolutional neural network accurately predicts landmarks, offering a promising tool for HFM diagnosis and treatment planning.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Craniofacial Surgery
Background:
- Deep learning for 3D cephalometry shows promise in typical facial structures.
- Limited research exists on applying these AI methods to craniofacial deformities like hemifacial microsomia (HFM).
Purpose of the Study:
- To develop and evaluate a novel deep learning method for automatic 3D cephalometric landmark prediction in HFM patients.
- To assess the accuracy and reliability of a 3D point cloud graph convolutional neural network for HFM analysis.
Main Methods:
- Utilized a PointNet++ model for automatic 3D cephalometry on computed tomography (CT) scans of 135 HFM patients.
- Employed a graph convolutional neural network to analyze relationships between points for landmark localization.
- Evaluated accuracy using Mean Distance Error (MDE) and Success Detection Rate (SDR).
Main Results:
- The system achieved a mean MDE of 1.46 ± 1.308 mm across 32 landmarks.
- 10 landmarks demonstrated SDRs exceeding 90% within a 2 mm threshold.
- AI system's MDE showed lower standard deviation (0.67) and coefficient of variation (0.43) compared to manual methods.
Conclusions:
- The 3D cephalometry system based on graph convolutional networks demonstrates suitability for HFM cases.
- This AI approach offers accurate landmark prediction, potentially aiding in HFM diagnosis and treatment.
- Expanding the HFM training dataset may further enhance the system's accuracy.
