Related Experiment Video
Updated: Oct 9, 2025
10:23
Author Spotlight: Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
3.2K
Automatic Localization of Landmarks in Craniomaxillofacial CBCT Images Using a Local Attention-Based Graph
Yankun Lang1, Chunfeng Lian1, Deqiang Xiao1
1Department of Radiology and BRIC, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA.
Medical Image Computing and Computer-Assisted Intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention
|December 20, 2021
Summary
This study introduces a novel two-network system for precise craniomaxillofacial (CMF) landmark localization in CBCT scans. The method accurately digitizes anatomical points, aiding deformity assessment and surgical planning.
Area of Science:
- Medical Imaging
- Computer-Aided Surgery
- Artificial Intelligence in Medicine
Background:
- Accurate landmark localization is crucial for craniomaxillofacial (CMF) deformity quantification and reconstructive surgery planning.
- Automated localization of numerous CMF landmarks is challenging due to severe deformities and partial anatomical absence in cone-beam computed tomography (CBCT) images.
Purpose of the Study:
- To develop and evaluate a novel cascaded deep learning approach for the simultaneous digitization of 60 anatomical CMF landmarks.
- To improve the accuracy and robustness of landmark localization in CBCT scans of patients with CMF deformities.
Main Methods:
- A two-stage cascaded network architecture was proposed.
- The first stage employed a U-Net with a local attention mechanism to generate landmark heatmaps and extract features.
- The second stage utilized a graph convolution network for binary classification to determine landmark existence based on extracted features.
Main Results:
- The proposed method achieved an average detection error of 1.47mm on 50 CBCT scan datasets.
- A false positive rate of 19% was recorded.
- The approach demonstrated superior performance compared to existing state-of-the-art methods.
Conclusions:
- The cascaded network approach effectively digitizes a comprehensive set of CMF landmarks from CBCT images.
- This method shows significant potential for enhancing the quantitative analysis of CMF deformities and guiding surgical interventions.
