Related Experiment Video
Updated: Sep 23, 2025
10:23
Author Spotlight: Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
3.1K
Multi-input adaptive neural network for automatic detection of cervical vertebral landmarks on X-rays
Yuzhao Wang1, Lan Huang1, Minfei Wu2
1College of Computer Science and Technology, Jilin University, Changchun, 130000, China; Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education, Jilin University, Changchun, 130000, China.
Computers in Biology and Medicine
|May 16, 2022
Summary
This study introduces MultiIA-UNet, an accurate method for cervical vertebral landmark detection in X-rays. It improves cervical spine motion analysis for better disease diagnosis.
Area of Science:
- Medical Imaging
- Computer Vision
- Biomedical Engineering
Background:
- Cervical vertebral landmark detection is crucial for measuring cervical spine motion and diagnosing diseases.
- Challenges exist due to variations in patient poses and X-ray angles, leading to similar bone appearances.
Purpose of the Study:
- To develop an accurate and effective method for cervical vertebral landmark detection.
- To leverage similar local features across different cervical spine X-rays for improved detection accuracy.
Main Methods:
- Introduced Multi-input Adaptive U-Net (MultiIA-UNet) with an improved U-Net backbone and adaptive convolution module.
- Employed a multi-input strategy for simultaneous feature extraction and a subspace alignment module to learn similar local features.
Main Results:
- Achieved state-of-the-art performance with a minimum average point-to-point error of 12.988 pixels on a dataset of 688 X-rays.
- Demonstrated improved accuracy in cervical spine motion angle parameter measurement (minimum symmetric mean absolute percentage of 26.969%).
Conclusions:
- MultiIA-UNet is an efficient and accurate method for cervical vertebral landmark detection.
- The method enhances the reliability of cervical spine motion analysis for clinical diagnosis.
