Related Experiment Video
Updated: Jun 23, 2026
00:10
Precision Measurements and Parametric Models of Vertebral Endplates
Published on: September 17, 2019
6.5K
A Projective-Geometry-Aware Network for 3D Vertebra Localization in Calibrated Biplanar X-Ray Images.
Kangqing Ye1, Wenyuan Sun1, Rong Tao1
1Institute of Medical Robotics, School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China.
Sensors (Basel, Switzerland)
|February 26, 2025
Summary
A new projective-geometry-aware network, ProVLNet, improves 3D vertebra localization in biplanar X-rays. This method enhances accuracy for 3D navigation in spine surgery by utilizing projective geometry.
Area of Science:
- Medical Imaging
- Computer Vision
- Spine Surgery Navigation
Background:
- Current deep learning methods for vertebra localization in biplanar X-rays primarily use 2D information, neglecting projective geometry.
- This limitation hinders the accuracy of 3D navigation crucial for X-ray-guided spine surgery.
Purpose of the Study:
- To develop a 3D vertebra localization method using calibrated biplanar X-ray images that incorporates projective geometry.
- To improve the accuracy and reliability of 3D navigation in spine surgery.
Main Methods:
- Proposed ProVLNet, a projective-geometry-aware network for 3D vertebra localization.
- Utilized a Siamese 2D feature extractor, a spatial alignment fusion module for integrating projective geometry, and a 3D landmark regression module.
Main Results:
- Achieved high identification rates of 99.53% (lumbar) and 98.98% (thoracic) on challenging datasets.
- Demonstrated low point-to-point errors of 0.64 mm (lumbar) and 1.38 mm (thoracic).
- Outperformed state-of-the-art methods in 3D vertebra localization.
Conclusions:
- ProVLNet effectively localizes 3D vertebrae in calibrated biplanar X-ray images by leveraging projective geometry.
- The proposed method offers superior performance for accurate 3D navigation in X-ray-guided spine surgery.
