Related Experiment Video
Updated: Jul 10, 2025
06:18
Pedicle Screw Placement Using an Augmented Reality Head-Mounted Display in a Porcine Model
Published on: May 24, 2024
2.2K
SpineHRformer: A Transformer-Based Deep Learning Model for Automatic Spine Deformity Assessment with Prospective
Moxin Zhao1, Nan Meng1, Jason Pui Yin Cheung1
1Department of Orthopaedics and Traumatology, The University of Hong Kong, Hong Kong.
Bioengineering (Basel, Switzerland)
|November 25, 2023
Summary
SpineHRformer accurately measures the Cobb angle (CA) for spinal deformities using AI, outperforming existing methods in landmark detection and severity grading. This offers a more reliable and efficient clinical tool for assessing scoliosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Orthopedics
Background:
- Manual Cobb angle (CA) measurement for spinal deformity is time-consuming and prone to errors.
- Existing automated methods show potential but are limited by factors like deformity severity and image quality.
Purpose of the Study:
- Develop a robust, learning-based approach for accurate and consistent CA measurement from posteroanterior (PA) X-rays.
- Surpass the performance of current state-of-the-art automated methods.
Main Methods:
- Introduced SpineHRformer, a novel deep learning model utilizing HRNet and transformer blocks.
- The model identifies key anatomical landmarks (vertebral endplates C7-L5) to calculate CAs.
- Trained and tested on a dataset of 1934 PA X-rays with varying spinal deformities and image quality.
Main Results:
- SpineHRformer achieved superior landmark detection (2.47 vs. 2.74 pixels MED) and CA prediction (0.86 vs. 0.83 PCC) compared to SpineHRNet+.
- Demonstrated improved sensitivity in spinal deformity severity grading across normal-mild, moderate, and severe categories.
- Exhibited greater robustness and accuracy than the previous state-of-the-art method.
Conclusions:
- SpineHRformer offers a significant advancement in automated Cobb angle measurement.
- The model's enhanced accuracy and reliability hold substantial potential for clinical applications in spinal deformity assessment.
- This AI-driven approach promises to improve the efficiency and consistency of scoliosis evaluation.
