Related Experiment Video
Updated: Aug 6, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
Automated Adolescence Scoliosis Detection Using Augmented U-Net With Non-square Kernels
Yujie Wu1, Khashayar Namdar2,3,4, Chaojun Chen1
1Department of Mechanical and Industrial Engineering, University of Toronto, Toronto, ON, Canada.
Summary
This study introduces an automated deep learning method for measuring the Cobb angle in scoliosis patients from X-rays. The new system improves accuracy and efficiency in diagnosing spinal deformities.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Orthopedics
Background:
- Scoliosis diagnosis relies on Cobb angle measurement, traditionally manual and prone to variability.
- Existing automated methods lack sufficient accuracy for clinical use.
- Accurate Cobb angle measurement is crucial for timely scoliosis treatment.
Purpose of the Study:
- To develop and validate a deep learning architecture for automated Cobb angle measurement.
- To improve the accuracy and efficiency of scoliosis assessment using X-ray images.
- To provide a robust tool for clinical settings.
Main Methods:
- A two-step deep learning architecture using Augmented U-Net for vertebral segmentation.
- A subsequent non-learning pipeline for landmark extraction and filtering.
- Validation on the AASCE-MICCAI challenge 2019 dataset and an internal dataset.
Main Results:
- The proposed method achieved a Symmetric Mean Absolute Percentage Error of 9.2%.
- Approximately 90% of estimations were within 10 degrees of ground truth on the challenge dataset.
- Comparable performance was observed on an internal validation dataset.
Conclusions:
- The developed architecture offers robust automated spinal vertebrae segmentation and Cobb angle measurement.
- The system demonstrates potential for generalization in real-world clinical applications.
- This automated approach can enhance the accuracy and efficiency of scoliosis assessment.
