Scholarlink

Objective <p>Pediatric scoliosis is the most prevalent spinal disorder, often leading to abnormal curvature and deformation of the spine. Early detection is essential for timely intervention, particularly in growing adolescents. In this study, we present a novel, fully automated, radiation-free method for Cobb angle evaluation, combining fringe projection profilometry with deep learning technologies.</p> Materials and methods <p>A three-dimensional reconstruction of the participant’s back surface is achieved using a seven-step phase-shifting algorithm. Convolutional neural networks are then utilized to extract asymmetry features from the 3D surface and predict the Cobb angle, a key clinical indicator of scoliosis severity. A total of 48 participants clinically diagnosed with scoliosis based on radiographic imaging were recruited from the hospital.</p> Results <p>The experimental results demonstrate a strong correlation between the predicted and actual Cobb angles, with a correlation coefficient of 0.94 and a coefficient of determination of 0.8796 during Adam’s forward bend test. The mean time required from scanning to Cobb angle prediction is approximately 3.3&#xa0;s.</p> Conclusions <p>The proposed evaluation method exhibits excellent discriminative capability and shows significant potential as an alternative to the traditional scoliometer for large-scale Cobb angle screening programs in schools.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automatic radiation-free evaluation of Cobb angle for spinal curvature based on fringe projection profilometry and deep learning technology

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月15日

查看全文
  • Chi-Kuang Feng,
  • Ching-Jung Hung,
  • Yen-Ju Chen,
  • Pei-Yu Su,
  • Guan-Ting Liu,
  • Cheng-Yang Liu

摘要

Objective

Pediatric scoliosis is the most prevalent spinal disorder, often leading to abnormal curvature and deformation of the spine. Early detection is essential for timely intervention, particularly in growing adolescents. In this study, we present a novel, fully automated, radiation-free method for Cobb angle evaluation, combining fringe projection profilometry with deep learning technologies.

Materials and methods

A three-dimensional reconstruction of the participant’s back surface is achieved using a seven-step phase-shifting algorithm. Convolutional neural networks are then utilized to extract asymmetry features from the 3D surface and predict the Cobb angle, a key clinical indicator of scoliosis severity. A total of 48 participants clinically diagnosed with scoliosis based on radiographic imaging were recruited from the hospital.

Results

The experimental results demonstrate a strong correlation between the predicted and actual Cobb angles, with a correlation coefficient of 0.94 and a coefficient of determination of 0.8796 during Adam’s forward bend test. The mean time required from scanning to Cobb angle prediction is approximately 3.3 s.

Conclusions

The proposed evaluation method exhibits excellent discriminative capability and shows significant potential as an alternative to the traditional scoliometer for large-scale Cobb angle screening programs in schools.