Article: Spinal segmentation algorithm for modelling Chinese digital human models Journal: International Journal of Vehicle Systems Modelling and Testing (IJVSMT) 2025 Vol.19 No.4 pp.335 - 352 Abstract: Low-dose spinal CT images often suffer from issues such as blurred boundaries, significant noise, and poor contrast, which complicate manual segmentation. Traditional spinal image segmentation algorithms, although fast, generally lack precision and require manual intervention. Meanwhile, deep learning-based methods require extensive datasets for support, limiting their widespread applicability. To overcome these limitations, this paper introduces the 3D-TSUnet, this method first employs traditional segmentation algorithms for pre-segmentation, followed by detailed segmentation using the refined 3D-Unet network. Comparisons with manual segmentation show a 98.28% reduction in self-intersections, 95.05% decrease in highly refractive edges, 89.59% reduction in spikes, and 96.48% reduction in incorrect partitions, with segmentation time reduced by 91.67%. These results demonstrate that the proposed network efficiently performs low-dose CT spinal segmentation, offering substantial practical value for developing Chinese human finite element models and advancing related research. Inderscience Publishers - linking academia, business and industry through research

Title: Spinal segmentation algorithm for modelling Chinese digital human models

Authors: Hongji Xiong; Cheng Chen; Yu Liu; Xiaofan Wu; Zhonghao Bai

Addresses: State Key Laboratory of Advanced Design and Manufacturing Technology for Vehicle, Hunan University, YueLu District, Changsha, 410082, China ' Products and Commodities Evaluation Center, Administration for Market Regulation of Hunan Province, TianXin District, Changsha, 410004, China ' Automotive Safety Technology Center, China Automotive Engineering Research Institute Co. Ltd., Yubei, Chongqing, 401122, China ' Automotive Safety Technology Center, China Automotive Engineering Research Institute Co. Ltd., Yubei, Chongqing, 401122, China ' State Key Laboratory of Advanced Design and Manufacturing Technology for Vehicle, Hunan University, YueLu District, Changsha, 410082, China

Abstract: Low-dose spinal CT images often suffer from issues such as blurred boundaries, significant noise, and poor contrast, which complicate manual segmentation. Traditional spinal image segmentation algorithms, although fast, generally lack precision and require manual intervention. Meanwhile, deep learning-based methods require extensive datasets for support, limiting their widespread applicability. To overcome these limitations, this paper introduces the 3D-TSUnet, this method first employs traditional segmentation algorithms for pre-segmentation, followed by detailed segmentation using the refined 3D-Unet network. Comparisons with manual segmentation show a 98.28% reduction in self-intersections, 95.05% decrease in highly refractive edges, 89.59% reduction in spikes, and 96.48% reduction in incorrect partitions, with segmentation time reduced by 91.67%. These results demonstrate that the proposed network efficiently performs low-dose CT spinal segmentation, offering substantial practical value for developing Chinese human finite element models and advancing related research.

Keywords: 3D-TSUnet; medical image segmentation; low-dose spinal CT images; supervised learning; Chinese human finite element model.

DOI: 10.1504/IJVSMT.2025.150169

International Journal of Vehicle Systems Modelling and Testing, 2025 Vol.19 No.4, pp.335 - 352

Received: 17 Jan 2025
Accepted: 21 Apr 2025
Published online: 02 Dec 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Spinal segmentation algorithm for modelling Chinese digital human models

Keep up-to-date