Journal of Applied Science and Engineering

Published by Tamkang University Press

ESCI jase impact factor scopus logo open access rate of Scopus journal

Addressing Low-Contrast Challenges: A Context-Aware Segmentation Algorithm for Robust 3D Lumbar Spine Reconstruction from Standard CT

 2026

 Volume 35

Xuemei Wang1,2

1School of Information Engineering, Shandong Huayu University of Technology, DeZhou, China, 253034

2Department of Information Technology, Pavel Sukhoi State Technical University of Gomel, Gomel, Belarus, 246029

Received: July 22, 2026
Accepted: August 24, 2026
Publication Date: September 25, 2026

上傳圖片

Architecture of 3D Attention U-Net for CT Image Segmentation

 Copyright The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are cited.

Download Citation: BibTeX | http://dx.doi.org/10.6180/jase.202612_35.070

Download PDF

Medical imaging through Computed Tomography (CT) performs critical functions because it provides essential support for both diagnosing and developing treatment plans for spinal disorders. The process of preoperative planning and surgical procedures depends on accurate lumbar spine segmentation together with 3D reconstruction of CT images. The process of lumbar spine segmentation faces major difficulties which arise from three factors: low-contrast areas, imaging artifacts, and different anatomical structures. The current methods are unable to solve these operational obstacles because they produce results with both reduced precision and diminished productivity. The research presents a new solution with its Context-Aware 3D Attention U-Net model which combines the Marine Predators Algorithm (MPA) for feature selection optimization, unnecessary elements while boosting learning efficiency. A hybrid loss function which combines include intensity normalization and resampling and Region of Interest (ROI) cropping to produce high-quality data input. The post-processing stage enhances segmentation output through processes which produce a seamless 3D reconstruction. The proposed model demonstrates superior performance compared to baseline methods while achieving a Dice score of 0.951 and an IoU of 0.907 and a precision of 0.963 and a recall of 0.944. The boundary error (HD95) measurement shows significant improvement because it decreased to 2.18 mm, which indicates better segmentation results and boundary detection performance. The method provides an accurate lumbar spine segmentation solution that people can use across various clinical settings which involve low-contrast CT scans.

Keywords: Lumbar spine segmentation, 3D reconstruction, Context-Aware 3D Attention U-Net, Marine Predators Algorithm, feature selection.

  1. [1] H. Xie, Z. Chen, J. Zhang, and T. Tan, (2025) “Multi-scale context U-net for breast cancer postoperative radiotherapy in patients with brachial plexus” BMC Medical Imaging 26(1): 8. DOI: https://doi.org/10.1186/s12880-025-02078-1.
  2. [2] W. Wang et al., (2024) “Transforaminal posterior lumbar interbody fusion microscopic safe operating area: a three-dimensional model study based on computed tomography imaging” Journal of Orthopaedic Surgery and Research 19(1): 342. DOI: https://doi.org/10.1186/s13018-024-04830-9.
  3. [3] A. Gubian et al., (2022) “CT-Navigated Spinal Instrumentations-Three-Dimensional Evaluation of Screw Placement Accuracy in Relation to a Screw Trajectory Plan” Medicina 58(9): 1200. DOI: https://doi.org/10.3390/medicina58091200.
  4. [4] A. Chan, E. Parent, J. Mahood, and E. Lou, (2022) “3D ultrasound navigation system for screw insertion in posterior spine surgery: a phantom study” International Journal of Computer Assisted Radiology and Surgery 17(2): 271-281. DOI: https://doi.org/10.1007/s11548-021-02516-9.
  5. [5] Y.-W. Jiang, X.-J. Xu, R. Wang, and C.-M. Chen, (2022) “Radiomics analysis based on lumbar spine CT to detect osteoporosis” European Radiology 32(11): 8019-8026. DOI: https://doi.org/10.1007/s00330-022-08805-4.
  6. [6] A. Poilliot, M. H.-P. Gay-Dujak, and M. Müller-Gerbl, (2023) “The quantification of 3D-trabecular architecture of the fourth cervical vertebra using CT osteoabsorptiometry and micro-CT” Journal of Orthopaedic Surgery and Research 18(1): 297. DOI: https://doi.org/10.1186/s13018-023-03760-2.
  7. [7] H. Frisk et al., (2024) “Automatic image registration on intraoperative CBCT compared to Surface Matching registration on preoperative CT for spinal navigation: accuracy and workflow” International Journal of Computer Assisted Radiology and Surgery 19(4): 665-675. DOI: https://doi.org/10.1007/s11548-024-03076-4.
  8. [8] B. Saravi et al., (2023) “Synthetic 3D Spinal Vertebrae Reconstruction from Biplanar X-rays Utilizing Generative Adversarial Networks” Journal of Personalized Medicine 13(12): 1642. DOI: https://doi.org/10.3390/jpm13121642.
  9. [9] R. Tao et al., (2026) “X2P-Net: Context-Aware 2D/3D Vertebra Localization” Bioengineering 13(2): 178. DOI: https://doi.org/10.3390/bioengineering13020178.
  10. [10] F. Hayat and Z. Yuqian, (2026) “MSAE-Unet: a multiscale attention-enhanced U-Net network for abdominal cancer segmentation using deep learning” Network Modeling Analysis in Health Informatics and Bioinformatics 15(1): 48. DOI: https://doi.org/10.1007/s13721-025-00684-1.
  11. [11] F. Abel et al., (2024) “Deep-learning reconstructed lumbar spine 3D MRI for surgical planning: pedicle screw placement and geometric measurements compared to CT” European Spine Journal 33(11): 4144-4154. DOI: https://doi.org/10.1007/s00586-023-08123-3.
  12. [12] D. M. Haida et al., (2024) “Hybrid-3D robotic suite in spine and trauma surgery – experiences in 210 patients” Journal of Orthopaedic Surgery and Research 19(1): 565. DOI: https://doi.org/10.1186/s13018-024-05044-9.
  13. [13] F. C. S. Altorfer et al., (2025) “Robotic pedicle screw placement with 3D MRI registration: moving towards radiation free robotic spine surgery” The Spine Journal 25(5): 1035-1041. DOI: https://doi.org/10.1016/j.spinee.2024.10.020.
  14. [14] A. Bayat et al., (2022) “Anatomy-Aware Inference of the 3D Standing Spine Posture from 2D Radiographs” Tomography 8(1): 479-496. DOI: https://doi.org/10.3390/tomography8010039.
  15. [15] CTSpine1K. 2026. URL: https://huggingface.co/datasets/alexanderdann/CTSpinelK/tree/main/raw_data/volumes (visited on 03/03/2026).
  16. [16] J. Chen et al., (2024) “TransUNet: Rethinking the U-Net architecture design for medical image segmentation through the lens of transformers” Medical Image Analysis 97: 103280. DOI: https://doi.org/10.1016/j.media.2024.103280.
  17. [17] F. Isensee, P. F. Jaeger, S. A. A. Kohl, J. Petersen, and K. H. Maier-Hein, (2021) “nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation” Nature Methods 18(2): 203-211. DOI: https://doi.org/10.1038/s41592-020-01008-z.