medigraphic.com
ISSN 2992-7749 (Electronic)
ISSN 2992-7897 (Print)
Órgano Oficial de difusión científica de la Asociación Mexicana de Cirujanos de Columna A. C. y de la Sociedad Iberolatinoamericana de Columna. SILACO.

Indizada en:

Scielo

2026, Number 3

<< Back Next >>

Cir Columna 2026; 4 (3)

Development of an AI-assisted semi-automated method for Hounsfield unit measurement in the lumbar spine. Preliminary study

Garay AL, Bazán PL, Cinalli M, Pérez GA

Language: Spanish
References: 9
Page: 209-214
PDF size: 622.28 Kb.


ABSTRACT

Introduction: osteoporosis is a global health issue, and limited access to dual-energy X-ray absorptiometry (DXA) hinders its timely diagnosis. The measurement of Hounsfield units (HU) in computed tomography (CT) has emerged as a complementary tool for estimating bone quality, particularly in settings where DXA is unavailable. Concurrently, the use of artificial intelligence (AI) in medical imaging has grown exponentially; however, automated models are often limited by implementation difficulties and scarce external validation. In this scenario, AI-assisted semi-automated methods could represent a viable alternative, although evidence remains limited. The objective of this study was to develop an AI-assisted semi-automated method to measure Hounsfield units in lumbar vertebrae and to assess its feasibility and agreement with manual measurement. Material and methods: the system was implemented in Python and included a preprocessing module and an automated analysis module that generated an elliptical region of interest (ROI), adjusted in a standardized manner to ≤ 2 mm from the cortical bone. Results: following a pilot test that allowed for the correction of segmentation errors, 36 vertebrae (L1-L4) were processed. The method demonstrated high correlation with manual analysis (r = 0.956), excellent agreement (ICC = 0.943), and reduced bias (+6.8 HU), with no proportional bias observed according to the Bland-Altman analysis. Conclusions: the semi-automated method yielded results equivalent to those of manual analysis, showing high agreement and minimal differences. These findings support its further investigation in studies with larger sample sizes.


REFERENCES

  1. Maeda SS, LLibre RDS, Arantes HP, de Souza GC, Molina FFC, Wiluzanski D, et al. Challenges and opportunities for quality densitometry in Latin America. Arch Osteoporos. 2021; 16: 23. doi: 10.1007/s11657-021-00892-y.

  2. Clynes MA, Westbury LD, Dennison EM, Kanis JA, Javaid MK, Harvey NC, et al. Bone densitometry worldwide: a global survey by the ISCD and IOF. Osteoporos Int. 2020; 31: 1779-1786. doi: 10.1007/s00198-020-05435-8.

  3. Bazán PL, Jordan RC, Cevallos JRM, Adaro AOG, Adaro JCS. Hounsfield units use in spinal surgery planning: Systematic review and meta-analysis. Coluna/Columna. 2022; 21: e264579. Available in: https://doi.org/10.1590/s1808-185120222103264579

  4. St Jeor JD, Jackson TJ, Xiong AE, Freedman BA, Sebastian AS, Currier BL, et al. Average lumbar Hounsfield units predicts osteoporosis-related complications following lumbar spine fusion. Global Spine J. 2022; 12: 851-857. doi: 10.1177/2192568220975365.

  5. Ahern DP, McDonnell JM, Riffault M, Evans S, Wagner SC, Vaccaro AR, et al. A meta-analysis of the diagnostic accuracy of Hounsfield units on computed topography relative to dual-energy X-ray absorptiometry for the diagnosis of osteoporosis in the spine surgery population. Spine J. 2021; 21: 1738-1749. doi: 10.1016/j.spinee.2021.03.008.

  6. Liu RW, Ong W, Makmur A, Kumar N, Low XZ, Shuliang G, et al. Application of artificial intelligence methods on osteoporosis classification with radiographs-a systematic review. Bioengineering (Basel). 2024; 11: 484. doi: 10.3390/bioengineering11050484.

  7. Ong W, Liu RW, Makmur A, Low XZ, Sng WJ, Tan JH, et al. Artificial intelligence applications for osteoporosis classification using computed tomography. Bioengineering (Basel). 2023; 10: 1364. doi: 10.3390/bioengineering10121364.

  8. Park H, Kang WY, Woo OH, Lee J, Yang Z, Oh S. Automated deep learning-based bone mineral density assessment for opportunistic osteoporosis screening using various CT protocols with multi-vendor scanners. Sci Rep. 2024; 14: 25014. doi: 10.1038/s41598-024-73709-w.

  9. Deng Y, Chen X, Yao N, Geng C, Yuan C, Chen Q, et al. Two-decade dialogue between artificial intelligence and osteoporosis: research trajectories and frontier projections under bibliometric and visual analysis. Front Med (Lausanne). 2025; 12: 1606361. doi: 10.3389/fmed.2025.1606361.




2020     |     www.medigraphic.com