Osteoporos Sarcopenia.  2024 Mar;10(1):28-34. 10.1016/j.afos.2024.01.002.

Discordance between dual-energy X-ray absorptiometry bone mineral density and spinal computed tomography texture analysis: An investigation into low correlation rates

• Kim MW
• Noh YM
• Huh JW
• Seo HE
• Lee DH

• KMID: 2554457
• DOI: http://doi.org/10.1016/j.afos.2024.01.002

Abstract

Objectives
This research delves into the application of texture analysis in spine computed tomography (CT) scans and its correlation with bone mineral density (BMD), as determined by dual-energy X-ray absorptiometry (DXA). It specifically addresses the discordance between the 2 measurements, suggesting that certain spinal-specific factors may contribute to this discrepancy.
Methods
The study involved 405 cases from a single institution collected between May 6, 2012 and June 30, 2021. Each case underwent a spinal CT scan and a DXA scan. BMD values at the lumbar region (T12 to S1) and total hip were recorded. Texture features from axial cuts of T12 to S1 vertebrae were extracted using gray-level co-occurrence matrices, and a regression model was constructed to predict the BMD values.
Results
The correlation between CT texture analysis results and BMD from DXA was moderate, with a correlation coefficient ranging between 0.4 and 0.5. This discordance was examined in light of factors unique to the spine region, such as abdominal obesity, aortic calcification, and lumbar degenerative changes, which could poten tially affect BMD measurements.
Conclusions
Emerging from this study is a novel insight into the discordance between spinal CT texture analysis and DXA-derived BMD measurements, highlighting the unique influence of spinal attributes. This revelation calls into question the exclusive reliance on DXA scans for BMD assessment, particularly in scenarios where DXA scanning may not be feasible or accurate.

Keyword

Dual-energy X-ray absorptiometry (DXA); Computed tomography hounsfield unit (CT HU); Bone mineral density (BMD); Morphometric texture analysis; Linear regression