Hounsfield unit measurements from different regions of interest in lumbar spine CT: correlation with DEXA-derived bone mineral density and fracture status.
TL;DR
Different lumbar CT ROI methods demonstrated comparable performance in assessing BMD and discriminating thoracolumbar fragility fracture status, and sensitivity analysis confirmed that exclusion of substituted vertebral levels did not materially affect the results.
Abstract
Objective
This study compared Hounsfield unit (HU) measurements obtained from different regions of interest (ROIs) on lumbar CT for assessing bone mineral density (BMD) and discriminating osteoporosis and thoracolumbar fragility fracture status, aiming to evaluate whether ROI selection meaningfully influences HU-based assessment.
Methods
Five HU measurement methods were evaluated using lumbar CT. Their correlations with dual-energy X-ray absorptiometry (DEXA)-derived BMD and their discriminative performance for osteoporosis and thoracolumbar fragility fracture status were analyzed. Receiver operating characteristic (ROC) curves were constructed, and pairwise comparisons of areas under the curve (AUCs) were performed using the DeLong test. A sensitivity analysis was additionally conducted after excluding substituted vertebral levels and recalculating mean lumbar HU using only non-substituted vertebrae.
Results
All five HU measurement methods showed significant correlations with DEXA-derived BMD, and no statistically significant differences in discriminative performance were observed among the five ROI methods in the primary analysis. In the sensitivity analysis excluding substituted vertebral levels, the AUCs remained comparable across methods for osteoporosis (0.816-0.826) and thoracolumbar fragility fracture status (0.677-0.689), with all pairwise DeLong tests showing no statistically significant differences (all p > 0.05).
Conclusion
Different lumbar CT ROI methods demonstrated comparable performance in assessing BMD and discriminating thoracolumbar fragility fracture status, and sensitivity analysis confirmed that exclusion of substituted vertebral levels did not materially affect the results. No ROI method showed statistically significant superiority, suggesting that ROI selection may be guided by clinical feasibility and workflow convenience rather than by diagnostic performance alone. ADVANCES IN KNOWLEDGE This study compares five commonly used lumbar CT HU measurement methods and shows that their performance is broadly similar. The findings suggest that ROI selection may be guided primarily by clinical feasibility and workflow convenience, although further external validation is required before routine clinical application.