Osteoporosis is a bone disease affecting both postmenopausal women and older men. Bone fractures caused by osteoporosis are a major health concern, creating a great economic burden to our society. Bone mineral density (BMD), a measure of bone mass by Dual-energy X-ray Absorptiometry (DXA), is a major risk factor for bone fractures. In addition to BMD, trabecular microarchitecture also contributes to bone strength and therefore is a risk factor for osteoporotic fractures. Recently, stochastic predictors derived from DXA scans have been found to correlate with trabecular microarchitecture in human vertebrae. In routine clinical scans of the human lumbar spine, posterior elements are always included during the posterior-anterior (PA) projection of DXA scans. To our knowledge, the influence of posterior elements on the relationship between stochastic predictors and trabecular microarchitecture has not been investigated. Therefore, the objective of this study is to examine the effect of posterior elements on the estimation of stochastic predictors using simulated DXA scans.

3D MicroCT images of human vertebrae from the lumbar spine of five tissue donors were obtained. Simulated DXA images of human vertebrae with and without posterior elements were generated from these 3D MicroCT images. Stochastic parameters such as correlation length and sill variance were calculated by fitting a theoretical model onto the experimental variogram of simulated DXA images. Linear regression analyses were performed to examine the correlations between microarchitecture of trabecular bone and stochastic predictors from DXA images of human vertebrae with and without posterior elements.

The sill variance of simulated DXA images without posterior elements was positively correlated with some of the microarchitecture parameters such as bone surface to volume ratio, trabecular separation, and negatively correlated with bone volume fraction, trabecular thickness, trabecular number. The sill variance of simulated DXA images of whole vertebrae was positively correlated with bone volume fraction, trabecular thickness, trabecular number, and connectivity density, and negatively correlated with the bone surface to volume ratio, trabecular separation. Although these correlations were not statistically significant, the correlations between the sill variance and microarchitecture parameters were mostly greater in the vertebral body without posterior elements than the whole vertebrae with intact posterior elements.

The outcome of this study indicates that it is necessary to remove posterior elements from DXA scans of the lumbar spine to improve the prediction of bone fractures using stochastic predictors