A new approach for quantifying localized bone loss by measuring void spaces.

The application of high resolution peripheral quantitative computed tomography (HR-pQCT) for the study of bone health has provided valuable insight into the role bone microarchitecture has in determining bone strength and fracture risk. However, conventional density and morphological parameters struggle to distinguish whether localized bone loss is present, visible as heterogeneous deterioration in the trabecular network. This is because current HR-pQCT parameters quantify a global average of properties in the cortical or trabecular compartment. This study proposes a new metric we term "void space" that segments volumes of localized deterioration in the trabecular bone network from HR-pQCT scans and quantifies void space as the void space to total volume ratio (VS/TV, %). A simple and fully automated protocol for segmenting and quantifying void space in HR-pQCT scans is presented, along with the assessment of accuracy, precision, and cross-calibration between generations of HR-pQCT systems. Finally, prevalence of void space and the association with standard HR-pQCT parameters is demonstrated using a large population-based cohort (n = 1236). Void space detection was found to be highly reproducible (accuracy >95%, least significant change <1.76% VS/TV) and correlation between scanner generations was strong (R2 = 0.87), although the first generation system struggled to identify small voids. Assessment of void space prevalence in the population-based cohort revealed that void spaces were more common in females than males, prevalence increased with age, and void spaces were typically systemic (occurring at both scan sites rather than only one). A comparison of group-wise differences between participants with and without void space demonstrated that individuals with void spaces had significantly worse trabecular properties for both sexes and at both scan sites. Specifically, the median trabecular bone mineral density, bone volume fraction, and trabecular number were below the 25th percentile of the population, while trabecular separation and inhomogeneity were above the 75th percentile of the population in participants with void spaces. Cortical bone characteristics did not differ between participants with and without void spaces. When the void space region was excluded from morphological analysis so that only the remaining "functional bone" was considered, trabecular properties of participants with void spaces were greatly improved, especially for those who were the greatest outliers. Void space is an intuitive morphological parameter that captures localized deterioration in the trabecular bone network, and has the potential to provide valuable insight into the assessment of bone fragility.