Investigation of localized bone density and microarchitecture changes in osteoporotic vertebral fracture from routine-CT scans

Abstract

During osteoporosis, bone mineral density and microarchitectures in different subregions of the vertebral body changed unevenly. Due to clinical imaging technology limitations, relatively few studies have explored the quantitative measurement of the uneven bone loss in routine computed tomography (CT) images. Routine CT image-based bone imaging biomarkers associated with osteoporotic fractures still lack investigation. This study aimed to develop a routine CT-based localized bone loss assessment method, and to identify the specific changes associated with osteoporotic vertebral fractures. Specifically, the objectives were 1) to develop and validate a surrogate method for localized bone quality loss evaluation from routine CT scans; 2) to investigate the feasibility and clinical association of the localized bone quality loss measurements with osteoporotic vertebral fracture prevalence; and 3) to investigate the association of localized bone density distribution in vertebral body subregions with osteoporotic fractures. In this study, the novel measure of focal bone mineral content loss was developed based on intensity profile analysis. To validate this approach, the measurement reproducibility, and the correlation with bone microarchitecture parameters in high-resolution CT were evaluated. Also, the association of focal bone mineral content loss with osteoporotic fracture was further investigated based on case-control study. Furthermore, a hybrid voxel-based morphometry approach for localized bone density distribution analysis was developed using the multi-scale thresholding, morphological ‘hit-or-miss’ transform and deep learning framework-based deformable registration. Finally, the association of localized bone density distribution in each vertebral subregion with the fracture prevalence was analyzed using the voxel-based morphometry. The results revealed that as a surrogate measure of localized bone microarchitectural quality in routine CT scans, the measurement of focal bone mineral content loss showed high scan-rescan reproducibility (root mean square standard deviation 0.011, intraclass correlation coefficient 0.97), and good correlation with the apparent microarchitecture parameters measured in high-resolution CT (bone volume fraction r = 0.79, P = 0.001; trabecular bone spacing r = 0.76, P < 0.001; porosity r = 0.79, P < 0.001). The focal bone mineral content loss also demonstrated a high association (OR=3.103, 95% CI 1.504~6.403, P<0.01) with the clinical vertebral body fracture prevalence, in an average bone mineral density independent manner (r=0.42, P<0.001). Moreover, the hybrid voxel-based morphometry approach showed high segmentation accuracy and registration precision (AQUIRC-dice score 0.9799±0.0059, AQUIRC-MSE 0.1487±0.0545) for localized vertebral bone density analysis. The case-control study further verified that the bone mineral densities in vertebral fracture patients are focally decreased in specific subregions of the vertebral body (around 30% of the total region) rather than evenly lost. In conclusion, the study found that focal bone loss is highly associated with osteoporotic fracture prevalence. Using the intensity profile-based analysis method achieved high reliability and effectiveness for focal bone loss evaluation in routine CT scans. Furthermore, the bone density in the specific regions of the vertebral body is more vulnerable to vertebral compression fracture than other regions. All the evidence indicated that the evaluation of localized bone quality is critical to the assessment of osteoporotic vertebral bone fragility, and this study provided the novel methods to analyze it.