Prediction of Femoral Neck Strength in Patients with Diabetes Mellitus with Trabecular Bone Analysis and Tomosynthesis Images.

Purpose To determine trabecular bone analysis values by using tomosynthesis images in determining femoral neck strength in patients with diabetes mellitus and compare its parameters between vertebral compression fracture and nonfracture groups. Materials and Methods The institutional review board approved this study, and written informed consent was obtained from all patients. Forty-nine patients with diabetes mellitus were included. Within 1 week, patients underwent dual x-ray absorptiometry (DXA), tomosynthesis, and computed tomography (CT) covering the T10 vertebral body to the hip joints. The trabecular patterns of tomosynthesis images were extracted, and the total strut length, bone volume per tissue volume, and five textural features (homogeneity, entropy, correlation, contrast, and variance) were obtained as the indices of tomosynthesis images. Failure load of the femoral neck, which was determined with the CT-based finite-element method (FEM), was used as the reference standard for bone strength. A forward stepwise multiple regression analysis for evaluating the availability of the tomosynthesis image indices was performed. The bone mineral density (BMD) at DXA and tomosynthesis image indices were compared between the vertebral compression fracture (n = 16) and nonfracture groups (n = 33) according to Genant semiquantitative morphometry methods by using one-way analysis of variance. Results The combination of BMD with the bone volume per tissue volume at the principal tensile group and the correlation at the principal compressive group showed the highest correlation to the failure load at CT FEM, and the correlation (r = 0.83) was higher than that between the failure load and the BMD alone (r = 0.76; P < .001). The averages of the bone volume per tissue volume and entropy at the principal tensile group in the vertebral compression fracture group were lower than those in the nonfracture group (P = .017 and P = .029, respectively), but there was no difference in BMD. Conclusion Tomosynthesis-based trabecular bone analysis is technically feasible and, in combination with BMD measurements, can potentially be used to determine bone strength in patients with diabetes mellitus. RSNA, 2016 Online supplemental material is available for this article.