Textural and geometric measures derived from digital tomosynthesis discriminate patients with vertebral fracture from those without

Document Type

Conference Proceeding

Publication Date

9-27-2024

Publication Title

J Bone Miner Res

Abstract

Digital tomosynthesis (DTS) is a limited-angle tomographic imaging modality providing a stack image of an object at high resolution and low radiation exposure. The purpose of this study was to examine the extent to which DTS derived textural and geometric properties of vertebrae discriminate patients with and without vertebral fracture. Under IRB approval, 93 postmenopausal women (age ≥ 50 years) with no history of bone disease other than osteoporosis were enrolled. The patients with vertebral fracture (Fx, N = 39) and those without (NFx, N = 54) were not different in age (65 ± 8 vs 64 ± 7 years; p > 0.2), BMI (25.1 ± 3.3 vs 25.1 ± 3.7 kg/m2; p > 0.9) or race distribution (9\30 vs 8\46 Black\Nonblack; p > 0.3). Lumbar spine bone mineral density (BMD) and trabecular bone score (TBS) were measured, and vertebral fracture assessment was performed from DXA scans. DTS of the spine was performed using a clinical system (Sonialvision Safire II, Shimadzu Inc) with the participant in supine position and central X-ray tube fixed at the T12-L1 level. DTS images were reconstructed with a voxel spacing of 0.28 x 0.28 x 1 mm. Fractal dimension (FD, a measure of texture complexity) and lacunarity (λ, a measure of texture heterogeneity) were calculated for cancellous bone using FracLac and ImageJ software. Mean intercept length (MIL, a measure of feature size) and line fraction deviation (LFD, a measure of orientation) were measured and degree of anisotropy (DA) was calculated (maximum MIL/minimum MIL). In addition, vertebral width was calculated at the narrowest section of the mid-vertebra using coronal images. DTS values for fractured T12 and L1 vertebrae were imputed from unfractured levels using a mixed model regression of each DTS variable by vertebral level from a superset of 131 patients with no fracture. DTS measurements of the T12 and L1 vertebrae were averaged for each subject. Differences between groups were assessed using t-tests or Wilcoxon tests based on data normality. Logistic regression models were constructed to examine the extent to which DTS predicts vertebral fracture status. BMD and TBS were higher, while DA and width were lower, in NFx than Fx (p < 0.02 to p < 0.003). Multiple logistic regression identified BMD, FD, λ, DA and width as significant predictors (p < 0.02 to p < 0.001) with AUC of ROC = 0.79 (compared to 0.67 for BMD alone) (Figure). These results support complementary use of DTS in assessment of bone quality and potentially of fracture risk.

Volume

39

First Page

152

Find It @ Sladen

Share

COinS