Digital tomosynthesis (DTS) derived textural parameters of human vertebral cancellous bone have been previously correlated to the finite element (FE) stiffness and 3D microstructure. The objective of this study was to optimize scanning configuration and use of multiple image slices in the analysis, so that FE stiffness prediction using DTS could be maximized. Forty vertebrae (T6, T8, T11, and L3) from ten cadavers (63-90 years) were scanned using microCT to obtain trabecular bone volume fraction (BV/TV) and FE stiffness. The vertebrae were then scanned using DTS anteroposteriorly (AP) and laterally (LM) while aligned axially (0°), transversely (90°) or obliquely (23°) to the superior-inferior axis of the vertebrae. From the serial DTS images, fractal dimension (FD), mean intercept length (MIL) and line fraction deviation (LFD) parameters were obtained from a 2D-single mid-stack location and 3D-multi-image stack. The DTS derived textural parameters were then correlated with FE stiffness using linear regression models within each scanning orientation. 3D-multi-image stack models obtained from Transverse-LM scanning orientation (90°) were most explanatory regardless of accounting for the effects of BV/TV. Therefore, DTS scanning perpendicular to the axis of the spine in an LM view is the preferred configuration for prediction of vertebral cancellous bone stiffness.
Keywords: Cadaver study; Fractal dimension; Lacunarity; Large scale finite element modeling; Line fraction deviation; Mean intercept length; Stiffness; Vertebral cancellous bone.