Whole human vertebral body creep is associated with DTS-derived texture parameters.

Document Type

Conference Proceeding

Publication Date


Publication Title

J Orthop Res


INTRODUCTION: Despite the potential importance of creep in development of vertebral fractures, little information is available in regards to clinically measurable predictors of creep properties in human vertebral bone. Digital tomosynthesis (DTS) combines relatively high in-plane resolution (150-300 μm) with about 1/5th the radiation dose as compared to computed tomography [1]. In this study, we examined DTS for its ability to provide image texture based predictors of creep properties in whole cadaveric vertebral bodies. METHODS: Under local IRB approval, thoracic 12 vertebrae were dissected and cleaned of soft tissue from 23 donors (13M/10F, 41-97y). Specimens were scanned using digital tomosynthesis (DTS), dual x-ray absorptiometry (DXA), and high resolution computed tomography (CT). Bone mineral density (BMD) was calculated from DXA in AP (BMD.AP) and LM (BMD.LM) directions and from CT in cancellous (cBMD), shell (shBMD) and integral (iBMD) volumes. DTS scans were performed in AP (producing a stack of coronal plane image) and LM (producing a stack of sagittal plane images) views while aligned axially (0°), transversely (90°) or obliquely (23°) to the super-inferior axis of the vertebrae. A cuboidal volume of interest from each image stack was analyzed using fractal (fractal dimension [FD], lacunarity [λ], slope lacunarity [Sλ]), mean intercept length (MIL) and line fraction deviation (LFD) methods as previously described [2]. For MIL and LFD parameters, Average (.Av), standard deviation (.SD), degree of anisotropy (.DA, ratio of maximum to minimum MIL or LFD), and maximum (.Max) were recorded within each slice, from which stack average and standard deviation were calculated (Fig 1). Within each scanning configuration, parameters causing high multicollinearity (as determined by a variance inflation factor>5) were eliminated. Specimens were then loaded to 1000N and held for 2 hours, load was removed, and recovery was recorded for another 2 hours. Creep deformation (Dcr, ptB-ptA Fig.2), creep recovery (Rcr, ptC-ptD Fig.2), residual displacement (Dres, ptD Fig.2), and residual from creep (Dres-cr, Dcr-Rcr) were calculated. In order to account for the dependence of creep displacement parameters on vertebral stiffness, displacements were normalized using the initial elastic deformation (ptA Fig.2) (i.e., Dcr-norm, Rcr-norm, Dres-norm, Dres-cr-norm). A function of the form “Dcr = a(1-EXP(-(t/τcr)n))+Ct” was fit to the creep portion of the data (pts A to B, Fig.2) to calculate time constant (τcr), stretch exponent (n) and creep rate (C) variables. Multiple regression models were constructed using a stepwise procedure to examine relationships between measures of creep and DTS texture parameters (JMP 10, SAS Institute). If a BMD variable was found to be significantly correlated to a creep variable, the BMD variable was introduced first and forced to stay in the model. Significance in multiple regression models was considered as p<0.05. RESULTS: Dcr was significantly correlated with BMD.LM, and Rcr with BMD.AP; these BMDs were forced as first terms in their associated models. The most explanatory set of variables for Dcr (R2adj=0.33) were BMD.LM (-, p<0.04) and Av(LFD.Max) (+, p<0.04) measured from oblique LM scanning configuration, while Rcr (R2adj=0.17) was only associated with BMD.AP (-, p<0.05). When normalized by stiffness to eliminate the contribution of elastic displacement, models for both Dcr-norm (R2adj=0.68; See Fig. 3) and Rcr-norm (R2adj=0.51) contained Av(λ) (p<0.0001,0.0003 respectively) and Av(MIL.DA) (p<0.0006,0.02 respectively), both from Axial AP scans. Both Dres (R2adj=0.47) and Dres-cr (R2adj=0.39) showed negative correlations with Av(FD) (p<0.0009,0.003 respectively) and SD(FD) (p<0.04,0.02 respectively) (Dres: Axial AP, Dres-cr: Transverse LM). The most explanatory models for Dres-norm (R2adj=0.49) and Dres-cr-norm (R2adj=0.65) demonstrated positive correlations with SD(MIL.Max) (p<0.0005,0.00003 respectively) and negative correlations with SD(LFD.Max) (p<0.02,0.02 respectively) in Trans erse LM scans. R2adj increased when displacements were normalized for stiffness (from 0.17-0.47 to 0.49-0.65). Creep rate was significantly correlated with all measured BMDs, with iBMD being the most significant, and the most explanatory model for creep rate (R2adj=0.61) included both iBMD (-, p<0.0001) and Av(LFD.DA) (+, p<0.003) (Axial AP scans). When alternative models from axial AP scans only were assessed for Dres-norm and Dres-cr-norm, it was noted that Dres-cr-norm could also be modeled with Av(λ) and Av(MIL.DA) (R2adj=0.57, p<0.0002,0.02) and Dres-norm with Av(λ) alone (R2adj=0.26, p<0.02). The most explanatory axial-AP model for C (R2adj=0.48) included a MIL.DA (+, p<0.05) term together with iBMD (-,p<0.002). DISCUSSION: We demonstrated that DTS-derived variables of cancellous bone texture are associated with creep properties of human vertebrae, independently from BMD variables. More strikingly, normalizing for stiffness eliminates BMD variables from models of Dcr and Rcr and only DTS variables remain, suggesting that DTS texture is related to vertebral creep independently from the elastic properties of the vertebra whereas BMD is not. Standard clinical DTS scanning configuration involves AP views and an axial scan direction. When models from the AP axial scans alone were considered, a unified set of variables (Av(MIL.DA) and Av(λ)) associated with all measured creep displacements could be identified. Lacunarity is a fractal measure related to the variability of the size of gaps in a porous texture [4], while MIL.DA represents anisotropy of trabecular structure. Both lacunarity [5] and anisotropy [6] are understood to significantly change with age and osteoporosis. The relationships of these measures of image texture with creep may help provide, independent from bone mass, important clinical indicators of the progression toward eventual vertebral deformity. (Figure Presented).



This document is currently not available here.