Methodology to Produce Specimen-Specific Models of Vertebrae: Application to Different Species

Abstract

Image-based continuum-level finite element models have been used for bones to evaluate fracture risk and the biomechanical effects of diseases and therapies, capturing both the geometry and tissue mechanical properties. Although models of vertebrae of various species have been developed, an inter-species comparison has not yet been investigated. The purpose of this study was to derive species-specific modelling methods and compare the accuracy of image-based finite element models of vertebrae across species. Vertebral specimens were harvested from porcine (N = 12), ovine (N = 13) and bovine (N = 14) spines. The specimens were experimentally loaded to failure and apparent stiffness values were derived. Image-based finite element models were generated reproducing the experimental protocol. A linear relationship between the element grayscale and elastic modulus was calibrated for each species matching in vitro and in silico stiffness values, and validated on independent sets of models. The accuracy of these relationships were compared across species. Experimental stiffness values were significantly different across species and specimen-specific models required species-specific linear relationship between image grayscale and elastic modulus. A good agreement between in vitro and in silico values was achieved for all species, reinforcing the generality of the developed methodology.

Optimizing computational methods of modeling vertebroplasty in experimentally augmented human lumbar vertebrae

“…Having greater definition of the cortical shell and trabecular bone alignment also allows a better regional representation of the load transfer through the vertebrae than the more homogenized predictions seen with the direct grayscale method. This improved agreement is much stronger than the agreement found in similar studies that used a comparable methodology to the direct grayscale method and comparable to methods that used more complex, specimen‐specific material properties for each model . The conversion between grayscale and Young's modulus was comparable to studies with similar methodologies; Robson Brown et al found an equivalent conversion factor of 0.0013/GPa (0.0009/GPa in the current study), which gives a similar average bone modulus of 0.33 GPa to that of the current study, 0.25 GPa.…”

“…Two approaches to generating the models were compared (Figure ). In the first (“direct grayscale method”), the images were downsampled to 1 mm 3 and the resulting grayscale of each voxel was used to define the local bone elastic modulus (Figure 3A,B), similar to a method used by Zapata‐Cornelio et al In the second (“bone volume fraction” method), an initial threshold was applied to the 82‐μm image stack to segment the trabeculae from the trabecular spaces (Figure 3D). The selected threshold was determined as the mean across all specimens using the optimize threshold feature of the BoneJ plugin for ImageJ (version 1.4.1), which optimized the connectivity against the threshold value.…”

“…Bone materials were modeled as an isotropic linear elastic material, where the Young's modulus of each element was correlated with the grayscale value of the corresponding voxel using a linear conversion factor. The relationship between the grayscale value and Young's modulus was optimized to provide the best fit between the FE‐predicted stiffness values and the corresponding experimentally derived stiffness values using an optimization toolbox with the method employed by Zapata‐Cornelio et al Initially, the 14 vertebra models were split evenly into a build set and a validation set, with the optimization of the conversion factor carried out on the build set and this factor was then applied to the models in the validation set. Following this validation step, all 14 vertebrae were then used to optimize the conversion factor.…”

Experimental and Computational Comparison of Intervertebral Disc Bulge for Specimen-Specific Model Evaluation Based on Imaging

“…Computed tomography images were imported into Simpleware ScanIP (v 7.0, Synopsys, United States) after their greyscalevalues had been rescaled to enable the use of previously calibrated bone properties (Zapata-Cornelio et al, 2017). The rescaled CT images at each compression step were rigidly registered to the CT images of specimens in their initial preloaded state ("preloaded CT images"), using the most caudal vertebrae as reference.…”

“…Bone was modelled as a linear elastic material, with element-byelement elastic modulus scaled using the underlying greyscale value of the CT image data, and Poisson's ratio of 0.3. In order to use a greyscale-based equation for the bone modulus (Zapata-Cornelio et al, 2017), the images underlying the highresolution segmentation were down-sampled to an isotropic resolution of 0.5 mm. The greyscale-based model requires first a normalisation of the greyscale to 0-255.…”