Development of specimen-specific finite element models of human vertebrae for the analysis of vertebroplasty

Abstract

The aim of this study was to determine the accuracy of specimen-specific finite element models of untreated and cement-augmented vertebrae by direct comparison with experimental results. Eleven single cadaveric vertebrae were imaged using micro computed tomography (microCT) and tested to failure in axial compression in the laboratory. Four of the specimens were first augmented with PMMA cement to simulate a prophylactic vertebroplasty. Specimen-specific finite element models were then generated using semi-automated methods. An initial set of three untreated models was used to determine the optimum conversion factors from the image data to the bone material properties. Using these factors, the predicted stiffness and strength were determined for the remaining specimens (four untreated, four augmented). The model predictions were compared with the corresponding experimental data. Good agreement was found with the non-augmented specimens in terms of stiffness (root-mean-square (r.m.s.) error 12.9 per cent) and strength (r.m.s. error 14.4 per cent). With the augmented specimens, the models consistently overestimated both stiffness and strength (r.m.s. errors 65 and 68 per cent). The results indicate that this method has the potential to provide accurate predictions of vertebral behaviour prior to augmentation. However, modelling the augmented bone with bulk material properties is inadequate, and more detailed modelling of the cement region is required to capture the bone-cement interactions if the models are to be used to predict the behaviour following vertebroplasty.

The use of preserved tissue in finite element modelling of fresh ovine vertebral behaviour

“…These measurements are sensitive to both the image resolution and the threshold level used to capture bone (Peyrin et al 1998;Hara et al 2002). Although the resolution used to image vertebrae in this study results in a low number of pixels per trabecular strut, it is comparable to that used in previous studies which involve the generation of specimen-specific FE models at the continuum level (Jones and Wilcox 2007;Wilcox 2007;Wijayathunga et al 2008). It has been shown to be sufficient to calculate the BV/TV on an element-by-element basis and provide models that are in good agreement with corresponding experimental results (Wilcox 2007).…”

“…The process was based on a method developed previously (Wilcox 2007) in which specimen-specific models of porcine vertebra were compared with corresponding experimental specimens and shown to be in good agreement in terms of the predicted stiffness and strength. For the VOI of each specimen, the stack of segmented images was down-sampled to a voxel size suitable for image processing on a PC and to generate a suitable FE mesh density of 1 mm £ 1 mm £ 1 mm that is superior to that suggested in previous mesh convergence tests, and is of a size commonly used in vertebral FE models (Jones and Wilcox 2007;Wijayathunga et al 2008). The VOI geometry was captured using thresholding and flood fill methods, surfaces were defined on the superior and inferior regions of the VOI, and the images were converted to an FE model using proprietary software (ScanIP, Version 4, Simpleware Limited, Exeter, UK).…”

“…At the continuum level, material properties are assigned to a model on an element-by-element basis using one of two methods. In the first, referred to here as the greyscale method, bone density is calculated from the intensity values or greyscale and directly used to determine mechanical properties such as elastic modulus or yield stress (Jones and Wilcox 2007;Wijayathunga et al 2008,). Whilst this is the most widely adopted method, it is necessary for all specimens to be imaged with the trabecular space in the same state, for example with the marrow intact.…”

Establishment and Validation of a Computed Tomography–Based Finite Element Model of Lumbar3-5 Vertebrae

“…To achieve this simplicity, the cortical surfaces of the vertebrae in these models are generally rough, with flat endplates. Some voxel-based models are smoothed on the surface to improve the geometric fit [21]. The initial mesh is generated from the image voxels, creating hexahedral elements.…”

Mechanical variables affecting balloon kyphoplasty outcome – a finite element study

“…Computational models have been used to better understand many factors affecting the biomechanical behaviour following augmentation (Baroud et al 2003;Polikeit et al 2003;Keller et al 2005;Lee and Teo 2005;Villarraga et al 2005;Rohlmann et al 2006;Chevalier et al 2008;Wijayathunga et al 2008). Factors such as cement volume, stiffness and level of height restoration, have been previously explored (Baroud et al 2003;Polikeit et al 2003;Keller et al 2005;Lee and Teo 2005;Villarraga et al 2005;Rohlmann et al 2006;Chevalier et al 2008;Wijayathunga et al 2008); however, not necessarily in a single model or study.…”