Validated FE models of vertebral bodies predict displacements and strains from axial impact loading

Finite Element (FE) method has been extensively applied to understand the behaviour of bone, including vertebral bodies (VBs), subject to loading with most studies so far having focused on quasi-static conditions. Dynamic studies are, however, important as they can provide insights into how injuries arise from high energy collisions, such as those that can be experienced in sport and road traffic accidents. The aim of this study was to develop and validate a methodology for FE modelling of VBs subject to impact loading. Seven (n=7) VBs from four different porcine spines were stripped of all soft tissues, cemented into polyoxymethylene pots, µCT-scanned and positioned in an impact cage. An impact was applied to the VBs via a falling mass of 7.4kg at a velocity of 3.1m/s. Surface displacements and strains were acquired from the anterior VB surface via DIC [Photron,UK] and the impact load was monitored with two loadcells, one cranial and one caudal. Specimen-specific FE models were created based on µCT images. Material properties were assigned based on Hounsfield units and the density-Young`s modulus relationship, validated by previous static experiments, was calibrated for the dynamic case using a factor (KImpact). All DOFs of the caudal pot were locked while the cranial pot was free to move vertically, with the experimental cranial load profile being applied to it, matching experimental conditions. Experimental and numerical load-displacement (LxD) curves were compared using Bland-Altman plots, RMSE and Lin’s concordance coefficient (CCC). With KImpact=0.0825, five models presented good agreement with experimental data, with average RMSE and CCC for LxD being, respectively, 0.036mm and 0.889. For peak load, maximum displacements and strains were, on average, 0.4mm and 0.035, respectively and most Bland-Altman plots showed dispersion around zero. The remaining two models showed poor agreement, with average RMSE and CCC for LxD being, respectively, 0.06mm and 0.32. However, overall strains levels, peak strains and peak strain locations were similar between experimental and numerical models. The majority of FE models had good agreement with experiments and, after calibration, they predicted the dynamic behaviour of VBs.