Elsevier

Journal of Biomechanics

Volume 45, Issue 2, 10 January 2012, Pages 394-399
Journal of Biomechanics

Short communication
Accuracy of finite element predictions in sideways load configurations for the proximal human femur

https://doi.org/10.1016/j.jbiomech.2011.10.019Get rights and content

Abstract

Subject-specific finite element models have been used to predict stress-state and fracture risk in individual patients. While many studies analysed quasi-axial loading configurations, only few works simulated sideways load configurations, such as those arising in a fall. The majority among these latter directly predicted bone strength, without assessing elastic strain prediction accuracy. The aim of the present work was to evaluate if a subject-specific finite element modelling technique from CT data that accurately predicted strains in quasi-axial loading configurations is suitable to accurately predict strains also when applying low magnitude loads in sideways configurations. To this aim, a combined numerical–experimental study was performed to compare finite element predicted strains with strain-gauge measurements from three cadaver proximal femurs instrumented with sixteen strain rosettes and tested non-destructively under twelve loading configurations, spanning a wide cone (0–30° for both adduction and internal rotation angles) of sideways fall scenarios. The results of the present study evidenced a satisfactory agreement between experimentally measured and predicted strains (R2 greater than 0.9, RMSE% lower than 10%) and displacements. The achieved strain prediction accuracy is comparable to those obtained in state of the art studies in quasi-axial loading configurations. Still, the presence of the highest strain prediction errors (around 30%) in the lateral neck aspect would deserve attention in future studies targeting bone failure.

Introduction

Osteoporosis and sideways fall are known to be two of the major determinants of proximal femur fractures among the elderly (Berry and Miller, 2008, Courtney et al., 1995, Greenspan et al., 1998, Parkkari et al., 1999). Current methods for fracture risk evaluation, based on densitometry and epidemiological parameters (Kanis et al., 2005), are not free from limitations (Lekamwasam, 2010, Silverman and Calderon, 2010, Watts et al., 2009). A major concern is that areal bone mineral density is the only bone strength determinant included (Ensrud et al., 2009, Gagnon and Ebeling, 2009). Subject-specific finite element (FE) models of bones from Computed Tomography (CT) data have been proposed to improve fracture risk prediction, since they can take into account the structural determinants of bone strength and the variety of external loads acting on bones (Cody et al., 1999). A preliminary requirement for clinical application is the in-vitro validation of FE models predictions (Viceconti et al., 2005).

The majority of FE validation studies in literature focused on bone strength prediction in quasi-axial loading configurations (e.g. resembling single stance configuration), while few works aimed at validating FE models in sideways fall configurations. Among those who addressed sideways fall (Keyak et al., 2001, Koivumäki et al., 2010, Majumder et al., 2009, Verhulp et al., 2008, Wakao et al., 2009), most of them try to directly predict bone strength, without preliminarily assessing the accuracy in elastic strain prediction, even if using a strain-based elastic limit criterion.

To the authors' knowledge, only one work (Lotz et al., 1991) reported FE strain prediction accuracy in sideways fall configurations. However, this work was limited to one femur and one loading configuration, while falling to the side may give rise to a variety of boundary conditions (Groen et al., 2008, Nankaku et al., 2005, Wakao et al., 2009), and achieved a limited strain prediction accuracy (R2=0.67, calculated from the raw data reported in the paper).

Recently, the authors developed a methodology to generate subject-specific FE models of femurs from CT data (Schileo et al., 2007, Schileo et al., 2008, Taddei et al., 2006). This methodology achieved under quasi-axial loading configurations a high in vitro strain prediction accuracy, comparable to the authors' knowledge only to (Bessho et al., 2007, Trabelsi et al., 2009).

The aim of the present work was to verify if the FE modelling procedure proposed in (Schileo et al., 2008), hereinafter called “reference study”, could accurately predict the strain levels elicited by low magnitude loads applied in vitro in sideways loading conditions.

Section snippets

Materials and methods

Three unpaired cadaver femora showing no deformities were obtained from IIAM (www.iiam.org) and embalmed in a 4% formalin solution (Öhman et al., 2008). They were scanned with CT (HiSpeed GE Co., USA, pixel size 0.59 mm, slice thickness 1 mm from femoral head to lesser trochanter, 5 mm elsewhere) and dual energy X-ray absorptiometry (DXA) (Eclipse, Norland Co., USA) (Table 1).

Experimental measurements

Strain increased linearly with load for each individual strain gauge, and each loading configuration: R2≥0.99 for 98% of the cases where strains reached a value of 100 microstrain or larger. Similarly, displacements measured by LVDTs increased linearly with load (R2≥0.85 for 94% of the cases where displacements reached a value of 50 μm or larger). This confirms that bone can be assumed to behave linearly with good approximation for the strain range and strain rates used in this study.

Comparison between predicted and measured strains

The FE

Discussion

The aim of the present study was to verify if a previously proposed FE modelling procedure (Schileo et al., 2008, Schileo et al., 2007) is capable to accurately predict in-vitro measured strains also in a range of sideways load configurations.

The main outcome of the present study is the achievement of a good strain prediction accuracy for the tested set of sideways load configurations, both in terms of correlation (R2>0.9) and error (RMSE%<10%). These results are comparable to those obtained in

Conflict of interest statement

None of the authors received nor will receive direct or indirect benefits from third parties for the performance of this study.

Acknowledgments

The present study was partially funded by EC Grant VPHOP (FP7-ICT2008-223865) and Emilia Romagna Region-University Research Programme 2007–2009.

References (37)

Cited by (95)

View all citing articles on Scopus
View full text