Escolar Documentos
Profissional Documentos
Cultura Documentos
Cardiff1
School of Mechanical and
Materials Engineering,
University College Dublin,
Belfield, D4,
Development of a Hip
Dublin, Ireland
e-mail: philip.cardiff@ucd.ie Joint Model for Finite
A. Karac
Faculty of Mechanical Engineering,
Volume Simulations
University of Zenica,
Fakultetska 1, 72000 Zenica, This paper establishes a procedure for numerical analysis of a hip joint using the finite
Bosnia and Herzegovina volume method. Patient-specific hip joint geometry is segmented directly from computed
tomography and magnetic resonance imaging datasets and the resulting bone surfaces
are processed into a form suitable for volume meshing. A high resolution continuum
D. FitzPatrick tetrahedral mesh has been generated, where a sandwich model approach is adopted; the
bones are represented as a stiffer cortical shells surrounding more flexible cancellous
R. Flavin cores. Cartilage is included as a uniform thickness extruded layer and the effect of layer
thickness is investigated. To realistically position the bones, gait analysis has been per-
A. Ivankovic formed giving the 3D positions of the bones for the full gait cycle. Three phases of the
gait cycle are examined using a finite volume based custom structural contact solver
School of Mechanical and implemented in open-source software OpenFOAM. [DOI: 10.1115/1.4025776]
Materials Engineering,
University College Dublin, Keywords: hip joint, gait analysis, bone segmentation, volume meshing
Belfield, D4,
Dublin, Ireland
3
In this context, castellated refers to the resemblance of the mesh to the where Ra 27:6 mm and Rf 26:4 mm have been determined
battlements on top of a medieval castle. by manually fitting spheres.
homogenous, and isotropic and the material distribution is dis- displacements field solutions [36], must also reach the prescribed
played previously in Fig. 3. The cortical bone is assigned a tolerance.
Youngs modulus of 17 GPa and a Poissons ratio of 0.3 The models have been solved in parallel on a distributed mem-
[2,47,14,6164], the cancellous bone is assigned a Youngs ory supercomputer using 32 CPU cores (Intel Xeon E5430 Quad
modulus of 800 MPa and a Poissons ratio of 0.2 [64,65], and the Core 2.66 GHz) in an approximate clock time of 10 h.
articular cartilage is assigned a Youngs modulus of 12 MPa and a Figure 7(a) displays the solution convergence for a typical
Poissons ratio of 0.45 [19,66]. model, showing that both the solver residual and the relative
As boundary conditions, shown graphically in Fig. 6, the pelvis residual reach the predefined tolerance of 107 in approximately
is fixed at the iliosacral and pubic symphysis joints, and the distal 160,000 outer iterations. Although a relatively aggressive under-
femur is displaced in the femur axial direction into the acetabulum relaxation factor of 0.05 has been applied to achieve convergence,
such that the resulting total hip joint force is as measured in vivo high frequency oscillations are still visible in the residuals, most
by Bergmann et al. [45] At midstance and toe-off, the hip joint likely due to the nonlinear nature of the contact as well as lower
force is twice the body weight, 1,611 N, in the femur axial direc- quality cells. Figure 7(b) shows the convergence of the contact
tion. For the heel-strike model, the femur axial force is 1,917 N, penetration to the predefined gap tolerance 9 106 m, where it
equivalent to 2.38 times the body weight. can be seen that the contact converges at approximately 60,000
The remaining femur and pelvis surfaces are specified as outer iterations and the remaining outer iterations are required to
traction-free. Custom boundary conditions with nonorthogonal converge the displacement increments.
corrections, previously described [36,60], are employed and the The collection of software employed in the current work is
gradient terms are calculated using the least squares method.4 The summarized in the Appendix.
models are solved in one load increment and inertia and gravity
forces are neglected.
For the contact procedure, the pelvis articular cartilage surface
is designated as the master and the femur articular surface as the
slave. The contact penetration distances have been calculated
using the contact-spheres approach [35]. A contact gap tolerance
of 9 106 m is employed, the penalty factor is 6 108 and the
contact correction frequency is 40.
3.3 Contact Stress Analysis. The von Mises stress distribu- articular surface faces with a pressure greater than 1 kPa. The
tion of the mid-stance model is shown in Fig. 8. The most highly average contact pressure is 6.28 MPa and has been calculated by
stressed areas, of 30MPa, are found in the ilium directly above dividing the total contact normal force by the contact area. The
the acetabulum, the acetabular roof bone, as well as near the fixed total contact normal force, Cn , is calculated by
iliosacral joint. The scale of Fig. 8 ranges from 0 to 10 MPa to
allow clearer viewing of the highly stressed regions. X Cf
Cn Cf rf (2)
Examining the contact pressure distribution, shown in Fig. 8(a), f
jCf j
three distinct contact regions are discernible, occurring in anterior
superior, posterior superior, and superior regions of the acetabu- P
lum. The maximum predicted contact pressure is 26 MPa occur- where f refers to the summation over all the faces of the articu-
ring in the most superior contact region. The predicted contact lar surface and Cf is the face area vector.
area is 3:96 104 m2 and has been calculated by summing the Inspecting the model contact gap, as shown in Fig. 8(b), the
anterior superior and posterior superior regions show the most
negative contact gap, relative to the superior region of the femoral
head.
When the von Mises stress distribution of the toe-off and heel-
strike models are examined, the most highly stressed areas occur
in the bone superior and posterior to the acetabulum. As with the
mid-stance model, the acetabular roof is highly stressed, in partic-
ular directly above the contact regions. Additionally, the bone
around the iliosacral joint experiences high stresses.
Inspecting the contact regions of the toe-off and heel-strike
models, shown side-by-side in Figs. 9, distinct contact areas are
once again visible. As with the mid-stance model, three contact
regions occur in the heel-strike model, where the maximum
predicted contact pressure is 26 MPa, the contact area is
3:83 104 m2 and the average pressure is 10.1 MPa. In contrast,
two contact regions occur in the toe-off model, where the maxi-
mum predicted contact pressure is 23 MPa, the contact area is
4:62 104 m2 , and the average pressure is 5.93 MPa.
To ensure the predicted results are mesh independent, an addi-
tional mid-stance simulation has been performed using a more
dense globally refined mesh. The refined mesh, containing a total
of 769,529 cells (25% more cells than the original mesh) with an
average articular surface cell breadth of approximately 0.42 mm,
predicted a maximum contact pressure of 26.67 MPa, which
is within 3% of the original mesh predictions, and stress
distributions close to the original mesh, demonstrating mesh
independence.
Table 1 Effect of Cartilage Thickness on the Contact Mechanics of the Mid-Stance Model
Cartilage thickness (in mm) Maximum pressure (in MPa) Average pressure (in MPa) Area (in m2)
4 Discussion been found that there is an inverse relationship between the carti-
lage thickness and maximum contact pressure, where the maxi-
Inspecting the regions of greatest stress in the mid-stance mum contact pressure increases by approximately 60% as the
model, the pelvis is relatively highly stressed in the acetabular cartilage thickness decreases by 33% from 0.72 to 0.48 mm. How-
roof, the ilium above the acetabulum and at the iliosacral joint. On ever, surprisingly at first, the middle thickness (0.6 mm) model
closer analysis of the femur, the body of the femur holds much of predicts the lowest average contact pressure. The reason for this
the stress where significant bending occurs. These predictions may be deciphered by noticing that the contact pressure increases
agree well with results from numerical models in literature in the thicker cartilage model in the anterior and posterior regions
[2,4,14,16,19,62,64,68], and it has been noted the large predicted of the acetabulum and lowers in the superior region. In reality, it
stresses in the vicinity of the iliosacral joint are contributed to by is expected that the physiological cartilage thickness is greater on
the rigid fixture boundary condition [69]. the superior femur head surface; therefore increasing the overall
Examining the predicted contact pressure distributions, the cartilage thickness may overestimate the thickness in the anterior
analyses suggest that the hip joint contact is not perfectly congru- and posterior regions of the acetabulum and femur head.
ent and that there are distinct regions of local high contact pres- Of the three phases of gait examined, as anticipated the largest
sure, agreeing with previous FE predictions [4,14]. Comparing the stresses and contact pressures occur in the heel-strike model, cor-
contact mechanics of the mid-stance model with values from liter- responding to the peak in gait cycle total joint forces.
ature, the predicted contact pressures are larger than the reported For all models, the predicted stress values and locations are
values. Peak contact pressures in literature vary from 1 to 18 MPa consistent with previous FE studies [2,4,14,16,19,36,62,64,68],
[2,4,14,16,19,62,64,68], whereas the maximum predicted contact however, the magnitude of maximum contact pressures are
pressure in the mid-stance model is 26 MPa. A possible explana- greater than previously predicted. The effect of cartilage thick-
tion for this difference may be attributed to the simplifying ness has been examined and as expected the maximum pre-
assumptions commonly made in the literature with regard to dicted contact pressure has been found to be inversely related
spherical congruent contact surfaces that overestimate contact to the cartilage thickness, with the case of the thinnest cartilage
areas; hence underestimating contact pressures. Additionally, producing the greatest maximum contact pressure. However, the
many of the published numerical models represent the contact largest average contact pressure has been found in the thickest
surfaces with relatively low resolution grids essentially averaging cartilage case, most likely due to the assumption of constant
local contact stress peaks. A contributing factor that may result in thickness cartilage resulting in underestimation of superior carti-
the current model overestimating the contact pressure may be lage thickness and overestimation of anterior/posterior cartilage
ascribed to simplifying assumptions with regard to the uniform thickness.
bone mechanical properties and the constant thickness articular All presented results show that this novel numerical model,
cartilage. Employing a stiffness of 17 GPa for the acetabular corti- based on the finite volume method as implemented in an open-
cal bone may overestimate the stiffness of the pelvis, reducing the source software, represent a powerful alternative to well-
natural congruency of the joint and increasing contact stresses. A established commercial FE-based software. As its object-oriented
more realistic representation of varying cartilage thickness in nature permits complex custom routines to be implemented in an
addition to spatially varying tomography-based bone properties efficient manner, further improvement of the current model will
would certainly result in a more faithful model. be focused on implementation of material properties based on CT
Analyzing the effect of cartilage thickness has shown a consid- Hounsfield pixel intensities and on employment of physiologically
erable influence on the predicted mechanics. As expected, it has realistic musculotendon loading models [36,44].
CT/MRI segmentation 3DSlicer [47] Open-source image segmentation and surface generation software
Surface processing Meshlab [50] Open-source software for general surface processing
Surface processing Metro [51] Open-source software alloying verification of decimated surfaces
Volumetric meshing ANSYS ICEM CFD [55] Commercial meshing software for generation of volumetric
meshes
Volumetric meshing OpenFOAM/fluent3DMeshToFoam OpenFOAM utility to convert mesh to OpenFOAM format
[42,43]
Volumetric meshing OpenFOAM/extrudeMesh [42,43] OpenFOAM utility to create cartilage volumetric mesh
Gait analysis OpenFOAM/visualiseGaitData [44] Custom OpenFOAM utility to convert gait data into a form suitable
for viewing
Gait analysis OpenFOAM/rotateRigidFemur [36] Custom OpenFOAM utility for relative positioning of the femur
using gait data
Solving OpenFOAM/elasticNonLinUL- Custom OpenFOAM application for numerical calculation of
SolidFoam [60] displacements, stresses and strains by finite volume method
Post-processing ParaView [59] Open-source visualisation software for post processing results
Acknowledgment [15] Oonishi, H., Isha, H., and Hasegawa, T., 1983, Mechanical Analysis
of the Human Pelvis and Its Application to the Artificial Hip JointBy Means
Financial support from TEKNO Surgical Ltd. is gratefully of the Three Dimensional Finite Element Method, J. Biomech., 16, pp.
acknowledged. The authors thank Mike Walsh and Damien 427444.
[16] Brown, T. D., and DiGioia, A. M., 1984, A Contact-Coupled Finite Element
Kiernan of the Central Remedial Clinic Dublin for conducting the Analysis of the Natural Adult Hip, J. Biomech., 17, pp. 437448.
gait analysis. [17] Afoke, N. Y. P., Byers, P. D., and Hutton, W. C., 1987, Contact Pressures in
the Human Hip Joint, J. Bone Joint Surgery, 69-B(4), pp. 536541. Available
at http://www.bjj.boneandjoint.org.uk/content/69-B/4/536.full.pdf
[18] Dalstra, M., and Huiskes, R., 1995, Load Transfer Across the Pelvic Bone,
Appendix J. Biomech., 28(6), pp. 715724.
[19] Russell, M. E., Shivanna, K. H., Grosland, N. M., and Pedersen, D. R., 2006,
The software employed in the current work are summarized in Cartilage Contact Pressure Elevations in Dysplastic Hips: A Chronic Overload
Table 2. Model, J. Orthop. Surg. Res., 1, p. 6.
[20] Shivanna, K. H., Grosland, N. M., Russell, M. E., and Pedersen, D. R., 2008,
Diarthrodial Joint Contact Models: Finite Element Model Development of the
Human Hip, Eng. Comput., 24, pp. 155163.
References [21] Demirdzic, I., Martinovic, D., and Ivankovic, A., 1988, Numerical Simulation
[1] Brekelmans, W. A. M., Poort, H. W., and Slooff, T. J. J. H., 1972, A New of Thermal Deformation in Welded Workpiece (in Croatian), Zavarivanje 31,
Method to Analyse the Mechanical Behaviour of Skeletal Parts, Acta Ortho- pp. 209219.
paedica, 43(5), pp. 301317. [22] Ivankovic, A., Muzaferija, A., and Demirdzic, I., 1997, Finite Volume Method
[2] Dalstra, M., Huiskes, R., and van Erning L., 1995, Development and and Multigrid Acceleration in Modelling of Rapid Crack Propagation in Full-
Validation of a Three-Dimensional Finite Element Model of the Pelvic Bone, Scale Pipe Test, Comput. Mech., 20(12), pp. 4652.
ASME J. Biomech. Eng., 117(3), pp. 272278. [23] Georgiou, I., Ivankovic, A., Kinloch, A. J., and Tropsa, V., 2003, Rate
[3] Anderson, A. E., Peters, C. L., Tuttle, B. D., and Weiss, J. A., 2005, Subject- Dependent Fracture Behaviour of Adhesively Bonded Joints, Vol. 32, Fracture
Specific Finite Element Model of the Pelvis: Development, Validation and of Polymers, Composites and Adhesives II, Volume 32 of European Structural
Sensitivity Studies, ASME J. Biomech. Eng., 127(3), pp. 364373. Integrity Society, A. Pavan, B. R. K. Blackman, and J. G. Williams, eds.,
[4] Anderson, A. E., Ellis, B. J., Maas, S. A., and Weiss, J. A., 2010, Effects of Elsevier, New York, pp. 317328.
Idealized Joint Geometry on Finite Element Predictions of Cartilage Contact [24] Karac, A., Blackman, B. R. K., Cooper, V., Kinloch, A. J., Rodriguez Sanchez,
Stresses in the Hip, J. Biomech., 43, pp. 13511357. S., Teo, W. S., and Ivankovic, A., 2011, Modelling the Fracture Behaviour of
[5] Bachtar, F., Chen, X., and Hisada, T., 2006, Finite Element Contact Analysis Adhesively-Bonded Joints as a Function of Test Rate, Eng. Fract. Mech., 78,
of the Hip Joint, Med. Biol. Eng. Comput., 44, pp. 643651. pp. 973989.
[6] Majumder, S., Roychowdhury, A., and Pal, S., 2004, Variations of Stress in [25] Demirdzic, I., and Muzaferija, S., 1995, Numerical Method for Coupled Fluid
Pelvic Bone During Normal Walking, Considering All Active Muscles, Trends Flow, Heat Transfer and Stress Analysis Using Unstructured Moving Meshes
Biomater Artif Organs, 17(2), pp. 4853. Available at http://www.biomaterials. With Cells of Arbitrary Topology, Comput. Methods Appl. Mech. Eng.,
org.in/ojs/index.php/tibao/article/view/221 125(14), pp. 235255.
[7] Silvestri. C., 2008, Development and Validation of a Knee-Thigh-Hip [26] Jasak, H., and Weller, H. G., 2000, Application of the Finite Volume Method
LSDYNA Model of a 50th Percentile Male, Ph.D. thesis, Worcester Polytech- and Unstructured Meshes to Linear Elasticity, Int. J. Numer. Methods Eng.,
nic Institute. 48, pp. 267287.
[8] Anderson, A. E., Ellis, B. J., Maas, S. A., Peters, C. L., and Weiss, J. A., 2008, [27] Fryer, Y. D., Bailey, C., Cross, M., and Lai, C. H., 1991, A Control Volume
Validation of Finite Element Predictions of Cartilage Contact Pressure in the Procedure for Solving Elastic Stress-Strain Equations on an Unstructured
Human Hip Joint, ASME J. Biomech. Eng., 130(5), p. 051008. Mesh, Appl. Math. Model., 15(1112), pp. 639645.
[9] Anderson, A. E., Ellis, B. J., Peters, C. L., and Weiss, J. A., 2008, Cartilage [28] Wheel, M. A., 1996, A Geometrically Versatile Finite Volume Formulation
Thickness: Factors Influencing Multidetector CT Measurements in a Phantom for Plane Elastostatic Stress Analysis, J. Strain Anal. Eng. Des., 31(2), pp.
Study, Radiology, 246(1), pp. 133141. 111116.
[10] Ota, T., Yamamoto, I., and Morita, R., 1999, Fracture Simulation of the Femo- Ivankovic, A., and Karac, A., 2012, Finite Volume Stress Analy-
[29] Tukovic, Z.,
ral Bone Using the Finite-Element Method: How a Fracture Initiates and sis In Multi-Material Linear Elastic Body, Int. J. Numer. Methods Eng., 93(4),
Proceeds, J. Bone and Mineral Metabolisms, 17(2), pp. 108112. pp. 400419.
[11] Yosibash, Z., Padan, R., Joskowicz, L., and Milgrom, C., 2007, A CT- [30] Demirdzic, I., and Martinovic, D., 1993, Finite Volume Method for Thermo-
Based High-Order Finite Element Analysis of the Human Proximal Femur Elasto-Plastic Stress Analysis, Comput. Methods Appl. Mech. Eng., 109, pp.
Compared To In-Vitro Experiments, ASME J. Biomech. Eng., 129(3), pp. 331349.
297309. [31] Demirdzic, I., Dzafarovic, E., and Ivankovic, A. I., 2005, Finite-Volume
[12] Taddei, F., Martelli, S., Reggiani, B., Cristofolini, L., andViceconti, M., 2006, Approach to Thermoviscoelasticity, Numer. Heat Transfer, Part B: Fundamen-
Finite-Element Modeling of Bones from CT Data: Sensitivity to Geometry and tals, 47(3), pp. 213237.
Material Uncertainties, IEEE Trans. Biomed. Eng., 53(11), pp. 21942200. [32] Bijelonja, I., Demirdzic, I., and Muzaferija, S., 2005, A Finite Volume Method
[13] Orwoll, E. S., Marshall, L. M., Nielson, C. M., Cummings, S. R., Lapidus, J., for Large Strain Analysis of Incompressible Hyperelastic Materials, Int. J.
Cauley, J. A., Ensrud, K., Lane, N., Hoffmann, P. R., Kopperdahl, D. L., and Numer. Methods Eng., 64(12), pp. 15941609.
Keaveny, T. M., 2009, Finite Element Analysis of the Proximal Femur [33] Cardiff, P., Karac, A., and Ivankovic, A., 2012, Development of a Finite Vol-
and Hip Fracture Risk in Older Men, J. Bone and Mineral Res., 24(3), pp. ume Contact Solver Based on the Penalty Method, Comput. Mater. Sci., 64,
475483. pp. 283284.
[14] Harris, M. D., Anderson, A. E., Henak, C. R., Ellis, B. J., Peters, C. L., and [34] Tropsa, V., Georgiou, I., Ivankovic, A., Kinloch, A. J., and Williams, J. G.,
Weiss, J. A., 2012, Finite Element Prediction of Cartilage Contact Stresses in 2006, OpenFOAM in Non-Linear Stress Analysis: Modelling of Adhesive
Normal Human Hips, J. Orthop. Res., 30, pp. 11331139. Joints, First OpenFOAM Workshop, Zagreb, Croatia.