Journal of Biomechanics 35 (2002) 767–773

Determination of orthotropic bone elastic constants using FEA and

modal analysis
W.R. Taylora, E. Rolandb, H. Ploegc, D. Hertigc, R. Klabundec, M.D. Warnera,
M.C. Hobathod, L. Rakotomananab, S.E. Clifta,*
a
Department of Mechanical Engineering, University of Bath, Claverton Down, Bath BA2 7AY, UK
b
Laboratoire de Genie Medical, Ecole Polytechnique Federale, Lausanne, Switzerland
c
Sulzer Orthopedics Ltd, Winterthur, Switzerland
d
UTC, Laboratoire de Biom"ecanique et G"enie Biom"edical, CNRS 6600, Compi"egne, France
Accepted 5 February 2002

Abstract

Finite element models have been widely employed in an effort to quantify the stress and strain distribution around implanted
prostheses and to explore the influence of these distributions on their long-term stability. In order to provide meaningful predictions,
such models must contain an appropriate reflection of mechanical properties. Detailed geometrical and density information is now
readily available from CT scanning. However, despite the use of phantoms, a method of determining mechanical properties (or
elastic constants) from bone density has yet to be made available in a usable form.
In this study, a cadaveric bone was CT scanned and its natural frequencies were measured using modal analysis. Using the
geometry obtained from the CT scan data, a finite element mesh was created with the distribution of density established by matching
the mass of the FE bone model with the mass of the cadaveric bone. The maximum values of the orthotropic elastic constants were
then established by matching the predictions from FE modal analyses to the experimental natural frequencies, giving a maximum
error of 7.8% over 4 modes of vibration. Finally, the elastic constants of the bone derived from the analyses were compared with
those measured using ultrasound techniques. This produced a difference of o1% for both the maximum density and axial Young’s
Modulus. This study has thereby produced an orthotropic finite element model of a human femur. More importantly, however, is
the implication that it is possible to create a valid FE model by simply comparing the FE results with the measured resonant
frequency of the CT scanned bone. r 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Bone; Orthotropic; Finite element; Modal analysis

1. Introduction elastic constants incorporated into such FE models

Material constants of an appropriate accuracy are an
essential pre-requisite for any successful finite element
(FE) analysis. FE models of bone have been widely
employed to explore the long-term effects of the
implantation of hip prostheses. These explorations have
used FE approaches together with algorithms to predict
bone remodelling (e.g. Orr et al., 1990; van Rietbergen
et al., 1993; Smolinski and Rubash, 1992; Skinner et al.,
1994; Huiskes, 1995; Weinans and Sumner, 1997).
Determination of the geometry and distribution of
*Corresponding author. Tel.: +44-1225-323-870; fax: +44-1225-
826-928.
E-mail address: s.e.clift@bath.ac.uk (S.E. Clift).
therefore needs to be performed before the models can
be used with any confidence. This validation is
commonly achieved by comparing some aspect of
experimental model behaviour with corresponding FE
predictions: Strain gauge measurements (e.g. Little et al.,
1986; Dalstra et al., 1995) and modal analysis (e.g. Hight
et al., 1980; Khalil et al., 1981; Couteau et al., 1998a)
have both been used in this capacity.
Strain gauges are capable of measuring average values
of strain over the gauge length of the device and can
thus provide a basis for comparison with numerical
predictions, but only at particular locations on the bone
surface. In contrast to the local nature of the validation
in strain gauge studies, modal analysis offers the
possibility of a global validation for an entire structure.

0021-9290/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 1 - 9 2 9 0 ( 0 2 ) 0 0 0 2 2 - 2
768 W.R. Taylor et al. / Journal of Biomechanics 35 (2002) 767–773

The effect of a variation in the Young’s Modulus of the
cortical and cancellous bone of a human tibia has been
explored by Hobatho et al. (1991) using an isotropic
materials model. Using CT scan supplied geometry,
values of Young’s modulus for the two different types of
bone were varied within the range of magnitudes from
the literature. It was found that very good agreement
between FE predictions and experimental values of
natural frequency could be obtained in this way.
Unfortunately, the distribution of density for the actual
bone tested was not incorporated into the model.
Couteau et al. (1998b) compared experiments and
numerical results for the first 4 natural frequencies of
a human femur. Errors of up to 8.1% were reported.
From the literature, it is well known that modal
analysis is capable of validating existing FE models of
structures with complex materials distributions (Cou-
teau et al., 1998a, b) but these techniques have never
been used to derive the properties of bone. The material
constants of bone are widely recognised as being better
modelled as orthotropic rather than isotropic (Rho,
1996). The aim of this study therefore was to both
validate our orthotropic FE model and determine the
quality of this validation using the above techniques. In
addition, this study aimed to establish whether modal
analysis is capable of determining the distribution of
material constants within bone structures.
(STAC SPMulti-8, V 2.1) and modal analysis software
(STAR 5.2) to give the natural frequencies of the bone
and the corresponding mode shapes. A frequency range
from 0 to 1000 Hz was investigated and five measure-
ments were averaged for each degree of freedom (DOF).
The modes of vibration were determined by identifying
the global peak values in the frequency domain.
Undamped frequencies of the system were calculated
using
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Fdamped ¼ Fundamped ð1  e2 Þ; ð4Þ
where e is the damping.
The bone was CT scanned (Phillips Tomoscan AV
120 kV scanner) with slice intervals of 1 mm in the
epiphysis and 5 mm in the diaphysis, using a 512  512
voxel slice density. An FE model of the bone was then
created from the CT scan data using MSC/PATRAN
(Version 8). The complete mesh consisted of 99732
nodes and 66442 quadratic tetrahedral elements and is
shown in Fig. 1.
The measured Hounsfield value (HU) was determined
for every element in the FE model by relating its co-
ordinate position to the CT scan data. The full range of
measured HU values were then divided into 42 different
material groups, based on intervals of 50 HU, ranging
from 50 HU to a maximum of 2050 HU (mid range
values: 25, 25, 75, up to 2025). Each element was then

2. Methodology

A fresh, frozen, human cadaveric femur was prepared

by removing all external soft tissues and then weighed.
The bone was mounted horizontally by its metaphyses
using soft elastic straps (resonant frequency o10 Hz,
mass 2 g) in order to simulate free–free boundary
conditions during vibration. A unidirectional piezo-
.
electric accelerometer (Bruel&Kjaer 4375, weight 2.3 g)
was fixed to the surface of the bone using beeswax. The
structure was then excited by hitting the femoral head
normal to its surface with an impact hammer containing
.
a force transducer (Bruel&Kjaer 8200, weighing 100 g)
at an angle so as to excite as many modes of vibration as
possible. Measurements were taken at 16 points evenly
distributed along the medial and lateral surfaces of the
bone, with all the required measurements at each
location recorded together before moving onto the next
location so as to leave as little beeswax residue as
possible. In order to measure the 3D acceleration of the
bone, the accelerometer was fixed successively to each of
the 16 points and in the orthogonal orientations, aligned
to a co-ordinate system defined by the shaft of the bone
Fig. 1. Tetrahedral finite element mesh generated from bone CT scan
and the epicondular axis. The impact force signal and
data. The four slices show the internal mesh distribution through the
the corresponding acceleration signal were conditioned relative indicated transverse sections. The mesh in sections A and B has
.
using charge preamplifiers (Bruel&Kjaer 2651). The an irregular surface as some element boundaries were out of plane
signal was characterised using a spectrum analyser across this section.
 W.R. Taylor et al. / Journal of Biomechanics 35 (2002) 767–773 769

placed into its relevant material group, but also assigned

Max.
one of 16 different radial orientations (8 in the shaft and
8 in the neck). The orientations of the material proper- Density
ties in shaft were aligned with the axial direction of each / kg m-3
CT scanned slice (assumed consistent with the long axis
of the bone). The centre of each CT slice section was 1000
then determined visually and from this centre point, the
slice was divided into 22.51 sectors (16 sectors in total, 2
each of 8 different radial orientations). The orientation
of each sector was assigned to those elements falling 0
within its boundaries. A similar procedure was followed 0 2050
in the neck of the bone where the long axis was defined
perpendicular to the neck resection. An initial estimate Hounsfield Units (HU)
of both density and elastic material properties was then Apparent Density Effective Density
assigned to each of these material groups before the Fig. 2. Schematic diagram showing the relationship between effective/
actual values could be determined by matching to the apparent density and Hounsfield number.
experimental results of the bone.
The material groups were firstly assigned a linear
density relationship from Rho et al. (1995) the region of most dense cortical bone and therefore be
3 equal to the maximum value of apparent density. The
reff ¼ A HU þ 1000 ðkg=m Þ; ð1Þ
second of these values was taken from the relationship
where reff was the effective density in kg/m3 and between Hounsfield units and effective density which
A ¼ 0:523 kg/m3 associates water at 0 HU with an effective density of
The total mass of the FE model was then calculated 1000 kg/m3. However, water has an apparent density of
using this literature value of A: This predicted a mass in zero and therefore on this scale 0 HU must be associated
excess of the actual bone mass. The calculation with 0 kg/m3. This is shown graphically in Fig. 2. As a
procedure was therefore performed again with an result, the properties of the bone elements in the model
arbitrarily lower value of A (0.4 kg/m3). Given these were able to be related to the apparent density, in a
two fixed values, an interpolative approach was then relationship that excludes fluid.
taken to determine the value of A needed to ensure that The final stage in the FE model construction was to
the mass of the FE bone model matched that of the calculate the elastic constants for the various materials
actual bone. This established the maximum value of groups from the apparent density distribution. A
bone density (reff;max ) and thus, together with the CT quadratic distribution of the elastic constants from the
distribution, the effective density distribution for the lowest to the highest the density values, as used by
initial FE model, which included marrow, was pro- Rubin et al. (1993), was employed as expressed in the
duced. It was important that this mass comparison following equations:
process occurred before any vibrational effects were
considered, since it was essential to have the mass of the r2app; i r2app; i
structure correct before the FE modal analysis could be E1;i ¼ E1;max ; E2;i ¼ E2; max ;
r2app; max r2app; max
performed.
The next stage in the FE model definition process was r2app; i r2app; i
E3;i ¼ E3; max ; G12;i ¼ G12; max ;
to calculate the material elastic constants from the r2app; max r2app; max
density. However, effective density includes mass of the r2app; i r2app; i
marrow, essentially fluid, which contributes to the G13;i ¼ G13; max ; G23;i ¼ G23; max ; ð2Þ
overall mass but not to the load carrying capability of r2app; max r2app; max
the structure; any mechanical properties based on
density must therefore be able to exclude the contribu- where E is the Young’s modulus in MPa, G the Shear
tion of any fluid. Hence, it was necessary to extract the modulus in MPa, rapp the apparent density in kg/m3, i
density of the bone alone from density of the combined the material index, max is maximum material number,
bone and marrow. This measure of density with no fluid and rapp;max maximum apparent density of the bone.
contribution is termed the apparent density. As with The process of varying the FE elastic constants in
effective density, the relationship between apparent order to match the numerical predictions with the
density and Hounsfield number was assumed linear. experimental values required the ratios of the elastic
Thus, two values were needed to determine its beha- constants to be supplied. Orthotropic elastic constants
viour. The first of these came from the observation that for the regions of highest density were taken from
the maximum value of effective density must relate to Couteau et al. (1998a), and are shown in Table 1.
770 W.R. Taylor et al. / Journal of Biomechanics 35 (2002) 767–773

Table 1
Initial values of elastic constants from Couteau et al. (1998a)

Initial elastic constants (GPa)

E1 E2 E3 G12 G13 G23

11.6 12.2 19.9 4.0 5.0 5.4

n12 ¼ 0:42; n23 ¼ n13 ¼ 0:23; where co-ordinate system is cylindrical

with directions (1) radial, (2) tangential and (3) axial.

Modal analysis was then performed on the FE model

using the ABAQUS FE software in order to predict the
natural frequencies of the bone for this initial set of
material properties. Free–free boundary conditions were
assumed in the model. After the vibration frequencies
were predicted for this initial model, the mechanical
properties of the bone were then varied until the first
natural frequency of the FE model matched the
experimentally measured undamped frequency of the
bone. This process involved varying the six maximum
elastic orthotropic constants of the bone (E1;max ; E2;max ;
E3;max ; G12;max ; G13;max ; G23;max ), whilst holding the ratios
of these properties constant. At each step, the remaining
material groups were re-calculated in accordance with
the new maximum material constants, and the process
was repeated until the frequencies matched. The values
for Poisson’s ratio were held constant throughout. The Fig. 3. Variation of effective density in the surface region of the
quality of the agreement between experimental and cadaveric left femur (linear scale).
numerical mode shapes was then quantified using the
modal assurance criterion (MAC) methodology (Ewins,
2000) as implemented in the FEMtools software
package (Dynamic Design Solutions, 1999). Table 2
In order to assess the quality of the material proper- Values of elastic constants predicted from the FE model
ties predicted by the FE model, the full orthotropic Elastic constants (GPa)
material constants of the bone were measured using
E1 E2 E3 G12 G13 G23
transmission ultrasound techniques. Ultrasound mea-
surements (as described by Ashman et al., 1984) were 13.4 14.1 22.9 4.6 5.8 6.2
taken using 2.25 MHz transducers (longitudinal and Assumed: n12 ¼ 0:42; n23 ¼ n13 ¼ 0:23; where co-ordinate system is
transverse, Panametrics V106RB, V154RM), an oscillo- cylindrical with directions (1) radial, (2) tangential and (3) axial.
scope (Tektronix, 3032) and an impulse generator (HP
3312A). Calibration of the system was performed on
small specimens of different materials and compared were similar to Couteau et al. (1998b), whose study
with theoretical values (Gray, 1957). A total of 10 cubic showed 301.6 Hz for the resonant frequency (cf. 285 Hz
specimens (approximately 5 mm3) were then cut from in this study). The mass of the cadaveric bone was found
our bone, from locations shown to be the most dense to be 622 g, and therefore the corrected relationship for
according to the CT scan data. Each specimen was effective density based upon the mass of this particular
orientated relative to the long axis of the bone, each with bone was
a 451 angled edge. The mass and dimensions of the
reff ¼ 0:464 HU þ 1000 ðkg=m3 Þ: ð3Þ
specimens were recorded, and finally the densities,
velocities, and material constants of the bone specimens An FE bone mass of 622 g was predicted for a
were calculated. maximum effective density of 1940 kg/m3, when using
2025 as the maximum material average HU value. The
predicted variation in bone density is shown in Fig. 3.
3. Results Modal analysis was performed on both the FE model
and the cadaveric bone and showed the second
The natural frequencies of the intact bone were harmonic of the first mode shape at a higher frequency
measured and are shown in Table 3. These frequencies than the second harmonic of the second mode shape.
 W.R. Taylor et al. / Journal of Biomechanics 35 (2002) 767–773 771

Table 3
Comparison of experimental results with FE modal analysis predictions

Mode Experimental results FE predictions of % Error Quality of

number Frequency (Hz) at agreement
rmax ¼ 1940 kg/ between mode
m3 shapes (%)
Damped Damping Undamped
frequency (Hz) frequency (Hz)

1 285 0.027 285 285 0.0 95

2 317 0.031 317 324 2.2 96
3 521 0.121 524 489 6.7 82
4 777 0.088 780 710 9.0 84

This result was not consistent with the findings of

Couteau et al. (1998b).
After completing the FE modal analysis using the
initially estimated elastic constants, the FE material
constants were systematically varied in order to match
the first natural frequency of the FE bone with the
modal analysis results. This produced a maximum axial
Young’s modulus (E3max ) of 22.9 GPa, representing an
increase in stiffness of approximately 13% over the
initial condition. The complete set of the resulting
maximum elastic constants is shown in Table 2. FE
modal analysis using these bone properties produced
predictions of the natural frequencies of the bone as
Fig. 4. FE predicted and experimental mode shapes for the first 4
shown in Table 3. In addition, the agreement between modes of vibration. Light grey displays the undeformed shape. Mode
the experimental natural frequencies and FE predictions 1: Bending in sagittal plane. Mode 2: Bending in frontal plane. Mode 3:
were calculated, showing a maximum frequency error of Torsion. Mode 4: Second bending mode in frontal plane.
9% over the first 4 mode shapes. The harmonic FE
modal shapes are shown in Fig. 4. Quality of agreement Table 4
between the shapes of the modes of vibration between Comparison of material constants from ultrasound measurements and
FE predictions
the FE model and the cadaveric bone are shown in
Table 3. The results show a minimum MAC value of Component Ultrasound measurement FE predictions % Difference
82% at the fourth mode of vibration (second harmonic E1 (GPa) 17.9 13.4 25.1
of the second mode shape). This signifies a high degree E2 (GPa) 18.8 14.1 25.0
of correlation of the vibration displacement patterns E3 (GPa) 22.8 22.9 +0.4
along the entire length of the bone for all frequencies. G23 (GPa) 7.11 6.20 12.8
In the final test, transmission ultrasound techniques G31 (GPa) 6.58 5.80 11.9
G12 (GPa) 5.71 4.60 19.4
were used to find the actual properties of the bone. The n21 0.28 0.42 +50.0
results of the most dense region of bone, relating to that n31 0.38 0.23 39.5
found in the CT scan data, are displayed in Table 4. A n12 0.26 0.42 +61.5
very high degree of correlation was found for both the n32 0.37 0.23 37.8
n13 0.30 0.23 23.3
density and the axial Young’s modulus (E3max ), both
n23 0.31 0.23 25.8
with o1% error. Other variables had higher degrees of r (kg/m3) 1932 1940 +0.4
associated error, with a maximum error of –62% found
for n12 : The more significant stiffness constants (E1;max ;
E2;max ; E3;max ; G12;max ; G13;max ; G23;max ), however, showed
a maximum of 25% error. that modal analysis can be used, together with FE
models incorporating CT scan data, to determine the
distribution of elastic constants throughout a long bone.
4. Discussion This study has assumed that the FE results should be
matched to the fundamental (resonant) frequency of the
This study has performed an FE and experimental bone. Modal frequencies have an increased error with
modal analysis and compared the determined elastic larger accelerometer masses, and hence only a unidirec-
constants with ultrasound results. It has been shown tional accelerometer was used in order to keep the mass
772 W.R. Taylor et al. / Journal of Biomechanics 35 (2002) 767–773
of the measuring equipment to a minimum. This meant
that multiple measurements were required for each
position along the length of the bone. In addition, the
higher frequencies are more susceptible to a higher noise
to signal ratio. Both these considerations support the
idea that the FE results should be matched with the first
frequency of vibration.
One of the limitations of the FE model in the current
study was the number of sets of material properties used
to represent the distribution of both mass and stiffness
throughout the bone. This was due to the assignment of
orthotropic constants—a process that required a differ-
ent material type for each set of elastic constants and
each material orientation. A more advanced model
could usefully incorporate properties that are able to
vary continuously with HU number. Fourty-two mate-
rial sets were regarded as sufficient for the stiffer cortical
bone, but further studies should increase this number
since only a small number of material sets covered the
range of the less dense cancellous bone. The effect of a
more sophisticated relationship linking Hounsfield
number with apparent density could very usefully be
explored. The model employed here used a quadratic
relationship for both cortical and cancellous bone. This
may well be appropriate for cancellous bone, but is
probably an underestimate for cortical bone.
It was necessary to supply the model with the ratios of
the initial elastic constants, which, in this case, were
taken from Couteau et al. (1998b). Our own ratios,
which were only acquired subsequently during ultra-
sound testing (as this required destructive testing of the
bone), suggest that the actual ratios for this bone are
somewhat different. In comparison with the values
calculated by ultrasound, the predetermined ratios result
in an underestimation of the stiffness values in the two
other mutually perpendicular directions. The signifi-
cance of this has yet to be quantified, but since it does
not affect the largest stiffness value (E3;max ), is not
expected to compromise the usefulness of the results.
Further studies could incorporate minimisation techni-
ques in order to calculate the effect of variation in these
ratios.
Since all bone elements were modelled using the same
material definition (although with varying densities), it
was assumed that all bone obeys the same quadratic
relationship between stiffness and density and has the
same orthotropic ratios. Associated with this global
material definition, was the fact that the structural
properties were almost transversely isotropic through-
out. As the structure of the bone changes from cortical
to cancellous, however, these material properties would
be expected to cease to be a function of the orientation
of the material and become more a function of the
cancellous architecture. In reality, the high degree of
anisotropy in cancellous bone can cause the ratios
between the elastic moduli to be as high as 1:4 (Linde
et al., 1990) whereas ratios throughout this bone are
approximately 1:1.7. Taking a standard set of initial
properties is a limitation in this study, but this effect
might be reduced if cancellous and cortical properties
were to be defined separately.
A major implication of this study is that it may offer
the possibility of determination of a complete set of
orthotropic elastic constants from a single modal
measurement, together with standard CT scanning data
and the FE approach detailed in this paper. Although
this necessitates prior assumptions regarding the ratios
of the elastic constants, which may vary from bone to
bone, current understanding and findings suggest that
this simple technique could yield useful results. The
development of this process would obviously require
further experimentation, but we fully expect this to be
feasible and the set-up is already in place for further
exploration.
By comparing FE predictions of fundamental fre-
quency with modal analysis results, this study has
demonstrated a viable technique for both validating
FE bone models and establishing the distribution of
orthotropic elastic constants throughout the bone.
Comparison of mode shapes between the resulting FE
model and the cadaveric bone displays a high level
agreement and therefore demonstrates the quality of the
results obtainable. The excellent agreement between the
FE predicted density and major stiffness component and
those determined for this bone using ultrasonic techni-
ques show this approach to be viable for the determina-
tion of the distribution of elastic constants.
Acknowledgements
Useful discussions with Dr. H. Bereiter (Raetisches
Kantons- und Regionalspital, Abteilung Orthopaedie,
Chur, Switzerland) and Dr. H. Jacob (Universitaetskli-
nik Balgrist, Zurich, Switzerland) are gratefully ac-
knowledged. Thanks are also due to Dr. Y. Deger of
Sulzer Innotec AG, Winterthur, Switzerland, who
determined the Modal Assurance Criterion values. The
funding for this study was provided by Sulzer Orthope-
dics AG, Winterthur, Switzerland.
References
Ashman, R.B., Cowin, S.C., Van Buskirk, W.C., Rice, J.C., 1984. A
continuous wave technique for the measurement of the elastic
properties of cortical bone. Journal of Biomechanics 17 (5), 349–
361.
Couteau, B., Labey, L., Hobatho, M.C., Vander Sloten, J., Arlaud,
J.Y., Brignola, J.C., 1998a. Validation of a three dimensional finite
element model of a femur with a customised hip implant In:
Middleton, J., Jones, M.L., Pande, G.N. (Eds.), Computer
Methods in Biomechanics and Biomedical Engineering, Vol. 2,
 W.R. Taylor et al. / Journal of Biomechanics 35 (2002) 767–773
Gordon and Breach Science Publishers, Amsterdam, pp. 147–154.
ISBN 90-5699-206-6.
Couteau, B., Hobatho, M.C., Darmana, R., Brignola, J.-C., Arlaud,
J.-Y., 1998b. Finite element modelling of the vibrational behaviour
of the human femur using CT-based individualised geometrical and
material properties. Journal of Biomechanics 31, 383–386.
Dalstra, V., Huiskes, R., van Erning, L., 1995. Development and
validation of a 3D finite element model of the pelvic bone. Journal
of Biomechanical Engineering 177, 272–278.
Dynamic Design Solutions n.v. (DDS), 1999. Interleuvanlaan 64, 3001
Leuven, Belgium.
Ewins, D.J., 2000. Modal Testing: Theory, Practice and Application.
(2nd ed.) Research Studies Press, Baldock, UK.
Gray, D.E. (Coordinating Ed.), 1957. American Institute of Physics
Handbook, McGraw-Hill, New York.
Hight, T.K., Piziali, R.J., Nagel, D.A., 1980. Natural frequency
analysis of a human femur. Journal of Biomechanics 13, 139–147.
Hobatho, M.C., Darmana, R., Pastor, P., Barrau, J.J., Laroze, S.,
Morucci, J.P., 1991. Development of a three dimensional model of
a human tibia using experimental modal analysis. Journal of
Biomechanics 24, 371–383.
Huiskes, R., 1995. Bone remodelling around implants can be explained
as an effect of mechanical adaptation. In: Galante, J.O., et al. (Ed.),
Total Hip Revision Surgery. Raven Press, New York, pp. 159–171.
Khalil, T.B., Viano, D.C., Taber, L.A., 1981. Vibrational character-
istics of the embalmed human femur. Journal of Sound Vibration
75 (2), 417–436.
Linde, F., Pongsoipetch, B., Frich, L.H., Hvid, H., 1990. Three-axial
strain controlled testing applied to bone specimens from the
proximal tibial epiphysis. Journal of Biomechanics 23, 1167–1172.
773
Little, R.B., Wevers, H.W., Siu, D., Cooke, T.V.D., 1986. A three-
dimensional analysis of the upper tibia. Journal of Biomechanical
Engineering 108, 111–119.
Orr, T.E., Beaupre, G.S., Carter, D.R., Schurman, D.J., 1990.
Computer predictions of bone remodelling around porous-coated
implants. Journal of Arthroplasty 5, 191–200.
Rho, J.Y., 1996. An ultrasonic method for measuring the elastic
properties of human tibial cortical and cancellous bone. Ultra-
sonics 34, 777–783.
Rho, J.Y., Hobatho, M.C., Ashman, R.B., 1995. Relations of
mechanical properties to density and CT numbers in human bone.
Medical Engineering Physics 17, 347–355.
Rubin, P.J., Rakotomanana, R.L., Leyvraz, P.F., Zysset, P.K.,
Curnier, A., Heegaard, J.H., 1993. Frictional interface micromo-
tions and anisotropic stress distribution in a femoral total hip
component. Journal of Biomechanics 26, 725–739.
Skinner, H.B., Kilgus, D.J., Keyak, J., Shimaoka, E.E., Kim, A.S.,
Tipton, J.S., 1994. Correlation of computed finite element stresses
to bone density after remodelling around cementless femoral
implants. Clinical Orthopaedics and Related Research 305, 178–
189.
Smolinski, P., Rubash, H.E., 1992. Bone remodelling around hip
implants. Critical Reviews in Biomedical Engineering 20, 461–483.
Van Rietbergen, B., Huiskes, R., Weinans, H., Sumner, D.R., Turner,
T.M., Galante, J.O., 1993. The mechanism of bone remodelling
and resorption around press-fir THA stems. Journal of Biomecha-
nics 26, 369–382.
Weinans, H., Sumner, D.R., 1997. Finite element analysis to study
periprosthetic bone adaptation, In: Lowet, G. et al. (Eds.), Bone
Research in Biomechanics. IOS Press, Amsterdam, pp. 3–16.