Home Search Collections Journals About Contact us My IOPscience

Frequency dependence of complex moduli of brain tissue using a fractional Zener model

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

2005 Phys. Med. Biol. 50 2799

(http://iopscience.iop.org/0031-9155/50/12/005)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 109.202.229.65
The article was downloaded on 19/08/2010 at 20:21

Please note that terms and conditions apply.

INSTITUTE OF PHYSICS PUBLISHING PHYSICS IN MEDICINE AND BIOLOGY

Phys. Med. Biol. 50 (2005) 2799–2805 doi:10.1088/0031-9155/50/12/005

Frequency dependence of complex moduli of brain

tissue using a fractional Zener model
M Kohandel1, S Sivaloganathan1, G Tenti1 and K Darvish2
1 Department of Applied Mathematics, University of Waterloo, Waterloo,

Ontario N2L 3G1, Canada

2 Center for Applied Biomechanics, University of Virginia, Charlottesville, VA, USA

E-mail: kohandel@uwaterloo.ca

Received 17 February 2005, in final form 4 April 2005

Published 25 May 2005
Online at stacks.iop.org/PMB/50/2799

Abstract
Brain tissue exhibits viscoelastic behaviour. If loading times are substantially
short, static tests are not sufficient to determine the complete viscoelastic
behaviour of the material, and dynamic test methods are more appropriate. The
concept of complex modulus of elasticity is a powerful tool for characterizing
the frequency domain behaviour of viscoelastic materials. On the other hand,
it is well known that classical viscoelastic models can be generalized by means
of fractional calculus to describe more complex viscoelastic behaviour of
materials. In this paper, the fractional Zener model is investigated in order
to describe the dynamic behaviour of brain tissue. The model is fitted to
experimental data of oscillatory shear tests of bovine brain tissue to verify its
behaviour and to obtain the material parameters.

1. Introduction

Mechanical properties of very soft tissues, such as brain, liver, kidney, etc, have been the
subject of many recent studies and have been used in a variety of different applications in
medicine such as hydrocephalus, robotic surgery, surgical operation planning, surgeon training
systems and traumatic brain injury (Fung 1993, Brett et al 1995, Burdea 1996). Brain tissue,
like most other soft tissues, exhibits viscoelastic behaviour. Due to numerous factors involved
in the deformation of the brain in various circumstances, there is a wide range of models in
the literature.
To obtain complete information about the viscoelastic behaviour of brain tissue, it is
necessary to have experimental data over a wide range of time scales. Relaxation and creep
experiments provide information on low rates of loading. However, in certain cases the
response of the brain to a loading for very short time or high rates of loading may be of
particular importance. For example, in a typical impact scenario, such as an automotive crash,
the brain is subjected to deformation impulses with a duration of only a few milliseconds
0031-9155/05/122799+07$30.00 © 2005 IOP Publishing Ltd Printed in the UK 2799
2800 M Kohandel et al

(Meaney et al 1996). Therefore, to investigate the response of the brain for very short times
of loading and characterize its viscoelastic behaviour, it is more appropriate to use dynamic
test methods.
The complex modulus concept is a powerful and widely used tool for characterizing the
viscoelastic behaviour of materials in the frequency domain. According to this concept, all
elastic moduli, as well as the Poisson ratio of the material, whether isotropic or anisotropic, can
be regarded as complex quantities. A knowledge of complex moduli, and more particularly
their components, the dynamic (storage) modulus, the loss modulus and the loss factor, and
their frequency dependence have been of interest to scientists and engineers for a long time.
The dynamic properties can be determined by measurements. In the case of brain tissue,
there are several experimental methods to obtain the dynamic moduli, see later sections. The
measured frequency dependence of the complex moduli of soft tissue materials is usually
interpreted by means of linear or nonlinear viscoelastic models. However, as we discussed in
a recent paper (Davis et al 2005), while these viscoelastic models seem to match well with
the experimental data, they involve an excessive number of material parameters. It is quite
well known that the classical viscoelastic models can be generalized to fractional models in
order to better describe the viscoelastic behaviour of materials (Bagley and Torvik 1983a).
By introducing fractional-order operators instead of integer-order operators in the constitutive
relations, the number of parameters required to accurately describe the experimental data can
be significantly reduced (Bagley and Torvik 1983a, Davis et al 2005). Fractional models have
successfully been fitted to experimental data within a broad frequency range for polymers
using only four parameters (Bagley and Torvik 1983a, Pritz 1998, 1999). Successful fits of the
models to time domain data have also been reported (see Davis et al (2005) and the references
therein). Fractional model (Kelvin–Voigt fractional model) has been recently used to study
the dynamic mechanical properties of agarose gel (Chen et al 2003, 2004) and liver tissues
(Taylor et al 2002, Kiss et al 2004).
In this paper, we use the fractional Zener model, which has only four parameters, to
describe the frequency dependence of brain tissue. The model behaviour will be investigated,
in the frequency domain, and compared to available experimental data of brain tissue for
complex moduli to obtain the mechanical parameters of brain tissue.

2. Theory

2.1. Complex moduli

The most general relation between stress and strain in linear viscoelasticity theory is given in
terms of the Boltzmann superposition principle (Tschoegl 1989),

t
σ (t) = M(t − τ )(τ ) dτ, (1)
−∞

where σ (t) and (t) are the stress and strain, respectively, and M(t) = σm (t)/0 is the memory
function, in which σm (t) is the stress response of the material to a strain impulse excitation of
0 magnitude. The above equation can also be written in terms of the relaxation function G(t)
using M(t) = dG(r)/dt + G(0)δ(0) (Tschoegl 1989).
Equation (1) defines the stress–strain relation in the time domain. One can determine
the general relation in the frequency domain by taking the Fourier transform of the equation.
Provided that the Fourier transform for σ (t), (t) and M(t) exist, one obtain
σ̃ (ω) = M̃(ω)˜ (ω), (2)
Frequency dependence of complex moduli of brain tissue 2801

where σ̃ (ω) and ˜ (ω) are the Fourier transforms of the stress and strain, respectively. This
equation represents the constitutive law in the frequency domain, which is simply Hooke’s
law generalized for dynamic loading. Therefore, the linear dynamic properties of materials in
the frequency domain can be characterized by the complex number M̃(ω), which is referred to
as the complex modulus of elasticity. The complex modulus is usually defined for harmonic
vibrations, but the more general definition is given (provided that the Fourier transform of the
stress and strain time histories exist) by equation (2), i.e.,
σ̃ (ω)
M̃(ω) = = Md (ω) + iMl (ω), (3)
˜ (ω)
where Md is the dynamic modulus and Ml is the loss modulus. The parameter η = tan φ =
Ml /Md is the so-called loss factor (φ is the phase angle).
In the definition of the complex modulus given by equation (3), no restrictions have been
placed on the type of deformation. Therefore, M̃(ω) may represent the complex form of
any modulus of elasticity, whether isotropic or anisotropic. However, in this paper only the
complex shear modulus is considered in order to compare with experimental data. One should
note that with the knowledge of the complex modulus the relaxation and creep functions can
be calculated and vice versa (Pritz 1999).
Any physically realizable system satisfies the principle of causality, in which no response
is expected from the system prior to the application of an excitation. One of the main
consequences of the principle of causality is that the real and imaginary parts of the frequency
response function of the real physical linear system are interrelated (Champeney 1973). The
relations are known as the Kronig–Kramers dispersion relations (or dispersion relations in
short) because the first such relations were given by Kronig (1926) and Kramers (1927) in
studies concerned with the theory of electromagnetic radiation. Mathematically, the relation
between the real and imaginary parts of the complex modulus can be derived from equation (1)
by taking into account the causality of the memory function M(t). Several forms of such
relations have been formulated for viscoelastic materials, a simplified form is given by
(Tschoegl 1989)

∞
2ω2 Ml (x)/x
Md (ω) = M0 + P dx
π 0 ω2 − x 2

∞ (4)
2ω Md (x)
Ml (ω) = − P dx,
π 0 ω2 − x 2
where P stands for the principal value of the integral and M0 = M̃(0) = Md (0) is the static
modulus of elasticity.

2.2. Fractional Zener model

Fractional derivative models have proved to be efficient in describing the dynamic behaviour
of materials, especially in the case of polymers (Pritz 1998, 1999, 2003) and soft tissues
(Taylor et al 2002, Chen et al 2003, 2004, Kiss et al 2004). Essentially, the generalization of
conventional viscoelastic models is obtained by replacing the viscous element with a fractional
element. In this paper we consider the Zener model, which consists of an elastic and viscous
element in series, all in parallel with another elastic element. Thus, the fractional Zener model
is obtained by replacing integer-order time derivatives of stress and strain in the constitutive
equation with fractional-order derivatives. The constitutive equation for the fractional Zener
model, or four parameter model as it is also known, is
σ (t) + τ α D α σ (t) = M∞ (t) + M0 τ α D α (t), (5)
2802 M Kohandel et al

where σ (t) and (t) are stress and strain, respectively, τ is the relaxation time τ = µ/M1 , µ is
the dashpot viscosity, M1 and M0 are the spring constants, and furthermore M∞ = M0 + M1
is the high-frequency limit of the dynamic modulus, which is equal to the initial value of the
relaxation modulus (Tschoegl 1989). The fractional derivative, say the αth-order derivative
of σ (t), is defined as (Podlubny 1999)


1 d t σ (τ )
D σ (t) =
α
dτ, (6)

(1 − α) dt 0 (t − τ )α
where 0
α
1. This equation yields the constitutive equation of the conventional Zener
model if α = 1.
The complex modulus of the model can be derived by transforming equation (5) into the
frequency domain. The derivative is easily transformed since (Bagley and Torvik 1983b)
F [D α σ (t)] = (iω)α F [σ (t)], (7)
where F denotes the Fourier transform. Hence, one obtain
1 + d(iωτ )α
M̃(ω) = M0 , (8)
1 + (iωτ )α
where d = M∞ /M0 . The dynamic modulus, loss modulus and loss factor are then given by
1 + (d + 1) cos(απ/2)ωnα + dωn2α
Md (ω) = M0 ,
1 + 2 cos(απ/2)ωnα + ωn2α
(d − 1) sin(απ/2)ωnα
Ml (ω) = M0 , (9)
1 + 2 cos(απ/2)ωnα + ωn2α
(d − 1) sin(απ/2)ωnα
η(ω) = .
1 + (d + 1) cos(απ/2)ωnα + dωn2α
Here ωn = ωτ is the normalized frequency. These equations will be used in conjunction with
the complex modulus experiments in order to determine the parameters of the model.

3. Results and discussion

In order to test the ability of the model to represent the dynamic behaviour of brain tissue,
experimental data are required. There are several experimental methods to study the dynamic
moduli of brain tissue such as the free vibration method, resonance method and forced
vibration method. One of the first experiments on dynamic behaviour of brain tissue was
performed by Fallenstein et al (1969) by studying the resonance of human brain samples in
simple shear. Galford and McElhaney performed free vibration tests on human brain samples
in compression (Galford and McElhaney 1970). Shuck and Advani (1972) studied forced
vibration of cylindrical samples of human brain in torsion. Wang and Weinmann (1972)
also performed in vivo forced vibration tests on a Rhesus monkey. A recent experiment was
performed by Arbogast and Margulies using oscillatory shear deformation on porcine brain
samples to study the anisotropy of the brain stem (Arbogast and Margulies 1998). Thibault
and Margulies (1998) studied the age-dependent material properties of the porcine cerebrum.
Bilston et al (1997) also studied the linear viscoelastic properties of bovine brain tissue in
shear. In a recent paper (Bilston et al 2001), the authors reported on experimental results for
large strain behaviour of brain tissue in shear and applied a differential constitutive equation
to fit the experimental data.
More recently, Darvish and Crandall studied the nonlinear viscoelastic effects in
oscillatory shear deformation (Darvish and Crandall 2001). They used small disc-shaped
Frequency dependence of complex moduli of brain tissue 2803

Table 1. Material constants for conventional and fractional models.

M∞ (kPa) M0 (kPa) τ (s) α
Conventional model 1.51 0.50 0.006 1
Fractional model 3.38 0.36 0.001 0.6

samples of bovine brain tissue in simple finite amplitude shear forced vibration (0.5–200 Hz).
The response of the samples to simple, double and triple harmonic inputs was determined in
order to characterize the nonlinearities up to the third order. By changing the input frequencies
independently, the linear and the third-order complex shear moduli were determined. Two
nonlinear viscoelastic models were considered to describe the dynamic viscoelastic behaviour
of brain tissue: a quasi-linear viscoelastic (QLV) model and a fully nonlinear viscoelastic
model with multiple hereditary integrals (MHI). The MHI model is nonlinear with respect
to strain and time, however, it requires too many material parameters, and as a result more
experimental data are needed. On the other hand, the QLV model is nonlinear only with
respect to strain, and it requires fewer material parameters. However, it has been shown that
this model is not sufficient for describing the nonlinearity in the test results of forced vibration
of brain samples (Darvish and Crandall 2001). They used two methods to determine the linear
complex modulus: by applying simple harmonic inputs and white noise input. A discrete
spectrum approximation was used to fit the experimental results.
In this section, we use the experimental results of Darvish and Crandall (2001) to test
the ability of the analytical results (obtained from the fractional Zener model) in providing
an accurate fit. In their experiment, resonance of the sample holders occurred at above
60 Hz that caused a ripple in the magnitude of the complex moduli close to that frequency
and disturbed the phase data above 30 Hz. Therefore, only the experimental data at low
frequencies (below 100 Hz for magnitude data and below 30 Hz for phase data) can be used
in the fitting processes. The four parameters of the model (see previous section) are identified
by fitting to the experimental data, using the NonlinearFit command available in Mathematica
(Wolfram Research 1996). The list of material constants for brain tissue is given in table 1.
In figure 1, we have shown the results for the magnitude of M̃(ω) and the phase
angle, i.e.,

|M̃(ω)| = Md (ω)2 + Ml (ω)2 , φ = tan−1 η. (10)
To compare, we have also plotted the results for the conventional Zener model (α = 1) as well
as the discrete spectrum model of Darvish and Crandall (2001). As can be seen from figure 1,
there is good agreement between the experimental data and analytical predictions.
In order to quantitatively compare the goodness of fits in the fractional Zener model and
the discrete model of Darvish and Crandall (2001), we have presented the results for mean
square error (MSE) in magnitudes and phases for two experimental samples in table 2. For the
magnitude, although the discrete model gives a better result, the fractional model also predicts
a reasonable fit. On the other hand, the fractional model predicts much better results for the
phase angle. One should also note that the number of parameters is significantly different in
these models. The fractional Zener model has four parameters compared to the discrete model
that has seven fitted parameters.
In summary, a fractional Zener model has been developed from the non-fractional
counterpart, and a method of calculating the complex modulus has been presented. From
the figure, we see that the fractional derivative gives rise to a model that provides a good fit to
experimental data for the complex moduli of brain tissue. The fractional model predictions fit
reasonably well with experimental data and, further, permit a characterization of the material
2804 M Kohandel et al

10
exp. sample2

Magnitude [KPa]
5 exp. sample3
exp. white noise
fractional model
conventional model
2 discrete model

0.5

1 2 5 10 20 50 100 f [Hz]

70
Phase [degree]

60

50

40

30

20

10

0
1 2 5 10 20 50
f [Hz]

Figure 1. Comparison of experimental data (two samples are shown as circles, and white noise
input is shown as joined points) with theoretical results of fractional Zener model (solid line),
conventional Zener model (dashed line) and discrete spectrum model (dotted line) of Darvish and
Crandall (2001).

Table 2. Goodness of fits in magnitudes and phases for fractional Zener model and discrete
spectrum model.

Experiment Model MSEmag (kPa2 ) MSEphase (degree2 )

Sample 2 Fractional 2.77 14.84
Discrete 0.42 324.30

Sample 3 Fractional 0.77 43.42

Discrete 0.10 300.45

using fewer parameters than those based on classical differential calculus. However, to reach
a definitive conclusion more experimental data will be necessary. A potentially fruitful area
of investigation is to study the nonlinear effects in the constitutive relation for brain tissue
(Darvish and Crandall 2001) using a fractional Zener model.

Acknowledgments

The financial support of the National Sciences and Engineering Research Council of Canada
and the Hospital for Sick Children Foundation is gratefully acknowledged.
Frequency dependence of complex moduli of brain tissue 2805

References

Arbogast K B and Margulies S S 1998 Material characterization of brain-stem from oscillatory shear tests J. Biomech.
31 801–7
Bagley R L and Torvik P J 1983a A theoretical basis for the application of fractional calculus to viscoelasticity
J. Rheol. 27 201–10
Bagley R L and Torvik P J 1983b Fractional calculus—A different approach to the analysis of viscoelastically damped
structures AIAA J. 21 741–8
Bilston L E, Liu Z and Phan-Thien N 1997 Linear viscoelastic properties of bovine brain tissue in shear Biorheology
34 377–85
Bilston L E, Liu Z and Phan-Thien N 2001 Large strain behaviour of brain tissue in shear: some experimental data
and differential constitutive model Biorheology 38 335–45
Brett P N, Fraser C A, Henningan M, Griffiths M V and Kamel Y 1995 Automatic surgical tools for penetrating
flexible tissues IEEE Eng. Med. Biol. 264–70
Burdea G 1996 Force and Touch Feedback for Virtual Reality (New York: Wiley)
Champeney D C 1973 Fourier Transforms and their Physical Applications (London: Academic)
Chen E J, Suki B and An K N 2003 Dynamical mechanical properties of agarose gel by a fractional derivative model
Summer Bioengineering Conference (Key Biscane, FL) ed L J Soslowsky
Chen E J, Suki B and An K N 2004 Dynamical mechanical properties of agarose gel modeled by a fractional derivative
model J. Biomech. Eng. 126 666–71
Darvish K K and Crandall J R 2001 Nonlinear viscoelastic effects in oscillatory shear deformation of brain tissue
Med. Eng. Phys. 23 633–45
Davis G B, Kohandel M, Sivaloganathan S and Tenti G 2005 Relaxation and creep functions in the fractional
viscoelastic Zener model Med. Eng. Phys. submitted
Fallenstein G T, Hulce V D and Melvin J W 1969 Dynamics mechanical properties of human brain tissue J. Biomech.
2 217–26
Fung Y C 1993 Biomechanics: Mechanical Properties of Living Tissues (New York: Springer)
Galford J E and McElhaney J H 1970 A viscoelastic study of scalp, brain and dura J. Biomech. 3 211–21
Kiss M Z, Varghese T and Hall T J 2004 Viscoelastic characterization of in vitro canine tissue Phys. Med. Biol. 49
4207–18
Kramers H A 1927 La diffusion de la lumiere par les atomes Atti del Congresso Internazionale dei Fisici, Como
vol 2 pp 545–57
Kronig R L 1926 On the theory of the dispersion of x-rays J. Opt. Soc. Am. 12 547–57
Meaney D F, Smith D H, Shreiber D I, Bain A C, Miller R T, Ross D T and Gennarelli T A 1996
Biomechanics analysis of experimental diffuse axonal injury Traumatic Brain Injury: Bioscience and Mechanics
ed F A Bandak et al (Larchmont, NY: Mary Ann Liebert Inc) pp 167–72
Nakada O 1960 Theory of nonlinear responses J. Phys. Soc. Japan 15 228
Podlubny I 1999 Fractional Differential Equations (New York: Academic)
Pritz K 1998 Frequency dependence of complex complex moduli and complex Poisson’s ratio of real solid materials
J. Sound Vib. 214 83–104
Pritz K 1999 Verification of local Kramers–Kronig relations for complex modulus by means of fractional derivative
model J. Sound Vib. 228 1145–65
Pritz K 2003 Five-parameter fractional derivative model for polymeric damping materials J. Sound Vib. 265 935–52
Shuck L Z and Advani S H 1972 Rheological response of human brain tissue in shear ASME J. Basic Eng. 94 905–11
Taylor L S, Lerner A L, Rubens D J and Parker K J 2002 A Kelvin–Voigt fractional derivative model for viscoelastic
characterization of liver tissue ASME Int. Mechanical Engineering Congress and Exposition (New Orleans, LA)
ed E P Scott
Thibault K L and Margulies S S 1998 Age-dependent material properties of the porcine cerebrum: effect on pediatric
inertial head injury criteria J. Biomech. 31 1119–26
Tschoegl N W 1989 The Phenomenological Theory of Linear Viscoelastic Behaviour: An Introduction (Berlin:
Springer)
Wang C H and Wineman A S 1972 A mathematical model for the determination of viscoelastic behaviour of brain
in vivo: 1. Oscillatory response J. Biomech. 5 431–36
Wolfram Research 1996 Mathematica 4 Standard Packages Wolfram Media (Cambridge: Cambridge University
Press)